Home
Search
Collections
Journals
About
Contact us
My IOPscience
Conversion of a multichannel analyser into an analogue signal averager
This content has been downloaded from IOPscience. Please scroll down to see the full text. 1975 J. Phys. E: Sci. Instrum. 8 915 (http://iopscience.iop.org/0022-3735/8/11/012) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 129.78.139.28 This content was downloaded on 04/10/2015 at 06:49
Please note that terms and conditions apply.
Conversion of a multichannel analyser into an analogue signal averager
L J Chapman? and 0 M Williams‘; f Department of Physics, University of New England, Armidale, New South Wales, Australia Department of Engineering Physics, The Australian National University, Canberra, ACT, Australia
Received 30 May 1975 Abstract An analogue to digital (AD) converter is described which may be used to transform a multichannel analyser into an analogue signal averager suitable for analysis of analogue signals in the frequency range from DC to typically 40 kHz and signal amplitudes in the millivolt to volt range. The AD converter operates on a triggered basis and is capable of sampling an analogue signal at a maximum rate of 100 kHz while maintaining a conversion linearity of better than 1% and voltage resolution equal to 2.5 % of the maximum permissible input signal level. Considerably better voltage resolution is obtained by averaging over several repetitions of the input signal and also at smaller sampling rates where more time is available for digitizing the analogue signal between successive samples. Operation of the signal averaging system has been extended to the range of tens of gigahertz by averaging suitably the samples recorded by a sampling oscilloscope.
1 Introduction Although multichannel analysers have been used widely in recent years for digital signal analysis, their versatility as general purpose laboratory instruments is limited in some current models by the lack of provision of an analogue signal averaging option. Other manufacturers include such an option by suitably rerouting the signal from the AD converter to the multiscaling section of the analyser. However, since the design of the AD converter is optimized for use in pulse height analysis operations, such a modification fails in general to exploit the full capabilities of the multichannel analyser. Fortunately, the rapid advances in integrated circuit technology have enabled this limitation to be overcome by reducing considerably the complexity of design of an independent AD converter suitable for transforming a multichannel analyser into a versatile analogue signal averager. In this paper we describe an AD converter which may be used to extend the operation of a multichannel analyser to analogue signal averaging applications over a wide range of signal frequencies and amplitudes. Several alternative methods are available for converting an analogue signal into digital form (see, for example, Hoeschele 1968). For the present application the choice is limited by two
requirements; that the AD converter be capable of operating at high speed in order that adequate temporal resolution of high frequency analogue signals be obtained, and that the design be sufficiently simple to enable construction within the average laboratory, We have chosen to develop a voltage-to-time AD converter satisfying these requirements. For signal averaging applications the voltage-to-time AD converter is coupled to a multichannel analyser operated in multiscaling mode and is triggered each time a new channel is addressed. A gating pulse of width proportional to the input signal level is then generated and AD conversion is achieved by allowing oscillator pulses of constant frequency to be delivered to the multiscaler channel only during the gating period. The channels are addressed successively at a constant preset rate and when the multiscaling sweep is synchronized to the occurrence of the analogue signal, averaging of that signal occurs over successive sweeps. Although the 0.1-1 %conversion accuracy typical of voltageto-time AD converters is less than that obtained by alternative methods (Hoeschele 1968), it is still quite adequate for most signal averaging applications and, furthermore, any disadvantages caused by the lesser precision are offset by the advantages of speed of operation and simplicity of design. Conversion speed is limited by the deadtime of the AD converter and also by the maximum rate at which oscillator pulses may be accepted by the multiscaler. The AD converter described in this paper is characterized by 6 ps deadtime and when coupled to a multiscaler capable of counting at 10 MHz, digitizes an analogue voltage within a period of (6+ 0-1N) ps, where N is the number of oscillator pulses delivered to the multiscaler. Conversion accuracy is better than 1 %. It is thus possible to sample an incoming analogue signal at a rate as high as 100 kHz while maintaining a voltage resolution of 2.5 % of the maximum permissible signal level. Considerably better voltage resolution of an input signal is obtained when smaller sampling rates are used since the lower temporal resolution required lessens the restriction on the maximum number of oscillator pulses that may be transmitted to the multiscaler between successive samples. The signal averaging system has been used for analysis of a wide variety of analogue signals over a frequency range extending from DC to typically 40 kHz and for signal amplitudes in the millivolt to volt range. Above 40kHz, the temporal resolution is limited by the finite deadtime. However, this restriction may be overcome simply for repetitive signals by using a sampling oscilloscope to view the analogue signal and averaging suitably the successive samples recorded by the oscilloscope. When this method is applied, the frequency range is extended to tens of gigahertz and thus the signal averaging system described in this paper is capable of analysing most common repetitive laboratory signals. 2 Circuit description A block diagram of the AD converter is shown in figure 1 and the principal waveforms of the timing sequence are shown in figure 2. The AD converter is triggered by a positive 3 V pulse each time a new multiscaler channel is addressed and samples the input analogue signal during a 1 ps aperture period. The hold voltage developed by the sample and hold circuit is applied to one input of the ‘stop’ comparator. Termination of the aperture period initiates a ramp voltage which rises at a constant rate from a slightly negative quiescent level. When the ramp voltage passes through zero volts, the ‘start’ comparator is triggered and opens the delay gate. The trailing edge of this gate opens the output gate allowing 10 MHz oscillator pulses to pass to the multiscaler channel. The ramp voltage continues to rise until the hold voltage level is reached, whence the ‘stop’ 915
L J Chapman and 0 M Williams
Sample and hold
An-
in
I
Hojd stop comparator
Sample gate
4 -
U
oscillator
Gated
Open Output gate [generator Reset
IOW
U
gate generutor
comparator
~
Delay gate generator
control
, gate Ramp " 5
--Lr--v A
40v
10
Ramp 0'4v] 0
n
De lay gate
0
out
Reset
Figure 1 Block diagram of the AD converter
gate
comparator is triggered and a reset gate is generated. The output gate is closed and since oscillator pulses pass to the multiscaler only during the gating period, a linear AD conversion is achieved. Further operation is inhibited during the reset gating period while quiescent conditions are restored. Both the AD converter described in this paper and the accompanying 10 MHz oscillator have been constructed on a single 5.5 in x 8 in printed circuit board using standard integrated circuit components. The circuit diagrams of the AD converter and the oscillator are shown in figures 3 and 4 respectively. Standard power supply levels as specified by the manufacturers are used for each integrated circuit, The circuit designs follow standard practice and only those features governing the operating characteristics of the AD converter are discussed below.
Gated 10 MHz
0 5v
0
I
0
I
"
'
Time (ps)
I
10
'
Figure 2 Timing sequence of the m:converter 10 mV/oscillator pulse; 100 kHz triggering rate
The AD converter shown in figure 3 has been designed's0 that the deadtime may be minimized subject to the requirements that adequate conversion linearity be achieved and undue design complication be avoided. The deadtime is determined by several factors: the ICI sample acquisition time of several
-15 V
Delay g a t e generator 'I ItI
150 p IO k
10 k
3 4 5 9 IO II 14
Figure 3 Circuit diagram. ICI,LH0043C sample and hold (National); ICZ6 , pA741 operational amplifier; 1c37, pA710 differential comparator; 1 c 4 5 8, pL9941 monostable multivibrator; 1c9 10, SN7472 JK Master-Slave flip-flop. 916
Unlabelled ICS are SN7400 quad NAND gates. Resistances in ohms ; unlabelled capacitances in microfarads. CI,Philips air dielectric concentric trimmer; Cz, 0,0022 pF (10 mV/ oscillator pulse)
Conversion of a multichannel analyser
Analogue signal Experiment
Ground
1.8 k
1.8 k
0
/Gated IO MHz
i
Trigger
C h o advance
3 Performance 3.1 Direct operation The AD converter may be combined with a multichannel analyser to form an analogue signal averager, as shown in figure 5(a). In this mode of operation the analyser is operated in multiscaling mode and is triggered in synchronism with the analogue signal. The multiscaler channels are then addressed successively at a constant preset rate. The input signal is sampled by the AD converter each time a new channel is addressed and the 10 MHz pulse train is transmitted to the particular channel addressed. Quiescent conditions are restored on completion of the sweep which is triggered each time the analogue signal is repeated. Signal averaging thus occurs as successive sweeps are added.
4*
n Multichannel n e l ~
~
l
scaler
Figure 4 10 MHz oscillator circuit constructed from a single SN7400 quad NAND gate. L has been constructed by winding 25 turns on a Neosid type F assembly. This circuit may be modified simply for crystal control by replacing the series LC combination by a suitable crystal microseconds, the time required for the ramp voltage to rise to zero volts, nonlinearity on the ramp voltage, and the time required to restore quiescent conditions. Nonlinearity is caused by switching transients and has been minimized in the present circuit by isolating the ramp generator from the switching components where possible, again within the constraint of retaining a basic simplicity of design. Thus, the delay gate has been introduced to ensure that the output gate remains closed during the period of the switching transient associated with the change of state of the 'start' comparator, IC?. Although nonlinearity caused by the transient associated with the trailing edge of the delay gate cannot be eliminated completely, the effect is not as severe as that caused by the 'start' comparator because of the less direct coupling between the delay gate generator ICS and the ramp generator ICB. A simple means of avoiding the nonlinearity in signal averaging applications is discussed below in 53.1. The delayed opening of the output gate may be balanced in the present circuit by offsetting the hold voltage by suitable adjustment of the 10 k!J potentiometer connected across pins 3 and 4 of ICI.If required, the offset level may be set so that the output gate is on the verge of opening when the analogue input terminals are grounded. The quiescent level of the gated ramp voltage is adjusted to typically -20 mV which is sufficiently negative to avoid spurious triggering by switching transients. The ramp is gated by IQ and rises at a rate set by the capacitor CZ.This has been chosen for operation in two ranges in the present circuit: at 0.0022 pF for rapidly varying signals, giving an AD conversion constant of 10 mV/oscillator pulse, and at 0.022 pF for slower signals (1 mvloscillator pulse). In the latter case the circuit deadtime is increased to 10 p s because of the slower rise of the ramp towards zero volts and also because of the longer period for quiescent conditions to be restored. The increase is not restrictive, however, since for the slower range of operation less temporal resolution is required thus increasing the available analysis time between successive samples.
AD converter
Trigger
Sweep trigger
Channel advance
Multichannel
Figure 5 Analogue signal averaging system. (a) Direct operation, (b) indirect operation The AD converter has been tested widely in combination with a Hewlett-Packard model 5401A multichannel analyser. This analyser is characterized by a 10 MHz maximum counting rate and a deadtime of 2.2 ps between successive multiscaler channels. Operation at the maximum AD converter triggering rate of 100 kHz thus almost matches the multiscaling capabilities of the analyser. The results of averaging a noisy 0.25 V analogue signal at the 100 kHz triggering rate are shown in figure 6. On a single shot basis, this signal was severely distorted
L
.------,*
I, 0
I
I
I
2
4
6
I
8
IO
Tiwe !ms)
Figure 6 Example of direct signal averaging at 100 kHz triggering rate (10 &channel). 10 mV/oscillator pulse; 87 Hz signal repetition frequency by 50 Hz mains hum. It is clear from figure 6 that the hum has been averaged satisfactorily over 20 sweeps and that considerable improvement in signal quality has been obtained after a 2 min accumulation period (10 000 sweeps). The rapid rise and rather slower decay associated with the signal illustrated are reproduced faithfully by the signal averaging system, indicating satisfactory high frequency response. Conversion linearity has been measured by recording the number of oscillator pulses from the AD converter as a function of input voltage. Differential linearity of better than 1% is 917
L J Chapman and 0 M Williams Maximum range a t
1.57
/
/i
- -Maximum range a t 100 kHz triggering rate
0
,v:
5p
(IO ps/channel)
Oscillator pulses
,
l5
'PO 1'6 (PSI
,
50
Time after triggering
Figure 7 AD conversion linearity. 10 mV/oscillator pulse; 50 kHz triggering rate
obtained for all but the first microsecond of oscillator pulse transmission, as is evident in figure 7. The initial nonlinearity has been traced to distortion of the ramp voltage caused by the switching transient associated with the trailing edge ofthe delay gate, as discussed above in $2. The distortion has been minimized in the present design within the limitation of retaining simplicity and in fact may be avoided in signal averaging experiments by offsetting the hold voltage to a level where ten oscillator pulses are transmitted to the multiscaler when the AD converter analogue input is grounded. Positive analogue signals will then be converted to digital form according to the linear part of figure 7 beyond the first ten oscillator pulses. The above procedure increases the system deadtime to an effective value of 6 ps so that at the maximum triggering rate of 100 kHz only 4 ps are available for distortion-free oscillator pulse transmission. Voltage resolution on a single sweep basis is thus reduced slightly to 2.5% of the maximum signal level (0.4 V maximum at 100 kHz triggering rate). Considerably better voltage resolution is of course obtained by averaging over many sweeps and also at slower rates of operation where more analysis time is available between successive samples of the input signal. The presence of a significant RMS noise component on an input analogue signal, as typified in figure 6, precludes the need to consider the effect of quantum error in the AD conversion process. However, quantum error can cause considerable distortion at low voltages when clean signals such as obtained from laboratory oscillators and pulse generators are digitized and averaged. This effect has been reduced in the present system by adjustment of the oscillator frequency to a value different from the nominal lOMHz, sufficient to avoid synchronism with the analyser channel advance frequency. When the two frequencies are synchronized such that a constant delay occurs between the opening of the output gate and transmission of the first oscillator pulse, then an input signal level corresponding to, say, 4.4 oscillator pulses will always be rounded to 4 oscillator pulses. A ramp voltage of small amplitude will thus be transformed into a staircase voltage where each step represents a single quantum jump. On the other hand, lack of synchronism ensures that an input analogue level equivalent to 4.4 oscillator pulses may be converted to either 4 pulses or 5 pulses according to the 918
instantaneous delay between the opening of the output gate and transmission of the first pulse. A more accurate representation of the input signal level is thereby obtained upon averaging over several sweeps. Unfortunately, quantum error distortion is never fully overcome because of the finite width of the oscillator pulses and the finite risetime and falltime of the output gate. These effects cause oscillator pulses of reduced width and amplitude to be delivered to the multiscaler when the oscillator pulse almost coincides with the opening or closure of the output gate. At the 10 MHz maximum counting rate, such pulses may not be recognized by the multiscaler and nonrandomness thus introduced. Some care should therefore be exercised when clean signals of low amplitude are averaged and it may be necessary to mix the signal with a white noise voltage in order to achieve distortion-free averaging. The system described here has been utilized for averaging a wide variety of analogue signals from both experimental equipment and standard laboratory instruments. In particular, the system has been used for studying the temporal growth of ionization current in gaseous prebreakdown experiments in which metastable particles have a major effect in determining the rate of ionization growth (Haydon and Williams 1975). The characteristic signal observed in such experiments has been shown in figure 6 in order to illustrate operation of the signal averaging system, The system has also been used on a single sweep basis for producing digital records of analogue signals. 3.2 Indirect opevation The maximum temporal resolution of the signal averaging system shown in figure 5(a) is determined jointly by the deadtime of the AD converter and the maximum rate at which oscillator pulses may be counted by the multiscaler. When the AD converter is coupled to a 10 MHz multiscaler, the minimum time between successive samples is limited to typically 10 ps. For slower multiscalers, the oscillator frequency must be reduced accordingly and voltage resolution can be retained only by increasing the time interval between successive samples, Both the maximum temporal resolution and the range of analogue signal frequencies suitable for analysis are thereby reduced. Direct signal averaging is suitable for analogue signals up to a maximum frequency of typically 40 kHz, inadequate temporal resolution being obtained at higher frequencies. However, it is possible to overcome this limitation for repetitive signals by operating the averaging system indirectly in conjunction with a sampling oscilloscope, as shown in figure 5(b). In this mode of operation, the sampling oscilloscope is used in normal fashion, taking a single sample from each repetition of the input analogue signal and storing that sample until the input signal is repeated. The multichannel analyser is again operated in multiscaling mode and is triggered in synchronism with the oscilloscope sweep. Following each sample, the oscilloscope strobe signal triggers the AD converter and advances the multiscaler to the next channel. The sample voltage stored by the oscilloscope is digitized by the AD converter and the resulting counts are stored in the particular channel memory. Hence, as the sampling oscilloscope sweep progresses, the sample information is effectively transferred to successive channels of the multiscaler memory. Quiescent conditions are restored at the end of the multiscaler sweep and signal averaging occurs over successive sweeps. Temporal calibration of the multiscaler signal is obtained simply from the oscilloscope calibration. Indirect signal averaging utilizing a Hewlett-Packard model 181A oscilloscope with type 1810A sampling plug-in unit is
Conversion of a multichannel analyser
References Haydon S C and Williams 0 M 1975 J. Phys. D: Appl. Phys. submitted for publication Hoeschele D F Jr 1968 Analog-to-DigitallDigital-to-Analog Conversion Techniques (New York: Wiley)
i .e ..
I
I
I
I
Time (ns)
Figure 8 Example of indirect signal averaging. 750 Hz signal repetition frequency; 0.1 nslchannel
illustrated in figure 8. The signal shown here represents the output of a Tektronix type 2601 pulse generator at minimum amplitude with a deliberate cable mismatch introduced between pulse generator and oscilloscope. The sample advance time was set at 0.1 ns. It is evident that a significant improvement in signal quality has been obtained by averaging over 100 sweeps, many more features being clearly defined in the averaged signal. The temporal response is determined by the aperture time of the sampling oscilloscope and for current models allows indirect signal averaging experiments to be conducted for repetitive signals to within the frequency range of tens of gigahertz. There can, however, be some restriction at low pulse repetition frequencies since only one sample is taken for each repetition of the signal voltage and long accumulation periods may be required. Conversion linearity is determined jointly by the sampling oscilloscope and the AD converter and it may be necessary for some oscilloscopes to suitably delay operation of the AD converter until the oscilloscope sample is fully established. 4 Conclusions The AD converter described in this paper has enabled a multichannel analyser to be employed as an analogue signal averager over the complete range of signal frequencies commonly encountered in the laboratory environment. The AD converter thus effectively transforms the multichannel analyser from a specialized digital instrument into a general purpose laboratory tool suitable for both digital storage of analogue information and analogue signal averaging applications.
Acknowledgments This work was conducted at the University of New England and supported by the Australian Research Grants Committee. We wish to thank Associate Professor W J Sandle for the suggestion that high frequency signal averaging may be conducted in conjunction with a sampling oscilloscope.
Journal of Physics E: Scientific Instruments 1975 Volume 8 Printed in Great Britain 0 1975 91 9
Search
Collections
Journals
About
Contact us
My IOPscience
Conversion of a multichannel analyser into an analogue signal averager
This content has been downloaded from IOPscience. Please scroll down to see the full text. 1975 J. Phys. E: Sci. Instrum. 8 915 (http://iopscience.iop.org/0022-3735/8/11/012) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 129.78.139.28 This content was downloaded on 04/10/2015 at 06:49
Please note that terms and conditions apply.
Conversion of a multichannel analyser into an analogue signal averager
L J Chapman? and 0 M Williams‘; f Department of Physics, University of New England, Armidale, New South Wales, Australia Department of Engineering Physics, The Australian National University, Canberra, ACT, Australia
Received 30 May 1975 Abstract An analogue to digital (AD) converter is described which may be used to transform a multichannel analyser into an analogue signal averager suitable for analysis of analogue signals in the frequency range from DC to typically 40 kHz and signal amplitudes in the millivolt to volt range. The AD converter operates on a triggered basis and is capable of sampling an analogue signal at a maximum rate of 100 kHz while maintaining a conversion linearity of better than 1% and voltage resolution equal to 2.5 % of the maximum permissible input signal level. Considerably better voltage resolution is obtained by averaging over several repetitions of the input signal and also at smaller sampling rates where more time is available for digitizing the analogue signal between successive samples. Operation of the signal averaging system has been extended to the range of tens of gigahertz by averaging suitably the samples recorded by a sampling oscilloscope.
1 Introduction Although multichannel analysers have been used widely in recent years for digital signal analysis, their versatility as general purpose laboratory instruments is limited in some current models by the lack of provision of an analogue signal averaging option. Other manufacturers include such an option by suitably rerouting the signal from the AD converter to the multiscaling section of the analyser. However, since the design of the AD converter is optimized for use in pulse height analysis operations, such a modification fails in general to exploit the full capabilities of the multichannel analyser. Fortunately, the rapid advances in integrated circuit technology have enabled this limitation to be overcome by reducing considerably the complexity of design of an independent AD converter suitable for transforming a multichannel analyser into a versatile analogue signal averager. In this paper we describe an AD converter which may be used to extend the operation of a multichannel analyser to analogue signal averaging applications over a wide range of signal frequencies and amplitudes. Several alternative methods are available for converting an analogue signal into digital form (see, for example, Hoeschele 1968). For the present application the choice is limited by two
requirements; that the AD converter be capable of operating at high speed in order that adequate temporal resolution of high frequency analogue signals be obtained, and that the design be sufficiently simple to enable construction within the average laboratory, We have chosen to develop a voltage-to-time AD converter satisfying these requirements. For signal averaging applications the voltage-to-time AD converter is coupled to a multichannel analyser operated in multiscaling mode and is triggered each time a new channel is addressed. A gating pulse of width proportional to the input signal level is then generated and AD conversion is achieved by allowing oscillator pulses of constant frequency to be delivered to the multiscaler channel only during the gating period. The channels are addressed successively at a constant preset rate and when the multiscaling sweep is synchronized to the occurrence of the analogue signal, averaging of that signal occurs over successive sweeps. Although the 0.1-1 %conversion accuracy typical of voltageto-time AD converters is less than that obtained by alternative methods (Hoeschele 1968), it is still quite adequate for most signal averaging applications and, furthermore, any disadvantages caused by the lesser precision are offset by the advantages of speed of operation and simplicity of design. Conversion speed is limited by the deadtime of the AD converter and also by the maximum rate at which oscillator pulses may be accepted by the multiscaler. The AD converter described in this paper is characterized by 6 ps deadtime and when coupled to a multiscaler capable of counting at 10 MHz, digitizes an analogue voltage within a period of (6+ 0-1N) ps, where N is the number of oscillator pulses delivered to the multiscaler. Conversion accuracy is better than 1 %. It is thus possible to sample an incoming analogue signal at a rate as high as 100 kHz while maintaining a voltage resolution of 2.5 % of the maximum permissible signal level. Considerably better voltage resolution of an input signal is obtained when smaller sampling rates are used since the lower temporal resolution required lessens the restriction on the maximum number of oscillator pulses that may be transmitted to the multiscaler between successive samples. The signal averaging system has been used for analysis of a wide variety of analogue signals over a frequency range extending from DC to typically 40 kHz and for signal amplitudes in the millivolt to volt range. Above 40kHz, the temporal resolution is limited by the finite deadtime. However, this restriction may be overcome simply for repetitive signals by using a sampling oscilloscope to view the analogue signal and averaging suitably the successive samples recorded by the oscilloscope. When this method is applied, the frequency range is extended to tens of gigahertz and thus the signal averaging system described in this paper is capable of analysing most common repetitive laboratory signals. 2 Circuit description A block diagram of the AD converter is shown in figure 1 and the principal waveforms of the timing sequence are shown in figure 2. The AD converter is triggered by a positive 3 V pulse each time a new multiscaler channel is addressed and samples the input analogue signal during a 1 ps aperture period. The hold voltage developed by the sample and hold circuit is applied to one input of the ‘stop’ comparator. Termination of the aperture period initiates a ramp voltage which rises at a constant rate from a slightly negative quiescent level. When the ramp voltage passes through zero volts, the ‘start’ comparator is triggered and opens the delay gate. The trailing edge of this gate opens the output gate allowing 10 MHz oscillator pulses to pass to the multiscaler channel. The ramp voltage continues to rise until the hold voltage level is reached, whence the ‘stop’ 915
L J Chapman and 0 M Williams
Sample and hold
An-
in
I
Hojd stop comparator
Sample gate
4 -
U
oscillator
Gated
Open Output gate [generator Reset
IOW
U
gate generutor
comparator
~
Delay gate generator
control
, gate Ramp " 5
--Lr--v A
40v
10
Ramp 0'4v] 0
n
De lay gate
0
out
Reset
Figure 1 Block diagram of the AD converter
gate
comparator is triggered and a reset gate is generated. The output gate is closed and since oscillator pulses pass to the multiscaler only during the gating period, a linear AD conversion is achieved. Further operation is inhibited during the reset gating period while quiescent conditions are restored. Both the AD converter described in this paper and the accompanying 10 MHz oscillator have been constructed on a single 5.5 in x 8 in printed circuit board using standard integrated circuit components. The circuit diagrams of the AD converter and the oscillator are shown in figures 3 and 4 respectively. Standard power supply levels as specified by the manufacturers are used for each integrated circuit, The circuit designs follow standard practice and only those features governing the operating characteristics of the AD converter are discussed below.
Gated 10 MHz
0 5v
0
I
0
I
"
'
Time (ps)
I
10
'
Figure 2 Timing sequence of the m:converter 10 mV/oscillator pulse; 100 kHz triggering rate
The AD converter shown in figure 3 has been designed's0 that the deadtime may be minimized subject to the requirements that adequate conversion linearity be achieved and undue design complication be avoided. The deadtime is determined by several factors: the ICI sample acquisition time of several
-15 V
Delay g a t e generator 'I ItI
150 p IO k
10 k
3 4 5 9 IO II 14
Figure 3 Circuit diagram. ICI,LH0043C sample and hold (National); ICZ6 , pA741 operational amplifier; 1c37, pA710 differential comparator; 1 c 4 5 8, pL9941 monostable multivibrator; 1c9 10, SN7472 JK Master-Slave flip-flop. 916
Unlabelled ICS are SN7400 quad NAND gates. Resistances in ohms ; unlabelled capacitances in microfarads. CI,Philips air dielectric concentric trimmer; Cz, 0,0022 pF (10 mV/ oscillator pulse)
Conversion of a multichannel analyser
Analogue signal Experiment
Ground
1.8 k
1.8 k
0
/Gated IO MHz
i
Trigger
C h o advance
3 Performance 3.1 Direct operation The AD converter may be combined with a multichannel analyser to form an analogue signal averager, as shown in figure 5(a). In this mode of operation the analyser is operated in multiscaling mode and is triggered in synchronism with the analogue signal. The multiscaler channels are then addressed successively at a constant preset rate. The input signal is sampled by the AD converter each time a new channel is addressed and the 10 MHz pulse train is transmitted to the particular channel addressed. Quiescent conditions are restored on completion of the sweep which is triggered each time the analogue signal is repeated. Signal averaging thus occurs as successive sweeps are added.
4*
n Multichannel n e l ~
~
l
scaler
Figure 4 10 MHz oscillator circuit constructed from a single SN7400 quad NAND gate. L has been constructed by winding 25 turns on a Neosid type F assembly. This circuit may be modified simply for crystal control by replacing the series LC combination by a suitable crystal microseconds, the time required for the ramp voltage to rise to zero volts, nonlinearity on the ramp voltage, and the time required to restore quiescent conditions. Nonlinearity is caused by switching transients and has been minimized in the present circuit by isolating the ramp generator from the switching components where possible, again within the constraint of retaining a basic simplicity of design. Thus, the delay gate has been introduced to ensure that the output gate remains closed during the period of the switching transient associated with the change of state of the 'start' comparator, IC?. Although nonlinearity caused by the transient associated with the trailing edge of the delay gate cannot be eliminated completely, the effect is not as severe as that caused by the 'start' comparator because of the less direct coupling between the delay gate generator ICS and the ramp generator ICB. A simple means of avoiding the nonlinearity in signal averaging applications is discussed below in 53.1. The delayed opening of the output gate may be balanced in the present circuit by offsetting the hold voltage by suitable adjustment of the 10 k!J potentiometer connected across pins 3 and 4 of ICI.If required, the offset level may be set so that the output gate is on the verge of opening when the analogue input terminals are grounded. The quiescent level of the gated ramp voltage is adjusted to typically -20 mV which is sufficiently negative to avoid spurious triggering by switching transients. The ramp is gated by IQ and rises at a rate set by the capacitor CZ.This has been chosen for operation in two ranges in the present circuit: at 0.0022 pF for rapidly varying signals, giving an AD conversion constant of 10 mV/oscillator pulse, and at 0.022 pF for slower signals (1 mvloscillator pulse). In the latter case the circuit deadtime is increased to 10 p s because of the slower rise of the ramp towards zero volts and also because of the longer period for quiescent conditions to be restored. The increase is not restrictive, however, since for the slower range of operation less temporal resolution is required thus increasing the available analysis time between successive samples.
AD converter
Trigger
Sweep trigger
Channel advance
Multichannel
Figure 5 Analogue signal averaging system. (a) Direct operation, (b) indirect operation The AD converter has been tested widely in combination with a Hewlett-Packard model 5401A multichannel analyser. This analyser is characterized by a 10 MHz maximum counting rate and a deadtime of 2.2 ps between successive multiscaler channels. Operation at the maximum AD converter triggering rate of 100 kHz thus almost matches the multiscaling capabilities of the analyser. The results of averaging a noisy 0.25 V analogue signal at the 100 kHz triggering rate are shown in figure 6. On a single shot basis, this signal was severely distorted
L
.------,*
I, 0
I
I
I
2
4
6
I
8
IO
Tiwe !ms)
Figure 6 Example of direct signal averaging at 100 kHz triggering rate (10 &channel). 10 mV/oscillator pulse; 87 Hz signal repetition frequency by 50 Hz mains hum. It is clear from figure 6 that the hum has been averaged satisfactorily over 20 sweeps and that considerable improvement in signal quality has been obtained after a 2 min accumulation period (10 000 sweeps). The rapid rise and rather slower decay associated with the signal illustrated are reproduced faithfully by the signal averaging system, indicating satisfactory high frequency response. Conversion linearity has been measured by recording the number of oscillator pulses from the AD converter as a function of input voltage. Differential linearity of better than 1% is 917
L J Chapman and 0 M Williams Maximum range a t
1.57
/
/i
- -Maximum range a t 100 kHz triggering rate
0
,v:
5p
(IO ps/channel)
Oscillator pulses
,
l5
'PO 1'6 (PSI
,
50
Time after triggering
Figure 7 AD conversion linearity. 10 mV/oscillator pulse; 50 kHz triggering rate
obtained for all but the first microsecond of oscillator pulse transmission, as is evident in figure 7. The initial nonlinearity has been traced to distortion of the ramp voltage caused by the switching transient associated with the trailing edge ofthe delay gate, as discussed above in $2. The distortion has been minimized in the present design within the limitation of retaining simplicity and in fact may be avoided in signal averaging experiments by offsetting the hold voltage to a level where ten oscillator pulses are transmitted to the multiscaler when the AD converter analogue input is grounded. Positive analogue signals will then be converted to digital form according to the linear part of figure 7 beyond the first ten oscillator pulses. The above procedure increases the system deadtime to an effective value of 6 ps so that at the maximum triggering rate of 100 kHz only 4 ps are available for distortion-free oscillator pulse transmission. Voltage resolution on a single sweep basis is thus reduced slightly to 2.5% of the maximum signal level (0.4 V maximum at 100 kHz triggering rate). Considerably better voltage resolution is of course obtained by averaging over many sweeps and also at slower rates of operation where more analysis time is available between successive samples of the input signal. The presence of a significant RMS noise component on an input analogue signal, as typified in figure 6, precludes the need to consider the effect of quantum error in the AD conversion process. However, quantum error can cause considerable distortion at low voltages when clean signals such as obtained from laboratory oscillators and pulse generators are digitized and averaged. This effect has been reduced in the present system by adjustment of the oscillator frequency to a value different from the nominal lOMHz, sufficient to avoid synchronism with the analyser channel advance frequency. When the two frequencies are synchronized such that a constant delay occurs between the opening of the output gate and transmission of the first oscillator pulse, then an input signal level corresponding to, say, 4.4 oscillator pulses will always be rounded to 4 oscillator pulses. A ramp voltage of small amplitude will thus be transformed into a staircase voltage where each step represents a single quantum jump. On the other hand, lack of synchronism ensures that an input analogue level equivalent to 4.4 oscillator pulses may be converted to either 4 pulses or 5 pulses according to the 918
instantaneous delay between the opening of the output gate and transmission of the first pulse. A more accurate representation of the input signal level is thereby obtained upon averaging over several sweeps. Unfortunately, quantum error distortion is never fully overcome because of the finite width of the oscillator pulses and the finite risetime and falltime of the output gate. These effects cause oscillator pulses of reduced width and amplitude to be delivered to the multiscaler when the oscillator pulse almost coincides with the opening or closure of the output gate. At the 10 MHz maximum counting rate, such pulses may not be recognized by the multiscaler and nonrandomness thus introduced. Some care should therefore be exercised when clean signals of low amplitude are averaged and it may be necessary to mix the signal with a white noise voltage in order to achieve distortion-free averaging. The system described here has been utilized for averaging a wide variety of analogue signals from both experimental equipment and standard laboratory instruments. In particular, the system has been used for studying the temporal growth of ionization current in gaseous prebreakdown experiments in which metastable particles have a major effect in determining the rate of ionization growth (Haydon and Williams 1975). The characteristic signal observed in such experiments has been shown in figure 6 in order to illustrate operation of the signal averaging system, The system has also been used on a single sweep basis for producing digital records of analogue signals. 3.2 Indirect opevation The maximum temporal resolution of the signal averaging system shown in figure 5(a) is determined jointly by the deadtime of the AD converter and the maximum rate at which oscillator pulses may be counted by the multiscaler. When the AD converter is coupled to a 10 MHz multiscaler, the minimum time between successive samples is limited to typically 10 ps. For slower multiscalers, the oscillator frequency must be reduced accordingly and voltage resolution can be retained only by increasing the time interval between successive samples, Both the maximum temporal resolution and the range of analogue signal frequencies suitable for analysis are thereby reduced. Direct signal averaging is suitable for analogue signals up to a maximum frequency of typically 40 kHz, inadequate temporal resolution being obtained at higher frequencies. However, it is possible to overcome this limitation for repetitive signals by operating the averaging system indirectly in conjunction with a sampling oscilloscope, as shown in figure 5(b). In this mode of operation, the sampling oscilloscope is used in normal fashion, taking a single sample from each repetition of the input analogue signal and storing that sample until the input signal is repeated. The multichannel analyser is again operated in multiscaling mode and is triggered in synchronism with the oscilloscope sweep. Following each sample, the oscilloscope strobe signal triggers the AD converter and advances the multiscaler to the next channel. The sample voltage stored by the oscilloscope is digitized by the AD converter and the resulting counts are stored in the particular channel memory. Hence, as the sampling oscilloscope sweep progresses, the sample information is effectively transferred to successive channels of the multiscaler memory. Quiescent conditions are restored at the end of the multiscaler sweep and signal averaging occurs over successive sweeps. Temporal calibration of the multiscaler signal is obtained simply from the oscilloscope calibration. Indirect signal averaging utilizing a Hewlett-Packard model 181A oscilloscope with type 1810A sampling plug-in unit is
Conversion of a multichannel analyser
References Haydon S C and Williams 0 M 1975 J. Phys. D: Appl. Phys. submitted for publication Hoeschele D F Jr 1968 Analog-to-DigitallDigital-to-Analog Conversion Techniques (New York: Wiley)
i .e ..
I
I
I
I
Time (ns)
Figure 8 Example of indirect signal averaging. 750 Hz signal repetition frequency; 0.1 nslchannel
illustrated in figure 8. The signal shown here represents the output of a Tektronix type 2601 pulse generator at minimum amplitude with a deliberate cable mismatch introduced between pulse generator and oscilloscope. The sample advance time was set at 0.1 ns. It is evident that a significant improvement in signal quality has been obtained by averaging over 100 sweeps, many more features being clearly defined in the averaged signal. The temporal response is determined by the aperture time of the sampling oscilloscope and for current models allows indirect signal averaging experiments to be conducted for repetitive signals to within the frequency range of tens of gigahertz. There can, however, be some restriction at low pulse repetition frequencies since only one sample is taken for each repetition of the signal voltage and long accumulation periods may be required. Conversion linearity is determined jointly by the sampling oscilloscope and the AD converter and it may be necessary for some oscilloscopes to suitably delay operation of the AD converter until the oscilloscope sample is fully established. 4 Conclusions The AD converter described in this paper has enabled a multichannel analyser to be employed as an analogue signal averager over the complete range of signal frequencies commonly encountered in the laboratory environment. The AD converter thus effectively transforms the multichannel analyser from a specialized digital instrument into a general purpose laboratory tool suitable for both digital storage of analogue information and analogue signal averaging applications.
Acknowledgments This work was conducted at the University of New England and supported by the Australian Research Grants Committee. We wish to thank Associate Professor W J Sandle for the suggestion that high frequency signal averaging may be conducted in conjunction with a sampling oscilloscope.
Journal of Physics E: Scientific Instruments 1975 Volume 8 Printed in Great Britain 0 1975 91 9
