Princeton Applied Research 160 Boxcar Integrator

[wishboneash]

Well, GAS has become my way of life; a late entrant to the TEA society. Does anyone here have any info on this beautiful looking piece of equipment - the PAR 160 boxcar integrator? Those 10 turn potentiometers... ah, so cool! I have been using lock-in amplifiers for a while now and seeing this available, I decided I must have it as it complements the lock-in amp in some ways. The only problem, I can't find any manuals online, user or service. I did get a manual for the PAR 162 which is a modular system, but this one is a completely integrated single channel unit.

No magic smoke when turned on, internal DC voltages all look good. Found one burnt out bulb which I am going to replace. Contacted Ameteksi who now own PAR, waiting for a response. If there are any boxcar integrator fanatics familiar with this unit, I would love to hear from you. Even better if you have access to any manuals that I could copy and return. Thanks.
Sam

[wishboneash]

I put a video demoing some features of this unit. Excuse my kludgy editing/unrehearsed narration, but hopefully you get the gist of this amazing piece of instrumentation after watching some of the experiments I conducted. I can think of many uses for it already without having to resort to using a fancy digital storage oscilloscope.

[Kleinstein]

The old instruments got pretty much replaced with digital methods.  Many DSOs offer waveform averaging that das essentilly the boxcar integration for different delays in one run. Modern digital lock-in amplifiers may also offer this function that is somewhat related. Compared to the DSO they offer a better dynamic range (e.g. 18 bit instead of 8 bit ADC), but usually lower BW.

[maxwell3e10]

Thanks for the video. Actually I was looking at boxcar integrators recently. SRS still makes one and there are probably similar more specialized instruments as well. Digital oscilloscopes can do waveform averaging but surprisingly inefficiently, recording only 1/10th of the waveforms, roughly speaking. It seems that it would be relatively easy just to build such integrator with CMOS switches. Found the specifications for PAR 160 online. It's actually more versatile than current SRS model 250. The gate time goes down to 10 nsec, so it would not be totally trivial to replicate with CMOS.  It's cool how it can reconstruct a MHz waveform on a chart recorder. I feel people were generally smarter before computers came along

[maxwell3e10]

It's interesting that Princeton Applied Research came up with a lock-in amplifier and a boxcar integrator at about the same time. Lock-in amplifiers are still widely used, but boxcar integrators faded from use. There were some averaging digital oscilloscopes or signal averagers made by Nicolet and perhaps others, but the most modern instrument along these lines seems to be FastFlight-2, which is long discontinued, as discussed here: https://www.eevblog.com/forum/testgear/4-gss-oscilloscope-with-averaging-and-very-fast-retrigger/.
But signal averaging is such a basic task. It would be relatively easy to have an ADC with a circular memory buffer that simply adds data on each pass, without needing a large system throughput. I am just wondering why such instrument optimized for averaging doesn't exist.

[Kleinstein]

Lock-in and Boxcar integrator are somewhat related. Many lock-in amplifiers don't multiply with a sine, but use a simple rectangular wave and filtering before. So this is a special case of the boxcar integrator with 2 long windows and taking the difference.
Some of the modern digital lockin amplifiers from Zürich Inst. can also do boxcar integration, likely as waveform averaging. One way to implement a digital lockin is to first do waveform averaging and than do the multiplication with the sine or what ever reference waveform is wanted. I don't know the current digial lockin's, but it would be a nice option to also offer waveform averaging and maybe choose the waveform later.

An instrument to do good / effective waveform averaging would be nice, though maybe still a relatively low volume market, and not many users may be aware of the possibilties. Much of this would be more like a little additional software for a digitizer. With some luck we may get scopes (especially those use a 12 bit ADC)  that get better with this. As the data rate to the output is low this could even work for an USB scope.

[ch_scr]

Thank you for the demonstration of the device. To me it seems like, for the case of a single sine component, it is done today with fft on a scope? But with a complex waveform, this technique shows it - very impressive.
Regarding the red and blue "time constant" dials, the aperture time selector ring above them has markings in red and blue as well. I belive with longer aperture times, the lower red dial is used instead of the blue one. They probably select RC components mounted directly to wafer switches - the need for two dials might come from limits of physical construction.

[maxwell3e10]

I tried to search for any oscilloscope with a fast averaging mode (without screen update). Apparently one of the first digitizing oscilloscopes, Tektronix P7001, had a hardware signal averager (HSA) module.  Then nearly 50 years later, in the programming manual for Tektronix MSO 4/5/6 series scopes, there is a mention of fast averaging hardware mode. But I can't find any info on its actual performance.

[wishboneash]

Thank you for the demonstration of the device. To me it seems like, for the case of a single sine component, it is done today with fft on a scope? But with a complex waveform, this technique shows it - very impressive.
Regarding the red and blue "time constant" dials, the aperture time selector ring above them has markings in red and blue as well. I belive with longer aperture times, the lower red dial is used instead of the blue one. They probably select RC components mounted directly to wafer switches - the need for two dials might come from limits of physical construction.

Ah, yes. That makes sense.
I did find an abbreviated datasheet/manual online which describes this as normal resolution and high resolution mode related to the aperture time.

I am still amazed at how cleanly it rejects uncorrelated signals. The analog meter beats any digital output and this works real time as well without any programming. Moreover, the analog output can be captured and post processed further. For the price, it's unbeatable!
Thanks.

[wishboneash]

The old instruments got pretty much replaced with digital methods.  Many DSOs offer waveform averaging that das essentilly the boxcar integration for different delays in one run. Modern digital lock-in amplifiers may also offer this function that is somewhat related. Compared to the DSO they offer a better dynamic range (e.g. 18 bit instead of 8 bit ADC), but usually lower BW.

As an undergraduate EE student in the 80s, we had a class lab project to build a boxcar integrator in our 3rd year. I believe we did build it, but I don't have any good recollection of how it worked or even understand the intricacies of such an instrument at that time. This was one reason I acquired it and wanted to play with it when I have the time to actually tool around with instruments without bothering about deadlines or grades.

[ch_scr]

Thank you for the demonstration of the device. To me it seems like, for the case of a single sine component, it is done today with fft on a scope? But with a complex waveform, this technique shows it - very impressive.
Regarding the red and blue "time constant" dials, the aperture time selector ring above them has markings in red and blue as well. I belive with longer aperture times, the lower red dial is used instead of the blue one. They probably select RC components mounted directly to wafer switches - the need for two dials might come from limits of physical construction.

Ah, yes. That makes sense.
I did find an abbreviated datasheet/manual online which describes this as normal resolution and high resolution mode related to the aperture time.

I am still amazed at how cleanly it rejects uncorrelated signals. The analog meter beats any digital output and this works real time as well without any programming. Moreover, the analog output can be captured and post processed further. For the price, it's unbeatable!
Thanks.

There is a "Model 162 Boxcar Averager" that seems similar in function, it has a detailed manual over there that contains schematics and descriptions. Could you do a small teardown of your device and share pictures of the insides? It might shed some light on how similar the devices are.

[wishboneash]

Thank you for the demonstration of the device. To me it seems like, for the case of a single sine component, it is done today with fft on a scope? But with a complex waveform, this technique shows it - very impressive.
Regarding the red and blue "time constant" dials, the aperture time selector ring above them has markings in red and blue as well. I belive with longer aperture times, the lower red dial is used instead of the blue one. They probably select RC components mounted directly to wafer switches - the need for two dials might come from limits of physical construction.

Ah, yes. That makes sense.
I did find an abbreviated datasheet/manual online which describes this as normal resolution and high resolution mode related to the aperture time.

I am still amazed at how cleanly it rejects uncorrelated signals. The analog meter beats any digital output and this works real time as well without any programming. Moreover, the analog output can be captured and post processed further. For the price, it's unbeatable!
Thanks.

There is a "Model 162 Boxcar Averager" that seems similar in function, it has a detailed manual over there that contains schematics and descriptions. Could you do a small teardown of your device and share pictures of the insides? It might shed some light on how similar the devices are.

Sure, I can look into doing a teardown of the 160. I had seen the 162 manual before, it appears to be modular like the Stanford Research ones.

[garrettm]

These old boxcar integrators are pretty neat. After watching wishboneash's demo, I found the scan option (sweeping of the sampling window) to be the most interesting feature.

In this mode, the boxcar essentially acts as a sample-and-hold circuit. This would allow a slow, but high resolution digitizer (like a 6.5 digit DMM) to sample the signal for high fidelity reconstruction. In this way, fast, repetitive AC signals can be digitized using a slow DC voltmeter. I'm curious if this method could be used for high accuracy AC voltage measurements?

At any rate, it’s possible that, for the price, an old boxcar integrator and benchtop multimeter can outperform most fast digitizers sampling at >= nyquist frequency even with significant averaging. It would be interesting to do a comparison and determine the ENOBs for each setup.

[wishboneash]

These old boxcar integrators are pretty neat. After watching wishboneash's demo, I found the scan option (sweeping of the sampling window) to be the most interesting feature.

In this mode, the boxcar essentially acts as a sample-and-hold circuit. This would allow a slow, but high resolution digitizer (like a 6.5 digit DMM) to sample the signal for high fidelity reconstruction. In this way, fast, repetitive AC signals can be digitized using a slow DC voltmeter. I'm curious if this method could be used for high accuracy AC voltage measurements?

At any rate, it’s possible that, for the price, an old boxcar integrator and benchtop multimeter can outperform most fast digitizers sampling at >= nyquist frequency even with significant averaging. It would be interesting to do a comparison and determine the ENOBs for each setup.

Yes, I assume one could reconstruct or measure AC waveforms as I showed, as long as the AC signal is repetitive (any complex waveform) and the long term clock jitter is small. I hope to do some experiments for low and high frequency AC signals.

[wishboneash]

I did a teardown of the PAR 160 which can be viewed here -  https://youtu.be/zGFD-I0_4gY. It appears to be quite different from the 162 (looking at its manual). This one is built up using a whole slew of boards plugged into slots and highly modular. Unfortunately, I couldn't shed much light on the details of the cards without a proper service manual except to guess about some basic functions.