Optical multichannel analyzer as a scanner for the ultracentrifuge

ARCHIVES

OF

Optical

BIOCHEMISTRY

AND

Multichannel

BIOPHYSICS

Analyzer 1. SINGLE

864-888

168,

as a Scanner BEAM

E. G. RICHARDS Department

of Biochemistry, The University Veterans Administration

(1973)

AND

of

for

the

OPERATION’ D. ROCKHOLT

Texas Health Science Center, and Hospital, Dallas, Texas 76216

Received

May

Ultracentrifuge

Biophysics

Section,

11, 1973

The use of a commercial optical multichannel analyzer as a scanner for the ultracentrifuge is described. A uv-sensitive silicon vidicon tube serves as the light detector. The 5 by 12.5 mm surface of the vidicon is divided into 500 channels, scanned about 30 times/set. An optical system was constructed which provided a reduced image of cell and counter-balance with the radius in the direction of channel number. The video signal from the vidicon is converted into digital data for each channel, the number of counts being proportional to the light intensity for that radial position. Storage registers are used to accumulate the counts for each channel, for any desired number of scans up to 9,999. For these studies, blue light from an H85C3 ac mercury lamp was used to illuminate the cell. Rotor and lamp pulses, visible on the real-time cathode-ray-tube monitor, were examined theoretically, and shown to have a negligible effect on the accumulated image. Only single-beam operation of the optical multichannel analyzer is described. The conversion of intensity data to absorbance as a function of radial position is described. The accuracy of the optical multichannel analyzer was verified by comparison of absorbances with values determined with a spectrophotometer. As a final test of the optical multichannel analyzer, the molecular weight of myoglobin was studied by sedimentation equilibrium. The reduced scatter in the experimental points permitted a more detailed analysis than usually performed. The final conclusion is that, even with single-beam operation at this early stage of development, the optical multichannel analyzer performs better than the double-beam, commercial scanner. Features already built into the optical multichannel analyzer lend themselves to automatic operation, either with a minicomputer or a specially constructed unit. Possible means of accomplishing double-beam and multiple cell operation are discussed.

The use of t,he absorption optical system in the ultracentrifuge has been generally limited to the study of dilute solutions of light-absorbing materials owing to its lesser accuracy when compared to the Rayleigh and schlieren optical systems. Both the ease of obtaining data and the accuracy of the results were improved by replacement of film as the detector with an automatic photo1 This investigation Grant HL14938 from Institute.

multiplier scanning system (1) and the use of double-beam electronics (2). I’urther improvement in electronic design led to the capability of running up to five cells simultaneously. Even with the commercial scanner and multiplexer equipment (3), any extensive study soon becomes bogged down in obtaining and processing the data from the many recorder tracts, which are still not as accurate as Rayleigh interferograms. With recent advances in electronic and computer technology, it has become possi-

was supported in part by the National Heart and Lung 864

C opyright All rights

0 1973 by Academic Press, of reproduction in any form

Inc. reserved.

Tl?LEVISION

hlc to digitize the optical density information at an appropriate place in the scanner ckcuit and either re:chord the information on magnetic tapo or procucss it, directly with a computer (4-6). A recent innovation involves digitizing thca light intcnsitJks obtained directly from the amplified phot,omultiplier output of the commclrcial B&man sc*anner and using a minkomput,t~r on lint: to convert. the intensity data to absorbnnccs (6). In this system, improvcbd atbcuravy is obtained by summing many pulses at cbac:hdrGrc;d radius. The syskm is able to handle data from Feveral dells simult,aneously and to process the data from stldimcntation equilibrium experiments on linct to givtg mol(~c*ular wcGght,s in a matkr of miriut,t3. A basic* probkrn in current scanner design (:onc(‘rns t.h(%us(~ of the photomultiplier as a light detector. Sinc:c: tha image is examined with the photomultiplkr placed behind a narrow slit, th(a cntircl imago can be seen only by mttchanically scanning the slit and tubcb assctmbly. Thus most, of the light, avail:lblr at th(> image planet is waskd. It has orcaurrod to many \vorkers that a t&vision (‘amr’r:l tube might 1~s usc>d as a scam1er d(ltwtrz, sinw it wr)uld pwmit t.ho r%xamination of t h(, entire imagca at. onto’. The design and c:onstruc*tion of suc*h :t system is a formidable task, beyond thcl c.npabilitics of most bioc~hcmkal rra~arc~h laborat,orirs. WC arc not, a\varc of any published work involving a sc*nnner clmploying a t&vision camera tube. Lkcc~ntly an inst,rument8, the> Optical IIlulticharmc~l Analyx(Ir (O1IA)2 that USPSa t&vision (*am(‘ra tubcb has becornci czommorcially availabl(~ (7, 8). It was designed to clxaminc> imngt5, auc*h as emission lint: spo(+trs, in \vhir*h thr light8 intcw3it.y is constant in on(’ dirc,c*tion but v-ark in th(b other. ‘1’11~ onu inc+lutks many fraturcls which mi& its us0 c!r3irabl(~ as a sc’ann(~r for the: ultrat*c~ntrifugc~ incaMing (a) that division of t hr tadial dirc*c+tirJll into 500 scpantto c*hanIlck, (b) ttlcl ( onv&ori of intcGty infor* The

name

O.1IA

is :I registered

trademark.

SCANNEll

865

mation into digital data, (c) the accumulation over selected timct pclriods of intensity information for ctac+ cahamirl into storage rcgistcq (d) digital t,o analog convertors whieh permit tho display on a suitable cathode ray tube monitor of the image from the c’amclra tub(b or t,hc information in the: storage rckgistcq and (0) input and output c*hannc!ls for programmcbd operation of data acAc.umulat,ion and transfer. ‘l’h(b purpr)sc of this paper is to assess the feasibility of using tho ON A as a sc:anncbr for th(% ultrawntrifuge. The modificat,ion and focusing of the optkal system arc dcsc:ribcd. optical density and sodimt%tat,ion quilihrium studios arc’ prctsont’ckd that. chow that this first version of a t,c+vision scttmif’r givc3 results that arc’ more ac:c:urate than any t.hat wo hav(L sc(~n published with the> wrnmt>rc*ial photomultiplicr scanner. l’c)ssibk improvcmcnts are discussed,including m&hods of multipk ~~91opclration.

The Beckman ,Ilodel 1‘; ultracentrifuge was cq~~ippetl with an early version of the absorption optical system, which used film to record the image and a 1185C3 mcrcllry arc lamp as the light source. The OM.4 Model 1205.4 console and Sfoticl 1205F {IV det,ector assm~bly (SSR Instrument,s Co., Santa Rfonica, CA) was r~sed as the scanner. Coupled to the O>IA wrre (a) a hIode 601 display monitor with a (i!$ in. cathode raytube (Tektronix, Inc., Beaverton, OR), (b) a modified Model 5055A digital recorder (Hewlett-Packard, Mountain View CR), and (c) :I Model i004H S-Y r~cortlel (Ilewl~~tt-l’a~k:lrtl). Cornpllt,er calculations were made through the use of a time-sharing system: The computer was a Cl)C 3200; ollr local terminal was a Teletype hlodrl MI<33 connected to a 701.4 telephone collpler (Omnit.er, Phoenix, AZ). On-lirrr plots were c~htainrtl wit.11 a 7’200A graphic plotter (Hewlctt-l’ack:crd 1 connected to the t elei ype. When relatively few culculat~ions involving logarithmic or trigononlrtric f\mctions were necessary, t,he TIP-35 pc)rkcl. (,:Jc111:1t or (T~eu,letl-Pnckurtl) w:js IlSPd.

Sperm whale ii~yogiobili was 0l)tained from Sigma, and horse heart myoglohin was purchased from I2lanr1. Human hemoglobin was l)rep:rrcd by Iysis of i-cd t~lood cells.

866

RICHARDS GENERAL

Modi$cation

AND

CONSIDERATIONS

of Optical System

The area scanned by the vidicon camera tube is only 12.5 by 10 mm, much smaller than the size of the image produced by the existing optical system. One has the choice of examining only part of the existing image, thereby achieving greater resolution, or modifying the optical system to produce a smaller image seen in its entirety. For the first evaluation of the OMA, we decided in favor of examining the entire image. To produce an image size of 12 mm from a span of 2 cm encompassing both reference holes, a magnification factor of about 0.6 is required. Several parameters were needed to calculate the required focal length and position of the camera lens from the equations presented by Richards et al. (9). The measured object distance u1 from the center of the cell to the center of the condensing lens was 8.25 cm. The focal length marked on the side of the lens used in our ultracentrifuge was 27% in. ; adding one-half the lens thickness gave 70.5 cm as the focal length fi required for first order optics. Assuming that this focal length was measured for the usual sodium yellow light at 596 nm, the focal length at any desired wavelength can be calculated from the dispersion data of fused quartz and the lens makers’ formula (10) for a plano-convex lens. For uv light at 260 nm, this gavefi = 69.25 cm. The final results from the calculations indicated that a focal length fi for the camera lens of about 20-2’2 cm would give an image of the desired size. With such a lens, however, the final image would be formed 30 cm above the support plate for the lens mount, a distance too small to be used conveniently with a 45’ mirror to direct the light horizontally. Thus, it was decided to dispense with the mirror and construct a vertical optical tube. The calculated values indicated that the camera lens would be positioned several centimeters above the plate, making it impossible to use a camera lens of shorter focal length.

ROCKHOLT

A scale drawing of the final optical system is shown at the top of Fig. 1. It should be noted that the camera lens lies to the left of the focal plane of the condensing lens, instead of to the right as in the conventional system. A comparison of calculated and measured parameters will be given later. At the bottom of Fig. 1 is a scale drawing of the optical tube together with the vidicon detector assembly. The optical tube was designed to permit vertical movement of the camera lens, a slight rotational adjustment of the tube on the support plate, any desired rotation of the vidicon housing, and the complete blocking of the light beam with a manual shutter. Adaptors with a variety of designs were constructed. One was used to mount a 35-mm camera for making photographs of the image. Others with greater heights permitted increasing the total optical distance, with corresponding greater magnification. For these first experiments with the OMA, it was decided to use the H85C3 mercury arc source already in our ultracentrifuge. The source housing was modified to incorporate a left-right adjustable slide with a 2.5-mm square hole. Interference filters of the desired wavelength were placed in a specially constructed holder attached to the housing with heat-insulating plastic. Preliminary trials with the 254 nm filter and a quartz camera lens demonstrated that our lamp emitted insufficient light of this wavelength for satisfactory operation of the OMA. Since an evaluation of the OMA does not depend upon the energy of the photons striking the vidicon, it was decided to confine our studies to visible light. We searched among three achromatic lenses (184, 193 and 220 mm focal lengths, Edmond Scientific Co., Barrington, NJ) and spacing adaptors (see Fig. 1) to find a combination that would present a satisfactory image with blue (435 nm) or green (546 nm) light. Various combinations were assessed with the aid of t,he ruled disc in a stationary rotor in a manner to be described below. A good image was ob-

TELEVISION

A

-

‘Jl6 c

B

C

SCANNER

D

E

F

G

H

I

104

FIG. 1. li’pper. Scale drawing of camera lens system used with the OMA. The scale in the vertical direction is exaggerated, being about 8 times that in the horizontal direct,ion. 1~1 and LI are the condensing and camera lenses, respectively; Jo and f~ denote their focal points. 0 is the object, 0’ is the apparent object as seen by the camera lens, and I is the final image. /,ower. Scale drawing of optical tllbe and vidicon housing. All dimensions arc in inches. The assembly as shown is mounted on its side; in use it, wordd be rotated !lO” counterclockwise. The optical tube C, constructed of heavy-walled aluminum tubing, was welded to a mounting flange. A sliding aluminum block H permitted adjustment ol thr camera lens (A) position. The light beam could be blocked when desired with a movablr shutter (I)). An adaptor cylinder E was used between the optiral tilbe and the vidicson housing I. The optical image was focllsed at the light-sensitive fare F of the vidiccm Ii. The double cross hatched regions, one of which is labeled (;, tlenotf, the focusing, drft‘c’tic)n, and nlignmcnt coils for t,he vidicon tube.

t,aincad with th(i 193 mm kns and the spacclr shown in Icig. 1.

iT it w(‘r(’ not’ for diod(x l(hakagc> caurrcint and/ or incident light. The signal assot-iatcd with t,hc intensity profile of the image is measured by th(l magnitude of the recharge current as This description of how the ORIA works thr &ctron bram r(‘scans each resolution will be limitcld to g;clncral operational princkmc~nt, of the mosaic array. The sharply ciplcs; spccGfic* d&ils of the circuit th(,ory focused bram is magnetically dcfktcd over nil1 bc omit,tcd. t,hc> dosircld aria of the image, rt~charging The opt’ical image> which represents thch ct:tc*h channc~l &mcnt in a zig-zag, noninterincident light signal is focused ontao thr laced pattclrn producing an output. video planar surfac*t> of th(B vidiron target. Tho signal. Th(b scaan pctriod rquircld for the size t,argct, is a silkon substrate: with a mosaic of th(l rastrr arca which covers the> projc&d array of silicon diodes having an average image is 32.X mscc. spacing of 12.5 pm. This diode array is Although several vidicon tube typtls arc initially charged to a uniform charge level by available, t,hc silicon vidicon camera t,ubc an electron scanning boam within the vidicon was st>lected for the OMA bc~ausc of its tube. The chargcx dcposit’ed by the electron very broad spectral range, high quantum beam would remain on the diodes indefinitely &ic~ic~nc~y, linear response ov(‘r :I wid(> dy-

868

RICHARDS

AND

namic range, and low geometrical distortion. A key feature of this array of silicon diodes which makes it so useful in the detection of an image generated by incident light is the efficient storage characteristic of the diodes. That is, the retention period is sufficiently long that no signal is lost for the duration of the scan period, but fast enough that changes in the image intensity which occur within the scan period can be detected. This method of electron beam replenishment in a silicon target to measure the signal intensity is a linear process unlike the case with standard vidicons and photographic film. The silicon vidicon is rugged and reliable because it cannot be damaged by exposure to high light levels. The particular model of the silicon vidicon tube used for these studies is a new experimental type that is sensitive to light over the wavelength range 200-1100 nm with quantum efficiencies ranging from 10 % at the extremes to over 70% at approximately 500 nm. The working surface of the silicon vidicon tube is essentially continuous, but consists of a rectangular array of 7.5 X lo5 silicon diodes within the 12.5 mm by 10 mm arca

ROCKHOLT

scanned. The diodes are approximately 5.0 pm in diameter on 12.5 pm centers. The discontinuous nature of the dioide surface is not pertinent for the rest of this discussion. The target area is divided into two equal rectangles with the image presented to one, while the other provides a partial dark current correction (see Fig. 2). The target is scanned in a zig-zag fashion with the electron beam. One can speak of the target being divided into an array of 500 lines, but each line is formed as the scanning beam makes an upward sweep from the dividing line to the upper edge then back down to the dividing line. The systems block diagram of the OMA is shown in Fig. 3. In the detector assembly the target signal is amplified and converted from a current to a voltage, then passed on to the console. The resulting signal is integrated over the line length, then the dark current signal from the continuation of the beam scanning the dark region downward from the dividing line and back is subtracted. The resulting voltage is compared with a ramp voltage, thereby converting the signal to a pulse width. This pulse is then used to

I.R. [[In-

rir

II I, -r 1 , , I 100 ; ’ 200 1 I

I I 300

rm

I I

br

1

400

n

I/

I

I

FIG. 2. Approximate scale drawing of absorption image on vidicon face. I.K. and OX. are the inner and outer reference holes, respectively, with rir and T.~ being the inner and outer reference edges. The vidicon tube is oriented so that. the scanning lines are perpendicular to the radial direction T, the channel number 12 then being a parameter proportional to r. To convert this figure to the image actually seen at the top of the optical tube, the observer faces the front of the ultracentrifuge and looks at the figure from the reverse side of the paper with the inner reference hole at the front of the machine.

TELEVISION

S69

SCANNER

I I digital printer or computer

--, i i

Model OMA

1205F Detector

, )

Head

input optics

I Model ; OMA 1 , I

1205A Console

preamp

FIG.

3.

Systems block diagram

gate a 16 MHz crystal clock to produce count,s in one of the 5-digit BCD counters in the memory units. As the scan continues the signal is processedin the same manner, but with the counts routed to different memory channels. At t,he complet’ion of the image scan in 32.S msec, t’he electron beam is returned to the origin and the scanning process repeated. If desired, the count’s for each line ma,y be accumulated into the separate storage registers, and the entire processrepeated. Any desired number of scans from 1 to 9999 can be selected by means of thumb wheels. There are t1v-o separate 500-word memories, called A and B, which can be used to store accumulated scans. Each channel can handle up to 99,999 counts. When this number is exceeded, the storage register cont,inues to accumulate starting again from zero. Thus the registers may be allowed to overflow any desired number of times, permitting the accumulation of a larger number of counts to reduce the error of background noise. With the “Full Scale Hold” button depressed, the accumulation is terminated as

of the OM.4.

soon as any of the 500 chamwls exceeds 98,000 counts. The number of scans for this accumulation are displayed when the “Accum Cycles” button is depressed. Thus, the operator may quickly determint the number of scans that n-ill nearly sat,uratc one of the channels. A digital to analog convertor is used to generate a signal appropriate for driving a cathode ray tube display monitor, thcwby providing rrbal time viewing of t,hc imagrl. One also has the option of examining the image stored in either of th(l mttmorics OI the arithmetic difference brtween them. On the front panel is a numerical display of channel number and t’he counts in that channel. A “cursor control” is used to ticlect any desired channel with its counts. The spot on the monitor that reprcwnts the channel

being

displayed

on the front

parwl

is

brightenrd, making it easy to dc%crminc the counts

for

any

desired

spot.

The

display

controls can be used to clspand the image horizontally about the cursor location, (‘11abling the user to examine fine st,ruc%urr. The pancll display van hc used to read thtr

870

RICHARDS

AND

contents of any desired channel in either memory or the difference between the memories for that channel. In real time, every tenth scan is displayed, making it possible to follow changing events in the desired channel. Output channels are available for attachment of an X-Y plotter or an appropriate recorder. Such plots are useful only for diagnostic purposes, but not for determination of absorbance values. Since the recording of data from the panel display is too time-consuming, additional methods for fast data output are available. We used a digital printer which prints the channel number and its counts at rat’es up to 10 lines/see. At lower speeds, it is possible to print the digital data and plot the analog output simultaneously. Many other output options are possible, including teletype print and tape punch, high speed tape punch, digital magnetic tape, and computer interfacing. The OMA is constructed with signal inputs and outputs available on the back panel to allow remote programming of most of its functions. Two additional features, which are not generally useful for ultracentrifugation, are also present in the OMA. First, one may display the logarithm of either memory or of the difference between them. However, the analog log circuit is not sufficiently accurate for our purposes. Second, one can sum the contents of any desired number of consecut,ive registers. This could be useful for determining the light flux within a desired scan region. Light Source and Rotor Pulsing

The image from a spinning rotor examined in real time on the monitor exhibited two kinds of pulses: one, a continually undulating pulse independent of rotor speed and, the other, a traveling pulse that moved to the left or right whose speed depended upon rotor velocity. The traveling pulse was not present with the rotor removed, but the

ROCKHOLT

undulating pulse remained. Replacing the ac mercury arc lamp with a constant-intensity lamp (ordinary hand flashlight) eliminated the undulating pulse, but the traveling pulses were still evident with a spinning rotor. Thus, the undulating pulse is caused by the ac light source, and the traveling pulse arises from the spinning rotor. A theoretical analysis of the two kinds of pulses was performed to determine their effect on accumulated scans. The results, presented in an Appendix, demonstrate that the pulses can be considered as random noise. Their effects can be reduced to a negligible error (less than 0.1%) if (a) more than 1000 rotor turns have occurred and (b) several hundred scans are accumulated. For fewer scans and/or a small number of rotor turns, the equations presented in the Appendix should be examined with the particular experimental parameters. Alignment

of the Optical System

The general alignment procedure of Schachman et al. (11) was modified for use with the OMA. The correct vertical position of the light source was verified for the light passing through the 254 nm filter. The uv image was examined visually with a piece of glass covered with phorphorescent materia?. (To use the OMA with this procedure would require a special mask with a spacing of about 10 mm for the parallel bars.) It is estimated that the light source could be positioned to within &l cm, which is equivalent to an angle of &l min for the light passing through the chamber. The light source height was, of course, incorrect for the 435 nm blue light used for all experiments reported on this paper. Thus, the blue light passing through the cell was slightly 3 A suitable material is sodium salicylate which can be deposited onto glass from a solution in methyl alcohol (for further discussion, see SAMSON, J. A. R., “Techniques of Vacuum Ultraviolet Spectroscopy,” p. 212, John Wiley and Sons, New York, 1967). A plastic film scintillator NE104 can be obtained from Nuclear Enterprises, Ltd., Glasgow, Scotland.

TELEVISION

8CANNE:I:

Sil

lens positions within 1 mm of each &her. divergent. The light source was positioned horizontally with the aid of a reflecting cell Since the meniscus reflection technique was more sensitive, the (barnera km was left in in a spinning rot’or. the position determined by it. The ruled At this point, the optical tube with camera disc method is theoretically more sound; it lrns in place was attached to the support could, perhaps, be improved by comparison plate. Shims were placed between the flange and thch plate to level t,he upper surface of of stored images from different camera lens positions. the tube, as dckrmincd with a small maTo determine the reliability of the focusing chinist’s level. The correct horizontal posiprocedure, t,hc remaining optical paramr+clrs tion of the tube was det’ermined for a spinshown in Fig. 1 were measured, then comning rotor containing a counter-balance and pared with values calculated with firsta ~~11 half-filled with water. The tube was order lens equations (9). Thtr measured shifted until, as determined by visual invalues \verc 1~~+ 2+ - 211+ (11 = S5.0 (‘1~1 spc>ction with unfilt’ercd light, the image was and v2 = 29.2 cm. From these values and cscbntclrc>dfront-back and to the right of the those given earlier, a value of S6.1 cm was ct,ntc>r-line. This orientation presents the calculated for u2 + u2, giving z12= 29.2 cm in correct format for the vidicon tube, as shown c~xcrllent agreement wit,h t hc mcasurc>d in P’ig. 2. With this arrangement, the light value. The final magnification factor was bclam passes through tho camt’ra lens offcaalculatc>d to be O.:i289. From this factor, an c*c>ntclr. Optical abtrrations would bc reassumed separation of 1.600 cm btkwerln th(> duccld by passing thr beam through the WItclr of thcb camera lrns and incorporating a refrrenctl t>dgcs of the countctrbalanc,cl, and a separation of 100 chann& bctwcbc>n thcb micromttc~r stage with the adaptor cylinder, refcrencc cdgc images, one cakulatcw a disE: (l:ig. I), to permit offsetting the vidicon tance of 11.66 mm for 500 chann& on thca rc4ativ(l to the image>. vidicon fa(s(‘, in good agreement with thus ‘1’1~~camclra luu focusing procedure using a manufact,urrr’s cstimatcd value of 11.75 IIHII. spinrling ccl1 half-filled with water was easily (It should brl noted that the distanc~t~ performed with the OMA operating in real scanned is adjustabk over a small range’, time. The meniscus region was magnified by plus or minus, from thch nominal vnluc~ ol proper adjustment of t’he display controls 13.5 mm.) u&l that dots rcprcsenting individual thanncls w(‘re cleanly separated from clach &hrr. With the light source moved forward, thtl correct position of tho camera lens rcprcscntEven though rotor and lamp pulsing ing even illumination on both sides of tht, produced a c*ontinually changing irnagct as nlckcus could b(l found within a minute. viewed on the monitor, th(, image produced ‘i’h(~ positioning of the light source’ involving from accumulated scans was r~~markably a similar procsc>durcl with the camera lens constant. With the controls appropriately drfocuscd also rcquirc>d an clqually short time. scxt the stored image could be rlxamined durTh(b ruled disc in a &tionary rotor was ing thtx c>arly stages of the accumulation. The also usctd as a11 aid for the focusing of the pulsing appearance of the growing ima,gcl dir;camcxra kns. ‘1 o diminish the light intensity appeared rapidly. neutral density filters of about 2.3 absorbance I&cause the upper and lowcr halvc~s of thcb units (0.5 % transmission) were placed with vidicon tube (sue Fig. 2) do not behave the intBerfcrcncae filter above the light source. identically, the internal dark current subThe camera lens was adjuskd to give the traction of thcb low\-cr part ia only approxisharpest image of the linc>s. mate. Therefore, afkr each ac*cumulat,ion of Th(a tn-o focusing methods gave camera an image,, the shutter \vas closed, md it dark

RICHARDS

AND

current image obtained with the same number of scans. The final image, representing their difference, was then printed onto tape with the high-speed printer. Figure 4, taken from a sedimentation equilibrium experiment with horse heart myoglobin, illustrates the necessity of the additional dark current correction. The figure is a photograph of plots obtained with an X-Y plotter taking transmission data directly from the OMA. A is the direct image, B is the image of the dark current (obtained with the shutter shown in Fig. 1 closed), and A - B is the difference as subtracted internally in the OMA. Also shown (C) is the hand-drawn plot from a separate trace taken with the cell rinsed and partially filled with solvent. There are controls within the OMA that serve to raise or lower the dark current image as seen in real time, as well as to change its profile. Because of the presence of blemishes within the tube and the discontinuous nature of the silicon diodes, it is not possible to ob-

ROCKHOLT

tain a smooth image. However, as will be demonstrated later, each line element behaves correctly as an individual phototube. A serious blemish, such as the one at about channel 30, is apparent as a large dip in t,he patterns A and B, but not in the corrected A - B pattern. In fact, this blemish served as an aide for focusing of the vidicon electron beam. The overall appearance of the subtracted pattern is much like light intensity patterns from the photomultiplier scanner (1). It should be noted that the intensity drops to zero in the dark space (channel 425) between the cell and outer reference hole. The pattern dips below 0 counts in the dark space (channel 50) between the inner reference hole and the top of the cell, indicative of the fact that, for this region of t’he vidicon, more light was present on the lower half than on the upper half. No attempt was made to find the source of this stray light or to reposition the lamp so as to produce more even intensity across the cell.

rkml 6.4

FIG. 4. Transmission equilibrium experiment

curves obtained with horse heart

6.6

Channel

No.

from the myoglobin.

OMA.

Obtained

from

a sedimentation

TELEVISION

Wit,h the 2.5.mm apert’ure at the light source and the 435nm interference filter (Type 14-53-7, Baird-Atomic, Inc., Cambridge, MA), some of the channels in the inner rcfercncc hole were saturated. We decided t’o use routinely a 0.6 neutral density f-iltcr (25 % transmission), which gave a sufficiently low intensity, so that more scans could be accumulated, thereby reducing the cffcct’ of rotor and lamp pulsing (see Appendix). This light level gave about 300 Z!I 20 counts/scan at the inner reference maximum, so that for “TO-300 scans about 80,000 counts were accumulated. Cwrection

oj 7’ransmission

Patterns

The absorbance of a solution is calculat8ed from t,he expression A = log lo/I, where 10 is the intensity of t’he light beam passing t,hrough a cc11wit.h solvent alone and I is the intcn&y of the light beam passing through the same or an idernical cell filled with the solution of int)crcst. Since we were limited to single beam operation for the experiments reported here, it was necessary to obtain a scparatc solvent pattern before or after each expcrimcnt (The transmission patterns were always corrected with a dark current accumulation of the same number of scans. ) Bcforc calculation of absorbance, it is necessary to compensate the patterns for va.riations in light’ intensity and the process of photon-to-(,ounts conversion. The height of the pattern in the inner reference hole proved to be an adryuat’e monitor for these variations. Since the reference holrs arc J&hapcd, the transmission patterns produced from them have only a narrow maximum. Thcrc were, however, three than ncls that had nearly the same number of couuts, so t,hry were averaged to produce a single rcfcrcncc value. Thus tfhe absorption *1 (n) for any channel number IL was cslculated from the relationship A (n) = log Klo (n)/I

(n),

s73

SCANNb:lt

0)

where lo (TL) is the number of counts in channel n for the solvent image, I(n) is the num-

her in the same channel for the solutzion image, and K is the ratio of the solution reference value to the solvent value. It! should be noted that the log term can bc separated into log (lo/l) + log K, which means that the correction changes the cl+ vation of the absorbance curves, but, not its profile. The corm&on factor was generally betwocn 0.99 and 1.01, and only once did it’ achieve a value as high as 1.03. Reproducibility

of Accumulated

Scans

The reproducibility of the transmission curves was readily assessed by storing a transmission image in one set of storagcb registers, then the same image in the other set. The difference, which is corrrctcd for dark current, can be examined on t’he monitor screen, or plotted directly. One set of registers can be crascd, a new image accumulated, and the diffcrencc again (lxamincd. 011~ can compare xucccssive imagrs or look for “drift” by comparing fresh images with a stored image. From examination of hundreds of difftrence patterns, wc found that a diffrrencc of up to scvcral thousand (but usually less) out of 70,000 counts often occurred. The diffcrcnccs increased as the time bctwccn images was incrcascd. Thcsc diffcrcnccs could have arisen hccausc of drifts in (a) lamp intensit,y, (1~) storage propcrtics of the vidicon target, or (c) clcctrical propertics of components involved in the conversion of the analog signal to digital numbers. (11s shown in tho Appendix, the effects of rotor or light source pulsing arc too small to bc ol~sc~~:iblt? when ticvera hundred w:ms rcprrscnting at least 1000 rotor revolutions arc a(*cumulated.) One would cspcct thcb “diffemme” images to rcstmblc :I miniature transmission pattern, but in hon~(’ casts they curved slightly up or down. I:or cxanlpie, in l:ig. 3 is .show-n a difference transmission pattern ahovc ow of the transmission curves. (The transmission patterns usccl hcrc include the dark cturcnt correction. ) E’or most of the chanricls in the irincr rcfcr-

874

RICHARDS

2:

AND ROCKHOLT

.

I*

.

g:

07.. :: o.oaF I 5.6

.2

C

;

.;

ABSORBANCE

“t 5.8

2 6.0

6.2

6.4

66

6.6

70

7.2

7.4

r km)

FIG. 5. Reproducibility of transmission dark current correction, B is the difference absorbance curve calculated for the B curve. t and b are the top and bottom of the cell, phosphate (pH 6.6); speed, 26,666 rpm.

plots. A is a single transmission curve, after between two transmission curves, and C is the i and o are the inner and outer reference edges, and m is the meniscus solution, 0.5~ sodium

ence hole and the cell, one observes a negative difference of about 1,000 f 200 counts or less. However, toward the bottom of the cell the pattern rises gradually until it is about -100 counts. One must assume that either the arc pattern within the lamp moved so as to alter the light intensity across the cell or that the properties of the silicon target or electron beam of the vidicon changed unevenly. Understanding and correcting the curvature in the difference patterns will require further work. A frequently occurring feature of difference patterns from scans taken immediately one after the other was appearance of “spikes” at radial positions where the intensity changed rapidly, namely the reference edges,the top and bottom of the cell, and the meniscus (see Fig. 5). These spikes are readily explained by assuming that the two images were slightly displaced from each other. This displacement, representing about Ho of a channel, is equivalent t’o a 4-pm shift of the rotor axis or a 2-pm shift in the scanning beam. Examination of a stationary target placed in the light path would enable one to evaluate the st’ability of the electron

beam and thereby determine the extent of its involvement in image displacement with a spinning rotor. The data for two images, plotted as a difference at the top of Fig. 5, is also plotted as a,bsorbance at the bottom. The average variations, except at the bottom of the cell, is less than about 0.001 units. In general average differences for successive scans of low absorbance solutions differed by no more than 0.003 units. Vchen the rotor was accelerated from 26,000 to 60,000 rpm, the image shifted about 5 channels toward higher channel numbers. This shift of 0.02 cm is the same as the generally accepted figure for the rotor stretching under these conditions, but additional information is required to distinguish stretching from a shift of the rotor axis. A routine check of rotor shift is not readily accomplished with the commercial scanner, but is simple with the OMA. RESULTS Accuracy

AND DISCUSSION

of Absorbances

from

the OMA

A critical performance test for the OMA is to verify that absorbances determined

TELEVISION

with it are linear with absorbances measured with a spectrophotometer. Since the counts obtained from accumulated scans are proportional to the light intensity, one obtains absolute absorbances without the need of calibration. For the first test, three hemoglobin solut’ions (obtained by careful dilution of a stock solution) with absorbances of about 0.3, 0.6, and 1 and also the solvent were individually centrifuged at 26,000 rpm. The image from 270 accumulated scans was printed out shortly after reaching speed. The absorbances for four widely separated channels were calculated according to Eq. (1). The calculated absorbances were not the same as those measured with a Cary 15 spectrophotometer, but’ a plot of OMA absorbance versus dilution factor or Cary absorbance was linear. We attributed the difference in absorbance values as measured by the two methods to uncertainty in knowing the effective wavelength of light passing through the 435nm interference filter that we used. Moreover, this wavelength is low on the right, side of the hemoglobin Soret peak. In searching for a dye with a flat absorption peak at the desired wavelength, we found that a mixture of green and black recorder pen inks (Part No. 366316-3 and 3541322, respectively, Honeywell) in water gave a very flat peak at 435 nm. In other words, the extinction coefficients for the polychromatic light passing through the filter were nearly the same. Moreover, the ink solut’ion had a relatively flat absorption profile, thereby serving as a neutral density filter and blocking about the same percentage of any residual light of other wavelengths that was passed by the filter. A series of seven solutions (obtained by dilution of a conccntrat’ed stock solut8ion) encompassing the absorption range from 0 to 2.4 and a water rcfcrence were centrifuged at 24,000 rpm. Again, the images from 270 accumulated scans w-fre printed out. Rather than analyzing the c>ntire cell, we selected 10 neighboring cbhsnncls in the bott,om third of

HCANNEK.

870

the cell, which is the region used for sedimentation equilibrium experiments. We also selected four positions 50 channels apart’, the first being at about 6.2 cm from the center of the rotor. The absorbance values were calculated according t’o Eq. (1 ), then divided by 1.2 to compensate for the 12-mm thickness of the cell in the direction of the light path. The results are plotted in Fig. 6 against the absorbance of the same solutions as determined by measurements with the Cary and dilution factors. For the calculations involving the 10 channels, only the average value is plotted, while individual values are plotted for the 4 channels to indicate the range. It can be seen that the plot is linear from 0 to 1.0 with a slope very near 1, indicating correspondence of Cary and OMA absorbances within the expcrimental error of the former. From examination of individual absorbance values from the lo-channel cluster, we find that absorbantes from 0 to 1 are accurate within f0.003 absorbance units. The error increases at higher absorbancrs, a behavior shared by photomult,iplier scanners. The error in absorbanccs measured in the> OMA can be understood in terms of t,hc manufacturers stated root-mean-square noise equivalent input, conservatjively cwtimatckd

I-------

4/:

.L--.I-L 0

-1

I

2

A, Cory

FIG. ci. Linearity- of absorbance measrlrements u-it,h the O,\ZA.

876

RICHARDS

AND

ROCKHOLT

at about 1 count/scan. If it is assumed that the noise level should be multiplied by 1.4 for the image obtained with the dark-current correction, there is a noise level of 380 counts for 270 scans. Since there were about 60,000 counts accumulated for 270 scans of the solvent, there would be an uncertainty of ~tO.003 at A = 0; f0.03 at A = 1, for which 6,000 counts are accumulated and f0.2 at A = 2, for which 600 counts are accumulated. From our measurements of absorbances from 0 to 1, it appears that our vidicon was operating with a noise level less than the manufacturer’s specifications. It should be possible to increase the accuracy of the measurement of absorbances above 1 by accumulating more scans, but we decided to limit our experiments to absorbances of about 1.2. As a test of the OMA the approach using a “neutral density” solution is better than using a solution with a narrow absorption maximum, because it allows the test of the OMA independent of any problems with stray light or the polychromatic nature of the light source. (With the green and black inks described here and the inclusion of a red ink, it should be possible to create within a few minutes a solution with a flat absorption

profile in any desired spectral region.) Once it is decided that the OMA is performing in a satisfactory manner, then it is appropriate to look for stray light by other means. Sedimentation

Velocity

Intensity patterns from sedimentation velocity experiments bear a superficial resemblance to absorption patterns, but they are not useful for interpretation until converted to absorbances. Two absorption patterns derived from an experiment with human hemoglobin are shown in Fig. 7. A computer program was used to perform the calculations indicated by Eq. (1). The OMA is especially suited to produce the information needed for a thorough analysis of sedimenting boundaries with any available computer program. However, when only a sedimentation coefficient is desired, the conversion of parts of the curve to absorbances in order to locate the halfconcentration level is an unnecessary tedium. A method, simple in principle, will now be developed. We will assume that the centrifuge cell receives uniform illumination and the solvent and velocity patterns all received the same total amount of light. The absorbance A, in the plateau for any given

,:’ ,:’ ,,...’ ,.”

2 :,/” o&-. 6.0

..,j

/

.‘.---T 6.2

;r’

/ ...*’

, 6.4

I

I

6.6

6.6

,

I 7.0

.

d

7. 2

r (cm) FIG. 7. Sedimentation velocity patterns from human hemoglobin. phosphate (pH 6.6); speed, 60,000 rpm; room temperature; pattern pattern B, 168 min.

Solvent, 0.5 M sodium A, 107 min at speed;

TELEVISION

pattern is obtainnd from Eq. (l),

0( ii

SCANNER

Sedimentation Equilibrium,

As a test of t’hc rc>producibility and accuracy of the ORIA, t’hree sedimentat)ion where I, is thtx solut’ion intensity in this equilibrium rxpcriments were performed rc@on. It is netscssary to find the channel with horse heart and sperm whale myonumber for which n = A,/2. The half- globins. To cover a wide absorption range, the conditions for high spctxd sedimttntation c*oncentration lfvcll ?‘r.5occurs at an intensity equilibrium n-crux scktcd, namely a speed IP clquivalent to A $2 so that of 44,000 rpm for a 3mm column height at room trmptraturcb. Several ONA scans wertl obtained aft’er sedimctntat,ion equilibrium Elimination of A P from t.hesetwo equations was achieved. Thck cell was carefully rinsed gives without, disassembly and filled with solvent to a slightly higher lt~~l. Baseline scanswcr(’ I, = (I,,&y2. obt’aincd a,ftcr this cc11had run about 30 min at the same spccbdused for t,ho cquilibThus Y,,.~is located at the intensity which is rium c~xpfrimcnt t#he geometrical mean of the intenskics of Thtk absorbanccb and channr~l number thrl supernatant a,nd plateau regions. Cordata, tog&cr with the necessary cxpcrrections would havcb to bc applied if the mental paramc+rs, were treated with t,he heights of the rc~feroncr:hole for the velocity computer time-sharing program dtv&pcd and solvent patterns differed significantly. by Dr. It. Il. DJWIL V~UCY of Y, ?, A and E’ivn intensity versus channel patterns 111A4 n-crc first printed in tabular form. obtained from the same velocity experiment, Then In 11 \vas plotjtcd vs 1.2- /.,,2 on-line: described above wcm used to locate the from thcl c*omputtr. lcrom c>xamination of channcll numbor corresponding to the halfthis plot various decisions could bc made as caonccntratjion level. The two patterns corto whkh of stvcral subroutines could bc responding to Fig. 7 gave within one channel most profitably utilizcbd, ineluding (a) the number the sam(’ position as determined discarding of certain points, (b ) fitt,ing of from t>xamination of the table which listed d(sircld regions with loash mean squarrbs channel numbers and absorbancfls. The five straight, linrs to gk apparent niolccular positions when plottc>d as log ro.5vs t gave a straight line with a rcasonablc deviation of wclights, ((a) changing the absorbance valuc~ the c~xpttrimcntal points. A value of 3.9 S at thck meniscus, or (d) the c~aleulat~ionof was found for s~o,~, somewhat less t,han the number, n-eight , and %-nvfrnge molecular value of 4.13 S obtained by Chianconc et al. weights ov(lr a span of that dcsircld number (12). However, the hemoglobin under our of points. With thcblast routine:, one also had condition of 0.5 M sodium phosphate (pH thcl option of sckcting a drsircd number of 6.6) was partially dissociated, as determined smoothing passesin which the central point of the cdlustcrwas movc>d toward t,hcl value from a sedimentation equilibrium cxpcric*alculatcd from thcb (urve fitting. nicbnt’. When fitting thcb In A vs 1,’ data with a It should bth fmphasizcld t’hat thr imcnsity profihs did not, change with speed, clxc*(:pt straight line, tht, program w-cbightsthe points for small a,pparcnt shifts in the radial diroc+- in proportion to tht, ahsorbancck.Since higher tion described ckarlier. Thus calculated ab- absorhancc~shnvcb a larger, not 3 smaller, sorbanccs were’ indt>pendcnt of sp(aed, in unc~crtainty, the SlOpc~ for 01w wtJ of datil caonstrastwith spcl&dtxpendent absorhanc~c~ was calculated \vithout wcbighting any of the d(lscribcd by othrr works (5). points. ‘Ihcrc~ hcling no significant differcnc~c~ A, = log (IO/I,),

878

RICHARDS

AND

in the calculated slope, we left the weighting method in the program. The plot of the logarithm of absorbance versus radius squared for one pair of scans for sperm whale myoglobin (summarized in second set of results from bottom of Table I) is shown at the top of Fig. 8. Since the data from absorbance values below 0.02 were more highly scattered and those above 1.0 deviated sharply downward, they were omitted from the plot. It is evident that the points curve downward to the right and cannot be fitted adequately by a straight line. Thus the points were fit with two straight lines, one from A = 0.02-0.2 and the other from A = 0.2-l. From examination of the second line, it is seen that the points still curve downward. The deviation plot shown in the middle of Fig. 8 represents the deviation in absorbance units of the experimental points and those calculated from the two skaight lines. At the bottom of the figure are shown apparent weight average and number average molecular weights W,&PP and M,,app, respectively) calculated CM from least-squares quadratic fits of fivepoint spans. (Such analyses are not generally performed with the commercial photomultiplier scanner because of scatter in the experimental points. ) The extrapolated values are

17,630 and 17,810, respectively. Since the values for apparent Z-average molecular weights were badly scattered, they are not shown. The results from eight scans representing three experiments are shown in Table I. The molecular weights with standard deviations obtained from slopes for absorbance values between 0.02 and 0.2 are shown in the column labeled Ml and those obtained from values between 0.2 and 1.0 are given under M,. The least mean squares deviations in absorbance for each slope are also given, they average about 0.002 for Ml and 0.005 for Mz. In the last two columns are given the intercept M,’ and slope K obtained from the apparent molecular weight M,,,, plotted against absorbance. (In these plots, only the absorbance range from 0.2 to 1 was used; if the points near A = 1 deviated strongly downward, they were discarded.) The molecular weight values for horse heart and sperm whale myoglobin based upon sequence determination are 16,951 (13) and 17,199 (14), respectively. Our Mz values differ somewhat from the sequence values, but a direct comparison is not valid owing to uncertainties in the partial specific volume [we used the value 0.741 of Theorell (15)] and in the purity of our samples, which we

TABLE MOLKCULAR

Scan No. Eq. Bl.

WEIGHTS

Ml x

Heart 1 5 2 5 3 4 Sperm Whale Run 1 6 8 7 9 Run 2 10 13 11 15 (Add 0.004) 12 14

OFMYOGLOBINS

ROCKHOLT

I BY SEDIMENTATION

Mz

L.M.S. AAa

10-4

x

EQUILIBRIUM

10-4

L.M.S. AA

Horse

a L.M.S.

is least

1.605 1.666 1.755

f f f

0.021 0.020 0.012

0.0021 0.0018 0.0013

1.721 1.748 1.753

f 0.007 zt 0.007 f 0.006

0.0041 0.0036 0.0033

1.672 1.661 1.697

1.630 1.540

f f

0.038 0.018

0.0066 0.0023

1.659 1.672

f f

0.012 0.008

0.0066 0.0036

1.865 1.730

1.690 1.749 (1.648 1.756

xt f f f

0.012 0.015 0.016) 0.014

0.0016 0.0016 (0.0020) 0.0014

1.678 1.684 (1.669 1.680

f f f f

0.098 0.007 0.006) 0.008

0.0047 0.0028 (0.0026) 0.0052

1.749 1.762 (1.730) 1.809

mean

square

deviation.

.

0.6 1 -0.9

-3.9 1.2 -1.5 -1.4 (-1.1) -2.3

TELEVISION

579

SCANNER

08 06

3

35

4

45

5

a r2

.

FIG.

8. Sedimentation

equilibrium

did not examine by other biochemical techniques. Further examinat’lon of the values in Table I is, however, useful. Because there was less scatter in the experimental points on the In A vs r2 plot, one can conclude t’hat the M, values are more reliable than the M1 values. In fact the scatter of the points for the M2 plots was sufficiently low that one could be satisfied with these molecular weight values. Further analysis of the molecular weights obtained at different levels in the cell revealed an apparent concentration dependence. For the first two experiments, there was sufficient scatter in the Mw,app vs A plots to make the extrapolated values of M,’ uncertain. The scatter in similar plots for the last experiment was satisfactorily low. However, the concentration of myoglobin at the bottom of the cell was calculated to be about 1 mg/ml, which is t’oo low to account for the rather large riegat~ivc K value. We invcst~igated three factors which

of sperm

whale

myoglohin

could lead to reduced absorbance mcasurements. First, the effect of adding a small absorbance at all positions in the cell was studied. Shown in parentheses near the bottom of Table I are the results when 0.004 absorbance units wcrc added to the data from the preceding set of scans. M1 de creases about 6 %, while Mz and M,’ de creaseonly slightly, but K is still too large. The presence of about 1 % st,ray light would lead to a progressive decrease in measured absorbances similar to what’ we observed. There is a possibility that a small amount of light at other wavelengths is passedby the filter. Another possible source of stray light is from reflcct’ions off lens surfaces or cell windows. Finally, a small amount of light could have leaked through the various seamsand openings of the optical system. While we did examine the effect of the various room lights on the stored images, a small leak could easily have been overlooked. I’inally, we considered the eff(bc*tof the

880

RICHARDS

AND

radial slit width per channel (about 0.004 cm) on measured absorbances, making use of a theoretical equation relating I t’o r. If it is assumedthat the light-sensitive surface of the vidicon is uniform, the slit width was much too small t’o have any effect on the absorbance. However, it is possible to construct a diode distribution that would lead to an error similar to what we observed. This source of error could be examined by mounting the detector assembly on a micrometer stage and comparing images displaced from each other by a small fraction of a channel separation. We were unable to investigate possible effects of distortions of the final image caused by camera lens aberrations or nonlinearity in the scanning pattern of the electron beam of the vidicon. CONCLUSIONS

The work reported here was aimed at assessingthe feasibility of employing the OMA as a scanner for the ultracentrifugc. The results demonstrate that its performance is better than instruments using a mechanically scanned photomultiplier tube (l-6). In fact, the data obtained from one scdimentation equilibrium experiment was sufficiently good that it was worthy of analysis by computer programs usually reserved for the Rayleigh optical system, which is generally acknowledged to be more accurate than the conventional scanner. This analysis revealed a discrepancy in absorbance measurement which will require further investigation to determine its cause. On-going investigation involving computer-interfaced commercial scanners are almost certain to uncover similar discrepancies as the measurements arc refined; likely sources of aberrations are the camera lens and the scanning screw. Based upon our work, we believe that the OMA coupled to a monochrometer with a suitable light source would function even better as a single-beam scanner. It could be used with light throughout the visible spectrum. Based upon the known spectral sensitivity of the particular vidicon that we used,

ROCKHOLT

we believe that the OMA could be used with uv light as low as 250 nm. The lower wavelength limit could be determined only by experimentation. The principal disadvantage of the OMA for use as a scanner with multicell operation is related to its advantage. Since the lightsensitive surface of the vidicon serves as an integrator for all the light striking it between successive scans, only one image can be examined during a given accumulation period. Comparing the vidicon and photomultiplier as light-sensing devices for multicell operation one can say that with the former only one image, but in its entirety, can be examined at a given time, while with the latter each of several cells can be examined during each turn of the rotor, but at only oneradial position. We have not demonstrated that the OMA can be used with double-beam operation or with several cells simultaneously in the same rotor. The photomultiplier scanner, however, has been readily adapted to doublebeam and multicell operation. Faced with the necessity of performing many experiments, many workers would not consider any scanner system unless it could be adapted for multicell operation. We have considered double-beam operation of the OMA by using wedged windows to separate the images from solvent and solution. A half-wedged window for use with double-sector cells is available from the manufacturer of the ultracentrifuge. If the two liquids are placed in separate cells, a latterly-wedged window in one cell would separate the two images. However, in the conventional optical system, the images from wedged or normal cells arc superimposed; a shutter mechanism placed near the focal plane of the condensing lens is required to block one image at a time (16). If a cylindrical instead of a spherical lens is used to focus the cell in the final image plane (one of the two configurations suggestedby Svenson (17) for use as a Rayleigh interferometer with cell-focusing), the two images are separated from each other, a configuration

TF:LEVISION

ideally suited for use with the OMA (see Fig. 2). Having constructed such a system and examined t,he image photographically, we believe that it has the potential of serving as a double-beam instrument. The use of this system would require modification of the ON-4 to separate the counts from the upper and lowtr half of the vidicon face into tha separate storage registers. We have abandoned further development of the split-beam syst’em in favor of two alternate solutions to mult’iple-cell operation currently under investigation by the manufacturer. One approach will involve the use of an image intensifier tube which would be placed between t,hr: optical image and the vidicon tube. The image intensifier tube has the potemiul of being gated by appropriate circuits so that only the desired hole or sector in the rotor is presented to the vidicon t.ubc. ‘Iht: other approach will involve pulsing a short-duration scnon lamp in phase with the desired cell. OINY a satisfactory method of mult)iopcrat,ion is found, the optical system will be oxamincd in greater detail, and methods of aut,omation of the OhIA, possibly with a nlinicomputrr, will be investigated. APPENDIX

&feet oj AC Arc Lamp To simplify the discussion the following conditions wcrc assumed: (a) the rotor is removed from the optical system, (b) the intensity of the light source is reduced to a level compatible wit’11 the O&IA, and (c) the image at the vidicon has negligible width in the direction perpendicular to the radial direction. The channels arc: read by the scanning beam at constant rate according to the function t = nAt,, where t is the time, n is the channel number, and At, is the scanning rate, 6.4 X 10-j scr*/c:hannel. The channels are numbered from 0 to 499, and the counts in channel 0 are read at t = 0. The time At, between scans is 512 At,, or 0.032763 sec. There is a small time interval between

SCANNKI:

8-4

the reading of t’he last, channel and the return to the first channel that is used to a(‘complish various electronic tasks. The number of light pulses per scan PI is obt.ained from the expression Pz = v&,

= 3.93216,

w 1

where vz is the on-off frequency (120 Hz ) of the 60-Hz ac mercury arc lamp. We will assumethat the light intensity of the lamp I(t) varies with time t according to the relationship I(t)

= lcl sin27r(v~ + o(),

(2A )

where a is an arbitrary phase constant and kl is the maximum intensity. (For angles in degrees instead of radians, 7r is rcplacrd by 180.) Between consecutive scans of the electron beam, each line element of the vidicon acts as an integrator for the light striking. it. If we assumethat the proc*rsses involved in the conversion of radiant energy striking the vidicon element to counts in the O&IA st)oragc registers are linear, the total number of counts ,V (n) generated for a given storage register during the>time intrrval from tl to ta is t? NC ?I) = x-2 I(l)di> (311) s $1 where l
= K (f? - t1) [ L TV1

cos

7rvz(tg+

t1 +

2a)

(42)

1

X sin 7w(t2 - fl,l ,

where K = lclk2/2. We now need to relate tl and tz to the scanning of successive channels. The channels are read at times t,,j obtained from tn,j = [TZ $- (j -

1)512]At,,

(5A)

wherej represents the scan number, I, 2, 3,

882

RICHARDS

AND ROCKHOLT

etc. It should be noted that j assumesthe value 1 while n goesfrom 0 to 499, then it is incremented to 2 while n goes again through the same range, and so on. We will let tz denote the current position of the scanning beam and 21 the time at which that channel was read on the previous scan. From examination of Eq. (5A), it is seen that tZ - tr is the same for all channels, representing At,, the time between scans. The value of tl + tz, needed for Eq. (4A), is obtained by summing Eq. (5A) for jandjl,togive

for j = 2, etc. The channel numbers corresponding to maxima can be calculated from t,he relationship vlAt,[2n + (2j 3)512] = -2, 0, 2, 4, etc., while those corresponding to minima are obtained from the left side of this equation set equal to the intervening odd numbers. The separation between successive maxima or minima is 1/ (YEAt,) = 130.21 channels. A plot of N(n) vs channel number for four successivescans is shown at the top of Fig. 9. The ordinate has been normalized to give an average value of 1. The undulating appearance of the real time display arises tl + tz = [2n + (2j - 3)512]At,. (6A) from an 8.833 channel displacement of the The real time display on the monitor can curve to the right (higher channel number) for successivescans. This displacement repbe constructed from Eqs. (4A) and (6A). It is assumedthat the scanning process has resents 0.06784 of the pulse width displayed been in operation for sufficient time after on the monitor. The same number is obexposure of the vidicon to light that all tained for the difference between the 3.93216 transient effects have been eliminated. Since pulses of the light source during each scan tz - tl is the same for all channels, the only and 4, the nearest integer. In fact the pulsing variation in Eq. (4A) occurs in the cosine frequency vC that would be obtained by counting the number of maxima occurring term. The first scan (j = 1) is begun with for any given channel during a measured (Y = 0 and n = 0 at tz = 0; the light intensity is 0 at this time. From substitution of time period can be obtained from the general appropriate values, it is found that 11+ tz relationship varies from -0.032768 set at channel 0 to vc = i/At, - vr, (7A) 0.031936 set at channel 499. The difference in these values represents 7.65~ radians, or where i is the nearest integer, in this case 4. nearly 4 complete cosine cycles with 4 With our parameters veis 2.07 Hz, which is maxima and minima. This variation in in good agreement with the value of 2.0 channel counts has the same frequency as obtained by measuring the time for 20 the on-off cycling of the light source. pulses on the monitor. This equation can be Let us now examine the variation in rearranged to solve for At,, giving an accuimage height for any channel. The value for rate measurement of the actual value for the sine factor is 0.2115, while the cosine the crystal clock of a particular instrument. factor varies between fl. Thus the second (If a number of pulses slightly greater than term varies between f5.61 X 10V4, which an integer were involved, v, would be negais added to the first term, 0.032768. The tive, and the pulses would appear to move total variation represents 3.4% of the aver- to the left instead of to the right.) age pulse height. From approximate height To determine the number of counts stored measurements on the monitor, we estimated in the registers after m scans, it is not necesthe pulse variation as 4 %, in good agreement sary to sum values of N(n) obtained from with the calculated value. successive values of tl + tz calculated from The changing profiles of the real time Eq. (6A). Instead, it is sufficient to use display are obtained by substitution of Eq. (4A) with values of tz(j = m) and successive channel numbers for j = 1, then tl(j = 0) obtained from Eq. (5A).

TEI,EVISION

SCANNER

200

XX8

300 Channel

No

FIG. 9. Theoretical real-time images. Upper, light-source pulsing. Lower, light source and rotor pulsing. The numbers 1-4 denote successive scans. Channel numbers to the right of the dotted line are not seen on the monitor.

We will now determine scan numbers that minimize the effect of the sinusoidal variation in light intensity. This is easily done by multiplying the number of pulses per scan, 3.93216, by integers, selecting those that give nearly integral numbers of pulses. A more informative method of accomplishing the same goal is to take the reciprocal of the difference between the nearest integer and the number of pulses; that is, l/ (4 - vlt,) = 14.7406 scans. Since a fractional number of scans has no meaning, wc take 15 scans. The sine fact’or for this number of scans is 0.0553. Thus, the second term has a maximum variation of f1.47 X lop4 compared t,o 0.4915 set, which is 3 parts in 10,000, a negligible error. E‘or twice 14.74 scans or 29.48, the nearest integer is 29, giving a sine factor of 0.102 and a fractional error about t,he same as the previous casr. If a number of scans half-way between succcssivc 14.74 incremcnt~s is taken, the sinecosine product has values ranging between f 1, and the term has its maximum value of f2.65 X 10-3. Such a number of scans is 110, but, the first term is sufficiently large, 3.604, that’ tha percent variation is only f0.15 %. Tha larger the number of scans, the smaller the variation. Still, it is better to select.a number t,hat is near a multiple of 14.74. h’ffect of h’$nning

Rot07

Let us now consider the origin of the traveling pulses from the spinning rotor

with a light source of constant intensity. To simplify the analysis, it is assumed that an aperture of negligible width is placed above the rotor containing a cell also of negligible width. Every time the rotor passesbeneath the aperture, light continues through the optical system and its effect is registered on the vidicon target. The effect of the light from successive revolutions is accumulated on t’he target until the scanning rlect,ron beam reads and replenishes the target. The difference in frequency of bhe light pulsesand the scanning processleads to t’he appearance of traveling pulsesin the real time display of successive scans. The origin and properties of these pulseswill now br explained. The number of rotor pulses P, per scan is obtained from Eq. (lA), with Y, as the rotor speed in revolutions per second. For sperds usually used with the ultracentrifugc, P, is a nonintegral number, say i + A. Since only integral numbers of light pulses arc accumulated on the target between successivescans, there can be only i or i + 1 pulses for each scan. When light from a rotor pulse strikes the target, the particular channel being read by the scanning beam will have accumulat’ed i + 1 pulses. Subsequent channels will accumulate the same number of pulses until a channel that was interrupted by a rotor pulse in the previous scan is rcachcd. Then i pulseswill be accumulated until light from a rotor pulse again strikes the target, whereupon i + 1 pulseswill be accumulated. A complete understanding of the pro~cssis

884

RICHARDS

AND

aided by the derivation of equations and the examination of figures. Let us begin our analysis with the rotor image striking the target at time t = 0, the same time that the first channel is read by the scanning beam, The rotor image appears on the target at times 2, given by the relationship

ROCKHOLT

10, one counts the number of rotor pulses that are encompassed by a moving line segment whose length is the number of rotor pulses equivalent to 1 scan, in this case, 7.0997 pulses. Both edges of this line segment always occur at the same channel number. The left-hand edge denotes the time at which that channel was read by the precedt, = mat,; m = 0,1,2, ... 7 (8-4) ing scanning beam, while the right-hand edge where At, is the time interval between rotor marks the current channel being read. To pulses (At, = 60/rpm). T’he channels are avoid the recording of the same pulse twice, read at times t, ,j obtained from Eq. 5. we will count the pulses encompassed by a The number of rotor pulses M(n, t) for line, including a pulse if it is at t’he right any desired channel can be obtained by edge, but not if it is at the left edge. At counting the rotor passages that occur bet = 0,8 rotor pulses lie beneath line segment tween the time period t, +.r < t 5 t, ,+ A. For line segment B, the left edge starts The total number of counts N(n) generated to leave rotor pulse number -7; the right for a given channel for each scan is given by edge of B denotes the channel number for the relationship the transition from 8 to 7 pulses. This channel number is found to be 7.192 t), N(n) = K,At,M(n, (g-4) by equating Eqs. (5A) and @A) with m = -7, j = 0. Higher numbered channels also where Kz is a constant derived from the accumulate 7 pulses until the right edge of geometry of the cell centerpiece and lightC meets the next rotor pulse, where again limiting aperture at the condensing lens 8 pulses are accumulated. This channel and from constants relating the conversion of number 72.115 is calculated from t’he same radiant energy to counts. (Values of N(n) equations with m = 1,j = 1. The location of normalized to an average value of 1 can be successive pulses in the first scan are calcuobtained from the expression N(n) = lated in the same manner by incrementing m. Mb, t)lpl..) The pulse width and separation are constant. We will now illustrate the calculations The first pulse for the second scan occurs required to give the real-time display as seen at channel number 64.923 corresponding to on the monitor. First, for 13,000 rpm, Y, = m = 8 with j = 2. The end of this pulse is 216.667 rev/set, At, = 4.6154 X 10e3 see; calculated from m = 1, j = 1. The position P, = 7.0997 rotor pulses/scan. The real of the additional pulses can be determined time pattern of N(n) as a function of n for in the same manner, but it can be seen that three successive scans is shown at the top of the pulse width and separation are the same Fig. 10, with Kt set equal to 1. The construcas for the previous scan. The overall appeartion of these patterns is illustrated in Fig. ance of the second scan is identical to the 11. Along the horizontal axis are plotted the first, except that all pulses are moved one time, channel number, and rotor pulse pulse width to the left. number. The horizontal axes are constructed Each subsequent scan is also similar, so that channel number 0 is read at time 0 just as a rotor pulse occurs. It should be with the pulses moving the same distance to the left. From visual examination of the noted that the rotor pulse number goes from negative to positive values in passing image in real time, one would infer that each through t = 0. With increasing time the pulse travels to the left, with no evidence of channel numbers go from 0 to 512, then start the discontinuous nature of the movement. over again. To construct the scans for Fig. One can determine the time between pulses

TELEVISION

SCANNER

13,000 rpm 3.5-

0

I

n

n

n

II

I

6

n

3.0:. 8

n

n

15

n

r!

..n--

4

uuu.--vz

lJ22

u

7r

u

u

u

u

u

2’ I

n

II

--LT

J

u

lr

u32 /

z

0

100

200

300 Channel

FIG. 10. Theoretical indicate

the

rotor

real-time number.

pulse

13,000

images

for

II

I

I

-4

rotor

III

-2

0

Rotor ioo I

20,000

pulsing.

200

300

I -0.02

I

400 I, - 0.01

Pulse

III,, 2 No

11. Schematic

on

the

figure

I

I s

0

I

100 ,

I 0.04

0.03

I

rpm

-8

-6

-4

IIIlllJlIL -2

6

diagrams

P”:se

illustrating

the time required for a given number of them to traverse a point on t’he monitor. ‘Ihe pulse time can also be calculatcd from the separation of pulses on each scan, the pulse movement between successive scans and At,. After simplification one obtains F,q. (7A), with v, in place of ~1 and 1,~ measuring

numbers

6

0 100 200 300 400 Chonnei No I , I I I,, 0 0 01 0 02

Rot% Fro. image.

The

4

11111111111111111 -10

500

rpm

r------h-* 1 I -6

I -0.03

400

No

8

IO

12

14

N:

the

effect

of rotor

prdses

on

t.he

real-time

the int’cger ncsrest to the rotor pulst~s/ scan. l:or 13,000 rpm one calculates -3.04 pulses/see (the minus sign denoting that the pulse moves to the left), compared to the value of 2.8 dchcrmincd from observation of pulw~son t,hc monitor. A speed of 20,000 rpm, P, = 10.928 rotor i is

886

RICHARDS

AND

pulses per scan, gives a negative pulse traveling to the right. Three scans calculated in the manner described above are shown in the center of Fig. 10 with the rotor pulse diagram given at the bottom of Fig. 11. There are 11 rotor pulses from n = 0 (line segment A) to n = 43.25, where the left edge of B intersects rotor pulse m = -10 (j = 0). There are 10 rotor pulses until the right edge of C intersects rotor pulse m = 1 (j = 1) at channel 46.875, giving 11 pulses again. Subsequent pulse widths and separations are the same. For the second scan, the drop to 10 rotor pulses occurs at channel 0, where m = 0, j = 1. This level continues until the right edge of D intersects rotor pulse m = 11 (j = 2), where a rise to 11 rotor pulses occurs. The appearance of the second and subsequent scans is the same, except that each scan is displaced one narrow pulse width of 3.625 channels to the right. The real time display looks like a seriesof negative pulses moving to the right. From Eq. (7A) one calculates 2.36 pulses/ see,in good agreement with the value of 2.38 determined from observation of the pulses on the monitor. We have demonstrated that, for speeds giving near integer values of rotor pulses/ scan, pulses traveling to the left or right are observed. For other speeds,successivescans may differ so greatly that visual examination fails to give any meaning to the rapidly changing patterns. For example, at a speed of 12,000 rpm (bottom of Fig. 10) there are 6.554 rotor pulses per scan, with a pulse width (high region) of 43.25 channels and a pulse separation of 78.125 channels. On the second scan the pattern is shifted one pulse width to the left, and the high and low region are out of phase. But the third scan resembles the first, being displaced 8.375 channels to the left. The fourth scan resembles the second with this same displacement. From the real time display one would infer t’hat twice as many pulses were present, with a movement to the left of 8.375 channels every two scans. From these values one can calcu-

ROCKHOLT

late a movement of 3.27 pulses/see, which is difficult to see on the monitor. Other types of pulse movements can be observed for the particular speeds available with the ultracentrifuge, including multiple images at some of the speeds below 2000 rpm. The nature of these patterns will not be investigated, for, as will now be demonstrated, any error in the stored image due to transient phenomena can be reduced to any desired level by the accumulation of sufficient scans. Let us suppose that s successive scans are stored. The number of rotor revolutions that occurred during this time period is obtained from m = v,.sAt,. If m is an integer, no error due to rotor pulses is possible. If m is not an integer, there can be at most a difference of one rotor pulse that contributed to the image for all the channels. (Where in the pattern these differences occur depend on the rotor speed, the phasing between the rotor pulsing, and the start of the scanning process.) Thus for speedsabove 10,000 rpm (166.7 rev/set), 183 scans is sufhcient to reduce any error to 0.1%. For lower speeds a satisfactory number of scans is readily calculated. In this discussion, we have ignored the geometry of cell and stationary aperture at the condensing lens, as well as the image decay characteristics of the vidicon tube. Consideration of these effects would lead to the rounding of the sharp edges of the traveling pulses and diminution of their height, but would have no effect on their number or apparent rate of movement. Thus the magnitude of the error due to rotor pulsing would be less than that described here. E$ect of Alternating Current Arc Lamp and Spinning Rotor Combining the effects of rotor and light source pulsing does not lead to a simple mathematical solution. The treatment of the traveling pulses originat’ing from the spinning rotor described in the last section needs to be modified for the effect of the arc lamp.

TELEVISION

The channel locations of the maxima and minima are calculated in the same manner, but the N(n) function must be modified to allow for the differing intensity of each rotor pulse. The light intensities obtained from Eq. (2A) are summed for the rotor pulses at times given by Eq. @A), giving N(n)

= :

$ sin’ ~vlmAt7. rm “1

(lo-41

The appearance of the first four scans (superimposed one on the other) at a speed of 13,000 rpm is shown at the bottom of Fig. 9. The first maxima in the N(n) function starts at a value of 0.998 (ml = -7, m2 = 0), drops to 0.958 (ml = -6, m2 = 0), incrcascs to 1.231 (ml = -6, m2 = l), decreasesto 1.028 (ml = -5, m2 = 1 ), and so on, alternating between 8 and 7 rotor pulses. In Fig. 9, it is seen that the traveling pulses that occur during any scan are of widely differing heights. Moreover, the height of any given pulse appears to increase and decrease in a regular fashion from scan to scan. The traveling pulsesappear to move to the left with the same velocity as for the case with constant light intensity. Their amplitude, however, varies between 0.987 and 1.266, whilr the amplitude of the region bet#ween the traveling pulses varies between 0.932 and 1.040. These variations are much greater than those for rotor pulsation or light source pulsation alone. In the regions between the eight-pulse maxima, one can not find the four maxima and minima seen with light source pulsation alone. The construction and analysis of realtime patsterns for other speeds would be tedious and time consuming. However, such analyses are not necessary, for it is evident that, so long as standing waves are not seen, sufficient scans can be accumulated to reduce the error to an acceptable level. We were unable to find a simple method of calculating the desired number of scans. It is evident that higher speedsgive more rotor pulses per scan, thereby more closely ap-

887

SCANNER

proximating the case for light source pulsation alone. In practice, one would accumulate an increasing number of scans, watching for the disappearance of any suspicious waves or bumps. There is, however, an observation which makes it easier to analyze a lower nwnber of scans. From Eq. (IOA) it is seen that rotor pulses occurring during scansj = 1 through s - 1 are accumulated in all channels. The total number of counts accumulated during this time period can be calculated, or one can assume an average value of 1 for thr sincsquare function. Then the residual valuc>sof N(n) are calculated for each channel from Eq. (1OA) written twice, once for rn, taking values forj = 0 and once for m taking values forj = s. ACKNOWLEDGMENTS The authors expresstheir appreciation t.o DrM. R. Zatzick of SSRInstrumentsCo. for the loan of the OMA usedin thesestudies,personalinstruction in its use, and many discussions concerning its mating with the ultracentrifuge. Weowethanks to Hewlett-Packard for loan of the S-Y and graphic

plotters. REFERENCES

1. HANLON,

S.,

JOHNSON,

Arch.

Ii.,

Biochem.

2. LAMERS, K., AND

LAMERS, K., LAUTERIMCH, AXD SCHACHMAN, H. K.

Biophys.

Biophys.

99, 157.

PUTNEY, F.,

STEINBERG,

H. K. (1963)

SCHACHMAN,

G.,

(1962)

Arch.

I. Z., Riochewt.

103, 379.

3. CHERVF.NKA, C. A. (1971) Fractions, Division of Beckman Instruments, Palto Alto, California. 4.

SPRMG,

Ann. 5. PEKAR, AND

Spinco Inc.,

S. P., AND GOODMAN, Ii. F. (1969) S.Y. Acacl. Sci. 164, 294. A. II., WELLER, II. E;., BYERS, Ii. A., FRANK, B. H. (1971) Anal. Niochem. 42,

516. 6. CREPE~U,

H., ?vf. J.

Ii.

~IEIIMAN,

~!IDELSTEIPI‘,

(1972)

Anal.

8. J., Biochem.

.\ND

60,

213. 7. Ka~asEc, 8.

OLSON,

9.

RICHARDS, M.W,

10. JKSKINS,

G.

H.

F. W. (1972) Res./Develop. 23, Jan., (1972) A,mer. Lab. 4, Feb., 57. F,.

G.,

K.

(1971)

F.

A.,

TEI,LXR, .\NU

Anal. WHITI’:,

I).,

~\NL)

PCHAV~T-

Biochem. H.

47.

41, I<.

18!). 11957)

888

RICHARDS

AND

Fundamentals of Optics, McGraw-Hill, New York. 11. SCHACRMAN, H. K., CROPPER, L., HANLON, S., AND PUTNEY, F. (1962) Arch. Biochem. Biophys. 99, 175. 12. CRIANCONE, E., VECCHINI, P., FORLANI, L., ANTONINI, E., AND WYMAN, J. (1966) Biochim. Biophys. Acta 127, 549. 13. DAUTREVAUX, M., BOULANGER, Y., HAN, K.,

ROCKHOLT

14. 15. 16. 17.

AND BIESERTE, G. (1969) Eur. J. Biochem. 11,267. EDMUNSON, A. B. (1965) Nature (London) 206, 883. THEORELL, H. (1934) Biochem. Z. 268, 46. Bulletin E-TB-019A (1965), Spinco Division, Beckman Instruments, Inc., Palo Alto, Calif. SVENSSON, H. (1950) Acta Chem. Stand. 4,399.