- Email: [email protected]
A MUBUS DMA-Interface for CCD optical arrays
Q
ROMJCRO
EU
EUROMICRO~ournal I_
1.
5 (197% 243-248 -_-
__I___.-
it is possible In fact, blr $.aving in cI’rc:uit a large part of tt,csc
INTRDDUCTION
In recent years., optical CC0 arrays with a number of interesting applications have appeared on the market. Two examples of possible applications are a CCD line camera 11 J ard an t5 I, automated spectrophotometer 12 I In using those of camp: icatjons arise. : rrct . elements + a number complicated, ofall, several, clock signals are required. Generating these signals with spepurpose candom logic has the drawback of cial requiring a completely new design if a diffeelement is to be used. rent, or newer, Next, the analog output appears at the outout gate in a relatively short time, depending on the cIockspeed. If the application is only visualisation, direct application of the output to an oscilloscope is satisfactory. This is the case If any other for the CCD line camera L 1 1 . treatment of the output is required, such as finding a maximum, or taking account of dark then it is obvious that dfgitrl storage current, of the results is necessary.
The DMA control counters already
* **
A universal a ROM.
clock
-
A DMJ\ controller face.
drive with
Researcher at the Nationaa? Wetenschappel i j k Onderzoek. Engineer at the Bel!Jische
circuit, a
based
Fonds Radio
by
derive clock
a considst-aComhin’ng
mostly from generation.
to suppl’, [see fig.
fi9e 1).
clock
are pulses which TV-ansfcr the CI)~1 A alrd Qjx sensors ir the shiftregistatits of &e optical They occur once every measurement sc3r’Aenters. and are followed by a number of pulses ot co, <,2 which drive the the two-phase clocks 831 and shiftregisters. The outputgate signal, @R.
is
reset
by
the
When 9R is low, but @3: or 92 is zero reference lever appears at is followed by the useful video @R and either @I or $2 are both
on
fifth
cltick
still high, the output. information low.
the This when
The integration time is fixed by the interval beThe tween two sets of transfer gate pulses. number of clock cycli for the shiftregisters must be at least equal to the number of cells, but may be larger to lenghten the ini.e. 256.
voor en
will then u,;ed for
upon The user is called signals to this element
inter-
MUBUS [ 31
to realize complexity circuits.
?. 1. Uescrje&xorr __-,__ of the CCDoperation “he CCD -1eme~ir wenaveu-,ed fs the C’CDIlOF from Fairchild 141. Besides two C&D analog shift rethe liDF cc ntai ns 256 1 i aht-sensi ti ve sisters, -The alecrr-ical charge which builds up cells. in each ceil is prol’ortional to the amount of witt: almost perfect linearity light received, There is also a trhen no saturation o:curs. strongly temperature dependant dark current which contributes trS zhe charge build-up. ‘nle are interested in that part to the surput signal Fol 1 owi ng a which is due to the light input. parallel transfer of the charrJes f awards the sequential sh’ fting of the CC> shift register, charges towards the output amplifier takes place.
This paper describes how the first ;omplication can be solved in a flexible and univer-al wav by using a ROM to store the clocking sequence. A Ik X 8 bit EPROM was originally used [ 1 1 , but has now been replaced by a much smal let- snd faster memory and a few extra counters. The answer to the second problem is of course to let a microprocessor handle the signal processing. Because of the speed at which data acquisition takes place, Direct Memory A.-cess (WA) will be that twa circuits requ i red _ It therefore appeal-r must be developed to allow convrnient use of optical CCD’s: -
--._
Televisie. 243
number of outputs at least equal to the number cf required signals (see fig. I). All that is rep, red, is a counter to generate PROM addressTo reduce the number of memory words re;:;rad, the PROM is divided into two parts : one contains four Q1 : 92 pulses al;d is repeated 32 times, the other contains the transfer gate
-
The mtput of the programmable dlvidcr $5 a squsre wave at twice the desired QR rate. This divider is controlled by the mfcroprccessor through its three "select" inputs. FROM address es are derided from counters 1 and 2. Two extra signals are incorporated in the PRW to control the two sequences. One is fed to counter 4 to fix the number of repetitions of Divis?on by 32 gives the the first sequence. pulse: to enable counter 2, and thereby allows transition to the second sequence. The secctnd control signal clears all coufiters at the end of this sequence, which completes a whole measurement cycle. This reset also ensures that the circuit will always start up correctly. Since two PROM addresses are necessary to produce a single output, the three most significant bits froze counter one form tne three least Tl!o five other significant DMA address bits. bits are generated by synchrono>qly cascading counter 3 to counter 1. To bring the clock vs,ltage levels up to specification, oper collector buffers are used. Moreover, to produce the correct 8R signal an AND operat+on with tnc basic clock is necessary (this avoids doubling the neProvision can be made cessary memory space). for 512 or 1024 elesen; arrays by simply using another output of counter 4. The shaoe itself of the cTock waveforms can be adjusted by changing the contents of the PROM.
Wh::s :he difference of these twir outputs is mdj th: Wsultiny signs? st!ll CQntdins both the z
-
M. Nyssen
246
L.%. The DHA control anri f&BUS inte++ce '17iispart of the system-="%?n%-i=f be described in tems of t1a general b?u,:k diagram
et
al.
The use of MUBUS considerably reduced the tim+ needed to develop the whole system, since U-L? could be made of a range of existing MUBUScompatible modblzs. We used a microprocessor tin.itcentaining a Motorola 6800, a 4kRAM memory, an EPROlvlmodule, and the debug nlon'ftor.[6]
3.
SOFTWARE
AND SOME RESULTS
3.1. Description of the software b'ocks As an example, consider the set-uF=ig. 5a. This set-up consists of an optica System focussing the light spectrum of an arb?trary lightTo perform such specsource I_ on a CCD array. tral measurements with the aid of 2 6800 mubus microcomputer a software package was developed. The programs were structured in a number of independant blacks, which could be Z.ested sepaThe microprocessor is used ':o take rately. over as many tasks as possible from :he human operator. When an intervention, suc'l as changing the light input is necessary, ati appropriate message is issued by the softwaw. The following blocks have been developed : - A basic initiator. This unit merely sends Z Co?iimZnTi IwoXEi the CCUto initiate one exposure-readout cycle. A waiting pei,iod is included to ensure DMA i,; finished bef&e exit from this block Cakes place. - Dark_c_urs$. substraction. Here two cy217s are cari%d-o
I Au>+tlrnPti_c rp_eed selection.
This unit wi I7 ?a% '3ut a nujb%r-d?‘fia?ic crcles unti', a sp&d setLing is found which optimaliy uses tne dynamic range of the CCD. If necessary, a messaye to adjust the light input is issued.
- __c.______' ol.tVJr.of results Two different formats are available, namely :
c.i
_.(.I,._”
““l”.-l~.l.-.---
PAR?J.LEL EmY
,-
--.
I-- .._.‘.,.... “. 1 _““_“.
I”“.--_~ Fig. 5a.
._I-
outout
A MUBUS DMA Interface _II_I~_ 7‘1 I‘$ 'II '0
__I.I) .*,
3.1.1. decimal ogrtput of the complr?te memot-y conten'~s. This output contains all the information present.
.I.t4.1:. 0
3.1.2. graphScal output of all or par!. of the 'nemoi-;, contents, giving a more visria:ly striking result.
*0x ,io: 44: :'I :?'. >(I 2 , '~7 _I Z", 70: :'I,2.' ;.,,:.,. 1'0 :..,ISi>:,. 80: .' 7 , l-, I,,, PO: :.I> 100: i(, :9 i' :; I:! ,/, ... 110: I'll-,4 ,x4 :a',I,, ^:I ,?‘ 130: I,:, 7,’ ,-,’ ?.’ ~> :>? - _ : 13c.r 70 L':,:,0 I ;-<, r. 7, 140: 7lri?7 .lij:;> I.7 :, . . ', “0 l,.. 120: Llli p% :::::?I. ‘, ..* ./, 1:; ../ 1*0: ill 1,OZ x 10 ?'I ?i, :n ?'r -i, :,I 1110: It, I'0 ?L :10 ?R '0 '" ',, LPO: *c, 10 20 ?? _:* 'i;' ‘I;! T?DQS111 22 Ff” ”/ ?O ‘” 2,: , ; ,, 210: ?[i "20; 2* ;i ;; ;g 5; :,g I,' :,,
1:;I ;!“ ‘0:
3.2. Some results T?e WI'17 preserit7rere a few results obtained in a speccroscopir set-up, in which light falls on the XD through a slit followed by a prism. In this 'way, every cell is covered by a small range of waue-lenghts, from 2 to 7 nm in our case. 7';~ CCC was placed in such a way that the whole visibir part of the spectrum was reached. Two types of sources were analysed in this-way :
-
247
.‘_.z:;‘-,.:,..7”y,
%. r'.
??
,,,
‘~_,.-i,..
___-____
.I...
.2_.
_._,__:.-~_.-__.:__
Fig. 5c discharge tube. The intensities are repre,ented on a scale from 0 to 255. Fig. 5c shows a plot of part of this spectrum. In this case, dark current has not been subtracted. It is obvious from fig, 5c that the precise cell on which a narrow spectral line falls is easily identified. In this way, the cell numbers can be calibrated in terms of wave-lenghts. - Sources with a continuous spectrum, such as an incandescent lamp. 1 . plot of such a $t"~m pattlal shows When a the exact energetic spectral ou&t'of the source ;s known, the spectral sensitivity of the CCD and the optical elements used
_,,._
-
-
-,
._
I_- _..r__._.-
.--
-
-2.-x:
.,
; ‘.,,1
-_1___1
. : : : : : :. : : :. :. :. :: :;* .:::. . :. .:: . .**a . . **a .. *. .. .s ::: .**, ..*.
f
________
;.,; 1
Fig. 5b.
: : : :: .m :: ::. .. .f .. ... ... . ::: ... ::::.. . . . .. ....*
.,
Y, (/ >i, ., , :, :
;>:I 2’301 2c 70 ;o :,o .n ?O .m ,; .z; 240: .ic 70 ;:: i:5( ,R 1.4 :y) .Iti ,3 ,il zc .!I, a XG: FEAh "ILII~ i--__. ?I%6, cri_____I 111
ilonochromatic sources, with one or several sharp peaks in their spectrum. Fig. 5b represents a Tistlng of data corresponding to the spectrum of a low-pressure mercury
_ ,,*,.
:'I,
Iii :.I :,o .‘ ‘0-. .<, :I? nr :::: :.:,
,,, -n ,7<) i
M.
248
4.
frlyssen
CDMCLUSI. ON.
We Peel that the connection of a CCD optical array to a m~craprocesso r 3per!s tip many interesFirst of al?, manipulation and tin;: prosj~ects. testing of LhP CCD unit is greatly simpl ified by $hF av;tilabie software, vrnich znakkes it tt-arssparent tu the operator. In addit.‘on, the use of a microprocessor makes signal p,-ocessing pus?ji,iE, such as subtracting the dark current or lnui tip1 ying by a spectral sens.itivjQ Factor. An&he-* app’ication is autorzatinc the p-ec-ise pocjtiwing carried out by a CCD camera 111 .
et
al
ROMJCRO
EU
EUROMICRO~ournal I_
1.
5 (197% 243-248 -_-
__I___.-
it is possible In fact, blr $.aving in cI’rc:uit a large part of tt,csc
INTRDDUCTION
In recent years., optical CC0 arrays with a number of interesting applications have appeared on the market. Two examples of possible applications are a CCD line camera 11 J ard an t5 I, automated spectrophotometer 12 I In using those of camp: icatjons arise. : rrct . elements + a number complicated, ofall, several, clock signals are required. Generating these signals with spepurpose candom logic has the drawback of cial requiring a completely new design if a diffeelement is to be used. rent, or newer, Next, the analog output appears at the outout gate in a relatively short time, depending on the cIockspeed. If the application is only visualisation, direct application of the output to an oscilloscope is satisfactory. This is the case If any other for the CCD line camera L 1 1 . treatment of the output is required, such as finding a maximum, or taking account of dark then it is obvious that dfgitrl storage current, of the results is necessary.
The DMA control counters already
* **
A universal a ROM.
clock
-
A DMJ\ controller face.
drive with
Researcher at the Nationaa? Wetenschappel i j k Onderzoek. Engineer at the Bel!Jische
circuit, a
based
Fonds Radio
by
derive clock
a considst-aComhin’ng
mostly from generation.
to suppl’, [see fig.
fi9e 1).
clock
are pulses which TV-ansfcr the CI)~1 A alrd Qjx sensors ir the shiftregistatits of &e optical They occur once every measurement sc3r’Aenters. and are followed by a number of pulses ot co, <,2 which drive the the two-phase clocks 831 and shiftregisters. The outputgate signal, @R.
is
reset
by
the
When 9R is low, but @3: or 92 is zero reference lever appears at is followed by the useful video @R and either @I or $2 are both
on
fifth
cltick
still high, the output. information low.
the This when
The integration time is fixed by the interval beThe tween two sets of transfer gate pulses. number of clock cycli for the shiftregisters must be at least equal to the number of cells, but may be larger to lenghten the ini.e. 256.
voor en
will then u,;ed for
upon The user is called signals to this element
inter-
MUBUS [ 31
to realize complexity circuits.
?. 1. Uescrje&xorr __-,__ of the CCDoperation “he CCD -1eme~ir wenaveu-,ed fs the C’CDIlOF from Fairchild 141. Besides two C&D analog shift rethe liDF cc ntai ns 256 1 i aht-sensi ti ve sisters, -The alecrr-ical charge which builds up cells. in each ceil is prol’ortional to the amount of witt: almost perfect linearity light received, There is also a trhen no saturation o:curs. strongly temperature dependant dark current which contributes trS zhe charge build-up. ‘nle are interested in that part to the surput signal Fol 1 owi ng a which is due to the light input. parallel transfer of the charrJes f awards the sequential sh’ fting of the CC> shift register, charges towards the output amplifier takes place.
This paper describes how the first ;omplication can be solved in a flexible and univer-al wav by using a ROM to store the clocking sequence. A Ik X 8 bit EPROM was originally used [ 1 1 , but has now been replaced by a much smal let- snd faster memory and a few extra counters. The answer to the second problem is of course to let a microprocessor handle the signal processing. Because of the speed at which data acquisition takes place, Direct Memory A.-cess (WA) will be that twa circuits requ i red _ It therefore appeal-r must be developed to allow convrnient use of optical CCD’s: -
--._
Televisie. 243
number of outputs at least equal to the number cf required signals (see fig. I). All that is rep, red, is a counter to generate PROM addressTo reduce the number of memory words re;:;rad, the PROM is divided into two parts : one contains four Q1 : 92 pulses al;d is repeated 32 times, the other contains the transfer gate
-
The mtput of the programmable dlvidcr $5 a squsre wave at twice the desired QR rate. This divider is controlled by the mfcroprccessor through its three "select" inputs. FROM address es are derided from counters 1 and 2. Two extra signals are incorporated in the PRW to control the two sequences. One is fed to counter 4 to fix the number of repetitions of Divis?on by 32 gives the the first sequence. pulse: to enable counter 2, and thereby allows transition to the second sequence. The secctnd control signal clears all coufiters at the end of this sequence, which completes a whole measurement cycle. This reset also ensures that the circuit will always start up correctly. Since two PROM addresses are necessary to produce a single output, the three most significant bits froze counter one form tne three least Tl!o five other significant DMA address bits. bits are generated by synchrono>qly cascading counter 3 to counter 1. To bring the clock vs,ltage levels up to specification, oper collector buffers are used. Moreover, to produce the correct 8R signal an AND operat+on with tnc basic clock is necessary (this avoids doubling the neProvision can be made cessary memory space). for 512 or 1024 elesen; arrays by simply using another output of counter 4. The shaoe itself of the cTock waveforms can be adjusted by changing the contents of the PROM.
Wh::s :he difference of these twir outputs is mdj th: Wsultiny signs? st!ll CQntdins both the z
-
M. Nyssen
246
L.%. The DHA control anri f&BUS inte++ce '17iispart of the system-="%?n%-i=f be described in tems of t1a general b?u,:k diagram
et
al.
The use of MUBUS considerably reduced the tim+ needed to develop the whole system, since U-L? could be made of a range of existing MUBUScompatible modblzs. We used a microprocessor tin.itcentaining a Motorola 6800, a 4kRAM memory, an EPROlvlmodule, and the debug nlon'ftor.[6]
3.
SOFTWARE
AND SOME RESULTS
3.1. Description of the software b'ocks As an example, consider the set-uF=ig. 5a. This set-up consists of an optica System focussing the light spectrum of an arb?trary lightTo perform such specsource I_ on a CCD array. tral measurements with the aid of 2 6800 mubus microcomputer a software package was developed. The programs were structured in a number of independant blacks, which could be Z.ested sepaThe microprocessor is used ':o take rately. over as many tasks as possible from :he human operator. When an intervention, suc'l as changing the light input is necessary, ati appropriate message is issued by the softwaw. The following blocks have been developed : - A basic initiator. This unit merely sends Z Co?iimZnTi IwoXEi the CCUto initiate one exposure-readout cycle. A waiting pei,iod is included to ensure DMA i,; finished bef&e exit from this block Cakes place. - Dark_c_urs$. substraction. Here two cy217s are cari%d-o
I Au>+tlrnPti_c rp_eed selection.
This unit wi I7 ?a% '3ut a nujb%r-d?‘fia?ic crcles unti', a sp&d setLing is found which optimaliy uses tne dynamic range of the CCD. If necessary, a messaye to adjust the light input is issued.
- __c.______' ol.tVJr.of results Two different formats are available, namely :
c.i
_.(.I,._”
““l”.-l~.l.-.---
PAR?J.LEL EmY
,-
--.
I-- .._.‘.,.... “. 1 _““_“.
I”“.--_~ Fig. 5a.
._I-
outout
A MUBUS DMA Interface _II_I~_ 7‘1 I‘$ 'II '0
__I.I) .*,
3.1.1. decimal ogrtput of the complr?te memot-y conten'~s. This output contains all the information present.
.I.t4.1:. 0
3.1.2. graphScal output of all or par!. of the 'nemoi-;, contents, giving a more visria:ly striking result.
*0x ,io: 44: :'I :?'. >(I 2 , '~7 _I Z", 70: :'I,2.' ;.,,:.,. 1'0 :..,ISi>:,. 80: .'
1:;I ;!“ ‘0:
3.2. Some results T?e WI'17 preserit7rere a few results obtained in a speccroscopir set-up, in which light falls on the XD through a slit followed by a prism. In this 'way, every cell is covered by a small range of waue-lenghts, from 2 to 7 nm in our case. 7';~ CCC was placed in such a way that the whole visibir part of the spectrum was reached. Two types of sources were analysed in this-way :
-
247
.‘_.z:;‘-,.:,..7”y,
%. r'.
??
,,,
‘~_,.-i,..
___-____
.I...
.2_.
_._,__:.-~_.-__.:__
Fig. 5c discharge tube. The intensities are repre,ented on a scale from 0 to 255. Fig. 5c shows a plot of part of this spectrum. In this case, dark current has not been subtracted. It is obvious from fig, 5c that the precise cell on which a narrow spectral line falls is easily identified. In this way, the cell numbers can be calibrated in terms of wave-lenghts. - Sources with a continuous spectrum, such as an incandescent lamp. 1 . plot of such a $t"~m pattlal shows When a the exact energetic spectral ou&t'of the source ;s known, the spectral sensitivity of the CCD and the optical elements used
_,,._
-
-
-,
._
I_- _..r__._.-
.--
-
-2.-x:
.,
; ‘.,,1
-_1___1
. : : : : : :. : : :. :. :. :: :;* .:::. . :. .:: . .**a . . **a .. *. .. .s ::: .**, ..*.
f
________
;.,; 1
Fig. 5b.
: : : :: .m :: ::. .. .f .. ... ... . ::: ... ::::.. . . . .. ....*
.,
Y, (/ >i, ., , :, :
;>:I 2’301 2c 70 ;o :,o .n ?O .m ,; .z; 240: .ic 70 ;:: i:5( ,R 1.4 :y) .Iti ,3 ,il zc .!I, a XG: FEAh "ILII~ i--__. ?I%6, cri_____I 111
ilonochromatic sources, with one or several sharp peaks in their spectrum. Fig. 5b represents a Tistlng of data corresponding to the spectrum of a low-pressure mercury
_ ,,*,.
:'I,
Iii :.I :,o .‘ ‘0-. .<, :I? nr :::: :.:,
,,, -n ,7<) i
M.
248
4.
frlyssen
CDMCLUSI. ON.
We Peel that the connection of a CCD optical array to a m~craprocesso r 3per!s tip many interesFirst of al?, manipulation and tin;: prosj~ects. testing of LhP CCD unit is greatly simpl ified by $hF av;tilabie software, vrnich znakkes it tt-arssparent tu the operator. In addit.‘on, the use of a microprocessor makes signal p,-ocessing pus?ji,iE, such as subtracting the dark current or lnui tip1 ying by a spectral sens.itivjQ Factor. An&he-* app’ication is autorzatinc the p-ec-ise pocjtiwing carried out by a CCD camera 111 .
et
al