Analog Oscilloscope advantages?

[alm]

The only I'm aware of is the Tek AM502 that is from the same era as the 7A22 (and essentially the same design). Only available on the used market, however, and needs a TM50x mainframe. Preamble Labs made a stand-alone version of the 7A13 (1855?), but I don't think they ever made a low bandwidth high gain version.

[EEVblog]

For a quite a long time I've considered doing a simple single ended pre-amp for scopes. Something like even x10 single ended would be handy.
A diff preamp with all the bells and whistles like a Tek ADA400 would be nicer, but yeah, as has been mentioned, much harder to do.

Dave.

[HackedFridgeMagnet]

Unfortunately right now I really need another project like a hole in the head

You need a hole in your head to breathe, so I'll take that as a yes.
Interested +1

[robrenz]

I highly agree. HERE reply #142 is an example. Keep in mind that this scope has a signal out that can feed your digital scope.

These signal outs were usually designed for driving things like counters, and not optimized for high-fidelity low-noise. I believe they convert the (internal) differential signal to single-ended by just throwing away half of it, for example. Did you actually try it and find the performance satisfactory?

I did now It looks pretty good to me. Testing below

The other interesting plugin in my opinion is the 7A13 (similar to the later Preamble differential amp). This was a 100 MHz differential plugin with normal sensitivity, but with excellent overload recovery specs, which is useful for measuring parameters like transistor saturation. It could connect one of the channels to a variable DC voltage, allowing huge DC offsets (eg. 10 V at 1 mV/div) without using DC coupling.

The 7A13 is a another thing of beauty. I suggest buying the later models that have the LED offset voltage display, not the older mechanical version. I paid $50.00 for one in perfect condition a few months ago.
I have a preamble 1855 from before Lecroy took over so it fortunately does not have the added Lecroy interface stuff. I limits its output to the scope to +/-500mV so the scope is never overdriven. It does have a 10X gain setting and if not terminated into 50 Ohms you gain an additional 2X of gain. It also has a built in offset voltage supply of +/- 15.9999V with digital readout and setting.

Preamble 1855


Upper trace is channel 1 of a 7A26 vertical amp driven by the calibrator at 0.4Vpp.  The lower trace is channel 2 fed from the signal out into a 50 Ohm inline terminator.  Output into 50 Ohm is aproximately 25mV/screen div. Output into 1Mohm is aproximately 0.5V/screen div.  Without the 50 Ohm termination the signal lost some BW as the upper leading edge was slightly more rounded than the channel 1 trace. Those traces are pretty much identical.


This is a 40mVpp signal from the calibrator into the 7A22 but I have used variable V/div on the Tek to get a 100X relation between the input signal and the Rigol.  This is to allow the measurement features of the DSO to give valid (divide reading by 100)readings on the signal. This would be used for very small signal levels. The rounding is from the upper BW limit.
 

This is the Rigol recieving the above signal without a 50 ohm terminator and reading 4V pp which is 100 times the actual 40mV signal.
That is 8 Tek screen divisions at 0.5V/div = 4.0Vpp on the Rigol. Variable on the Tek vert amp allows making it match exactly.


This is the same 40mVpp signal input to the Tek but adjusted with Variable V/div to give the Rigol exactly 40mVpp. This allows the DSO measurements to be used directly.


This is a roughly 16µVpp noise signal on the Tek adjusted to give a 1000X to the Rigol which reads 16.8mVpp.

[Conrad Hoffman]

There's no question that you can build a far better differential preamp than the old Tek units. I started to do this, but by the time you put in the filtering and other features it's not a trivial project. The noise level of the Tek units is high, thus at high gains you'll be filtering down to 10 kHz or less, sometimes much less. OTOH, for the problems I need to solve this hasn't proved a limitation. I don't know if it's still around, but the Tek stand-alone preamp was about $1k and it's a pita to use. The plug-ins are just so much handier. (I used to use a SRS preamp at work and that wasn't bad if you can find one cheap.)

There was a time when the big scopes showed up at hamfests en mass, and at the end of the day if they didn't sell they were just left to be picked up as junk. I don't see them around as much and the condition of what's left isn't very good. Many units have been scavenged for tubes and you could never afford to re-tube one.

FWIW, I have a 350 MHz 7xxx mainframe and most of the plug-ins a person could want, but I use the old 545B most of the time just because it's adequate and comfortable. I keep a spare 545B around with the diode recovery test plug-in permanently installed because it's handy when sorting surplus diodes. Yes, I'm a packrat.

[PaulStoffregen]

Another nice advantage of analog scopes is fast and natural-feeling response to the knobs.

Many newer digital scopes are pretty fast, but some are terribly slow.  Admittedly much of the slowness happens when you turn on fancy features that analog scopes don't have.  But analog scope just "feel" more real or natural, since there's never a software-based lag.

[David Hess]

Where are you going to get a $185.00 differential amplifier that goes down to 10µV/div with a 5µV noise floor at 3kHz upper BW limit? Get your signal set up on the analog scope and then capture it on the digital scope with the digital scopes triggering if you need to capture a single shot event.

5uV rms in a 3kHz noise bandwidth is an abysmal input-referred noise of 91.2nV Sqrt/Hz (a little less worse in a 3kHz -3dB bandwidth). You could better that by > an order of magnitude with a $15 instrumentation amplifier chip and a few bucks worth of passives as a preamplifier (using x1 probes only!) for your DSO.

But a simple instrumentation amplifier is not going to have the same massive input overload capability and interface to passive probes as easily.  The 7A22 and even more the 7A13 have considerable slide-back capability as well.

There's no question that you can build a far better differential preamp than the old Tek units. I started to do this, but by the time you put in the filtering and other features it's not a trivial project. The noise level of the Tek units is high, thus at high gains you'll be filtering down to 10 kHz or less, sometimes much less. OTOH, for the problems I need to solve this hasn't proved a limitation. I don't know if it's still around, but the Tek stand-alone preamp was about $1k and it's a pita to use. The plug-ins are just so much handier. (I used to use a SRS preamp at work and that wasn't bad if you can find one cheap.)

A 7A13 or 7A22 could also be upgraded for lower noise.  I have considered doing this as well as adding chopper stabilization to the input stage to remove drift and 1/f noise as a consideration but they both perform well enough for me so far.

Design and build something from scratch and it will look more like the Preamble 1855 that robrenz has.

There are lots of inexpensive Chinese high voltage differential amplifiers available now so they are probably the best option for most people.

[GK]

Where are you going to get a $185.00 differential amplifier that goes down to 10µV/div with a 5µV noise floor at 3kHz upper BW limit? Get your signal set up on the analog scope and then capture it on the digital scope with the digital scopes triggering if you need to capture a single shot event.

5uV rms in a 3kHz noise bandwidth is an abysmal input-referred noise of 91.2nV Sqrt/Hz (a little less worse in a 3kHz -3dB bandwidth). You could better that by > an order of magnitude with a $15 instrumentation amplifier chip and a few bucks worth of passives as a preamplifier (using x1 probes only!) for your DSO.

But a simple instrumentation amplifier is not going to have the same massive input overload capability and interface to passive probes as easily.  The 7A22 and even more the 7A13 have considerable slide-back capability as well.

I wasn't suggesting that a simple low noise in-amp would have the versatility of the plug-ins mentions. I was just pointing out that the noise spec given wasn't something to write home about and that very low noise measurements are not strictly confined to the domain of specialized equipment.   

I'd be interested in seeing a successful chopper implementation that can remove 1/f noise. Back in the old days even the solid-state choppers typically switched at a rate well below typical 1/f corner frequencies. Then you have to demodulate with a low pass filter having a corner frequency much lower than the actual switching frequency to reduce the ripple to an acceptable degree. A wide bandwidth amp might have a 1/f corner frequency as high as 100kHz or more.

[David Hess]

Where are you going to get a $185.00 differential amplifier that goes down to 10µV/div with a 5µV noise floor at 3kHz upper BW limit? Get your signal set up on the analog scope and then capture it on the digital scope with the digital scopes triggering if you need to capture a single shot event.

5uV rms in a 3kHz noise bandwidth is an abysmal input-referred noise of 91.2nV Sqrt/Hz (a little less worse in a 3kHz -3dB bandwidth). You could better that by > an order of magnitude with a $15 instrumentation amplifier chip and a few bucks worth of passives as a preamplifier (using x1 probes only!) for your DSO.

But a simple instrumentation amplifier is not going to have the same massive input overload capability and interface to passive probes as easily.  The 7A22 and even more the 7A13 have considerable slide-back capability as well.

I wasn't suggesting that a simple low noise in-amp would have the versatility of the plug-ins mentions. I was just pointing out that the noise spec given wasn't something to write home about and that very low noise measurements are not strictly confined to the domain of specialized equipment.

I do not remember why the input noise specification for the 7A22 is as high as it is; it seems like it should be lower.  I know I looked into it in connection with improving its drift but I was only considering noise near and below the 1/f noise corner and it was a while ago.

I'd be interested in seeing a successful chopper implementation that can remove 1/f noise. Back in the old days even the solid-state choppers typically switched at a rate well below typical 1/f corner frequencies. Then you have to demodulate with a low pass filter having a corner frequency much lower than the actual switching frequency to reduce the ripple to an acceptable degree. A wide bandwidth amp might have a 1/f corner frequency as high as 100kHz or more.

Linear Technology has a couple of examples in application notes showing how it can be done.  In this case a pair of chopper stabilized operational amplifiers configured as integrators sample the positive and negative inputs and outputs before and after the JFET input buffer pair and adjust their independent offset voltages.  The integration constants are chosen so that only frequencies below the 1/f noise corner are corrected.  It is more complicated in this case because these amplifiers use complex cascode designs to support wide input voltage ranges and from what I remember, the 7A22 input stage starts as a differential amplifier and not a pair of buffers.

It may not be possible to avoid long overload recovery times.  On a new design I think this could be handled by clamping the inputs at a lower level.

On the 7A22 I would also try to replace the input JFETs with a modern lower noise ones.

[GK]

Linear Technology has a couple of examples in application notes showing how it can be done.  In this case a pair of chopper stabilized operational amplifiers configured as integrators sample the positive and negative inputs and outputs before and after the JFET input buffer pair and adjust their independent offset voltages.  The integration constants are chosen so that only frequencies below the 1/f noise corner are corrected.  It is more complicated in this case because these amplifiers use complex cascode designs to support wide input voltage ranges and from what I remember, the 7A22 input stage starts as a differential amplifier and not a pair of buffers.

It may not be possible to avoid long overload recovery times.  On a new design I think this could be handled by clamping the inputs at a lower level.

On the 7A22 I would also try to replace the input JFETs with a modern lower noise ones.

How high a 1/f corner and what chopping frequency? Do you have the app. note numbers handy? Just pondering some of the potential design hurdles and charge injection back into the input, due to the potentially high chopping frequencies required, immediately comes to mind. With chopper stabilization you can always kill demodulation noise trough the DC path by just low pass filtering with a low enough corner frequency, but once switching interference enters the AC path you're screwed.

1/f noise and DC drift has always been an issue in the design of sensitive analogue oscilloscope front-ends. Given how old the fundamentals of chopper stabilization are I'm sure if it was so easy these venerable old oscilloscopes of ours would have been designed this way, essentially free of DC drift and 1/f noise, from the year dot.
   
 

[mazurov]

Analog scopes are cheap so one can easily have several. The advantage is the same as having several DMMs - you can do more in less time.  Once every year, I calibrate my freq.counter against a GPSDO, this procedure takes several days and having dedicated scope showing the difference between signals saves a lot of time.

[David Hess]

Linear Technology has a couple of examples in application notes showing how it can be done.  In this case a pair of chopper stabilized operational amplifiers configured as integrators sample the positive and negative inputs and outputs before and after the JFET input buffer pair and adjust their independent offset voltages.  The integration constants are chosen so that only frequencies below the 1/f noise corner are corrected.  It is more complicated in this case because these amplifiers use complex cascode designs to support wide input voltage ranges and from what I remember, the 7A22 input stage starts as a differential amplifier and not a pair of buffers.

It may not be possible to avoid long overload recovery times.  On a new design I think this could be handled by clamping the inputs at a lower level.

On the 7A22 I would also try to replace the input JFETs with a modern lower noise ones.

How high a 1/f corner and what chopping frequency?

This actually uses chopper stabilized amplifiers so the chopping is further removed.  When I did it, I used the LTC1150 which chops at about 550 Hz and I already had a known low 1/f corner frequency of like 2 Hz for the LT1007 or 3.5 Hz for the LT1028.  There are better alternatives now like the LTC2057 which chops at a much higher 100 kHz.

One problem with applying this to an old amplifier is incomplete specifications on the old transistors so it is difficult to tell how much improvement can be made without detailed measurements.

Do you have the app. note numbers handy?

I just end up looking them all up from memory when asked this.  I should keep a list handy.   

Page 10 of Application Note 21 shows what I started from in designing a differential input and output version.  This example shows up in a lot of other LT documents like page 23 of the LTC2057 datasheet.  I recently ran across fully differential input example in Design Note 36.

Page 3 of Application Note 45 shows an example of a chopper stabilized FET pair.  Page 14 of Application Note 61 shows the same example.  I thought I saw a more recent implementation using an LSK389 although Application Note 124 shows the same idea without using chopper stabilized amplifiers.

Just pondering some of the potential design hurdles and charge injection back into the input, due to the potentially high chopping frequencies required, immediately comes to mind. With chopper stabilization you can always kill demodulation noise trough the DC path by just low pass filtering with a low enough corner frequency, but once switching interference enters the AC path you're screwed.

I never observed this as a problem when I did it but I was correcting noise and drift in low noise bipolar designs.  The overall noise was considerably improved over either amplifier alone.  When I trimmed the integration constant for lowest noise, the resulting gain versus frequency for the parrallel stages closely agreed with the transition frequency which would yield lowest integrated noise.  Measuring such low levels of integrated input noise was surprisingly easy.

1/f noise and DC drift has always been an issue in the design of sensitive analogue oscilloscope front-ends. Given how old the fundamentals of chopper stabilization are I'm sure if it was so easy these venerable old oscilloscopes of ours would have been designed this way, essentially free of DC drift and 1/f noise, from the year dot.

Offhand I can think of several reasons they may not have done it:

1. They lacked high performance monolithic integrated chopper stabilized amplifiers.
2. There were not *that* many high sensitivity analog oscilloscope front ends and in many (all?) cases, drift over temperature is significant in later stages as well which is why their calibration includes multiple balance adjustments.  This is an issue even with low sensitivity front ends (I can see my old DSOs warm up over several minutes when first turned on and it goes without saying for my CA vertical plug-in.) although I have noticed before that maximum sensitivity seemed like it was often limited more by drift than noise (the CA vertical plug-in again).  At low frequencies it is often difficult to tell the difference between drift and noise anyway.  Drift is a significant problem with the 7A22 and 7A13 so they handled it by making all of their balance adjustment user accessible from their front panel.
3. Drift and noise were already low enough.  This is arguable since as these specifications got better, they allowed for higher input sensitivity.
4. There is a potentially serious overload recovery problem.  If the front end is driven into overload, the chopper integrator will windup causing a massive increase in recovery time.  Luckily it should be a later stage which overloads.

[GK]

With an internal switching frequency limited to 100 kHz the LTC2057 it isn't going to correct for a typical wide-band amplifier 1/f corner frequency in the region of 1 kHz to 100 kHz. Modern bioplar-input op-amps with 1/f corners below 10Hz are a different matter. What is the 1/f corner of the Tek plug-ins under discussion?

[johansen]

more than one trigger.

the  Tek475 has a feature where the B trigger can have different sweep rates, and can be triggered in several different ways.
example:
http://johansense.com/induction_heater/6/P1020776.JPG
http://johansense.com/induction_heater/6/P1020775.JPG

first half is 10uS/div, the second half can be triggered from ext, delay, ch1, ch2, or from what's on the screen.
a second mode allows you to scroll through the delayed sweep. i think this feature was intended to allow you to trigger on a clock, and read digital data, i never used that feature.

[w2aew]

more than one trigger.

the  Tek475 has a feature where the B trigger can have different sweep rates, and can be triggered in several different ways.
example:
http://johansense.com/induction_heater/6/P1020776.JPG
http://johansense.com/induction_heater/6/P1020775.JPG

first half is 10uS/div, the second half can be triggered from ext, delay, ch1, ch2, or from what's on the screen.
a second mode allows you to scroll through the delayed sweep. i think this feature was intended to allow you to trigger on a clock, and read digital data, i never used that feature.

Later models replaced the "Mix" mode with a separate trace from timebase B. Still others added a dual delaying timebase. One major use for these features was to zoom in on an area of the waveform. These B timebase modes were accompanied by some very nice delay and trigger controls  I've made extensive use of these features for many applications over the years, before the advent of the digitizing scopes.

[David Hess]

With an internal switching frequency limited to 100 kHz the LTC2057 it isn't going to correct for a typical wide-band amplifier 1/f corner frequency in the region of 1 kHz to 100 kHz. Modern bioplar-input op-amps with 1/f corners below 10Hz are a different matter. What is the 1/f corner of the Tek plug-ins under discussion?

As near as I can tell, the corner frequency for the 7A22 will be about 10 kHz and for the faster 7A13 about 1 kHz but if the corner frequency is higher, then won't chopper stabilization will make a larger difference?

[David Hess]

more than one trigger.

the  Tek475 has a feature where the B trigger can have different sweep rates, and can be triggered in several different ways.
example:
http://johansense.com/induction_heater/6/P1020776.JPG
http://johansense.com/induction_heater/6/P1020775.JPG

first half is 10uS/div, the second half can be triggered from ext, delay, ch1, ch2, or from what's on the screen.
a second mode allows you to scroll through the delayed sweep. i think this feature was intended to allow you to trigger on a clock, and read digital data, i never used that feature.

Later models replaced the "Mix" mode with a separate trace from timebase B. Still others added a dual delaying timebase. One major use for these features was to zoom in on an area of the waveform. These B timebase modes were accompanied by some very nice delay and trigger controls  I've made extensive use of these features for many applications over the years, before the advent of the digitizing scopes.

I have never been fond of mixed sweep which the 7B53A timebases for the 7000 series also support and much prefer the delayed sweep of the 465B.  I have been told mixed sweep was useful in video applications.

What I would like to see in a DSO is dual sweep capability like the 547, 556, and the 4 bay 7000 series support where you can display the same or different signals with two different sweep and trigger settings.  Various implementations of zoom on newer DSOs kind of allow this but are not as easy to use and not really a substitute.  You can tell when you ask the salesperson about this capability and they respond, "Sure!  Just like, um, hmm.  I am not sure how to do that."  Dual delayed sweep is handy as well and missing on DSOs.

My poor solution was just to get two of the same DSO and stack them on top of each other.

[FrankenPC]

This analogy just popped into my head.  TL;DR: Sometimes, the visceral experience is its own reward.

The  milling machine analogy.  I converted an old milling machine to CNC operation a few years back.  Now, it can basically operate in one of three modes:

1)  Full manual:  I bolt the material down to the bed, mount a cutting bit, and cut away.  This is a crude but visceral experience.  I still use this mode quite a bit.  Manually controlling the machine gives me instant feedback.  I can tell if I'm cutting too fast, or if the bit still has a good cutting edge by the sound and vibrations.  It's useful.  Setup time to do any given operation is really fast. 

2) Augmented:  When I need precision and repeatability, I use my three axis digital readout to tell me where I was and where I'm going.  It also gives me a good idea of how fast I'm cutting.  This is really good for working on small parts and allows me to quickly go back to zero if I need to repeat steps.  With this mode, I don't need to spend any additional time doing setup.  The whole process is still very quick and easy to do, but I have to pay careful attention to XYZ coords and jot down figures as I go.

3) Full auto:  I build a model in CAD/CAM, generate G-CODE, mount and zero out everything.  I might even do test runs to verify everything is going to work properly.  This takes much more time.  But the results are spectacular and incredibly precise.  Also, if I want to repeat the operation, I can do so with great precision and reasonable speed.  But, if I want to cut an entirely different part, I have to cycle through the whole setup/test process again.

1 = old school analog scope
2 = old school analog scope augmented with smart triggers and  measurement cursors
3 = Digital scope

I love all three.  I use all three.  It depends on what I need to get done and what experience I wish to have.  For me, using test equipment is entertainment in and of itself.  Getting my hands on a old Tektronix 7000 mainframe can be glorious.

[GK]

With an internal switching frequency limited to 100 kHz the LTC2057 it isn't going to correct for a typical wide-band amplifier 1/f corner frequency in the region of 1 kHz to 100 kHz. Modern bioplar-input op-amps with 1/f corners below 10Hz are a different matter. What is the 1/f corner of the Tek plug-ins under discussion?

As near as I can tell, the corner frequency for the 7A22 will be about 10 kHz and for the faster 7A13 about 1 kHz but if the corner frequency is higher, then won't chopper stabilization will make a larger difference?

Not to my understanding. At those frequencies the chopper stabilizer should have almost no influence at all.
 

[David Hess]

With an internal switching frequency limited to 100 kHz the LTC2057 it isn't going to correct for a typical wide-band amplifier 1/f corner frequency in the region of 1 kHz to 100 kHz. Modern bioplar-input op-amps with 1/f corners below 10Hz are a different matter. What is the 1/f corner of the Tek plug-ins under discussion?

As near as I can tell, the corner frequency for the 7A22 will be about 10 kHz and for the faster 7A13 about 1 kHz but if the corner frequency is higher, then won't chopper stabilization will make a larger difference?

Not to my understanding. At those frequencies the chopper stabilizer should have almost no influence at all.

This only helps with drift and noise at low frequencies, specifically flicker noise, which is otherwise a problem.  I would not go to this much trouble without changing the 7A22 input JFET for a lower noise one like an LSK389 or LSK489.

On the faster 7A13 which lacks a cutoff frequency low enough to be useful for these sorts of measurements, it would just improve drift which would still be helpful.

[GK]

This talk of chopper stabilization has given me cause to read up a bit more about the technique in my old analog computing books. For my analog computer project I developed the patchable "summer amplifier" stage several months ago. The specifications call for a +/-100V output voltage swing, so I designed a composite amplifier with a IC op-amp first stage and a discrete, HV op-amp second stage. I've been slack lately working on building the computers auxiliary units instead of the fundamental operational blocks, but I'm really glad that I haven't sent the board designs off for manufacturer now, because the chopper stabilization route turns out to be technically much better.

I'm just finished a preliminary design of a much improved "summer amplifier". The previous composite approach was taken to benefit from the low Vos and Ib of the integrated op-amp acting as the 1st stage. For adequate stability margins the HV discrete op-amp stand alone was compensated for a conservative GBW product of 1 MHz. However when wrapped up into the global feedback loop of the IC op-amp 1st stage the global feedback loop of the composite pair has to be compensated to at least 1/3 the GBWP of the 1 MHz stage to maintain an adequate phase margin.

However a better approach appears to be stabilizing the discrete HV op-amp with a chopper. That way the full 1 MHz GBWP can be maintained while keeping Vos and Ib to negligible levels.

Here is my preliminary circuit. I have applied the chopper stabilization topologically identically to how it was applied to high voltage, differential-input vacuum tube computing amplifiers back in the 50's through to the 70's:



The net input voltage offset error of the discrete HV op-amp is essentially reduced by the gain of the DC-coupled chopper amplifier path, which = 1010 Av.
For simulation purposes I have deliberately inserted a offset error voltage source of 50mV (the voltage source in series with the gate of jfet J1).

Here a sim of the power-up sequence, as the initial 50mV offset error is "tuned" out by the chopper (for this test all signal inputs to the left are left floating):   



The response of this computing amplifier of course is completely flat down to DC. For the chopper amp I have selected AD8538. This part is cheap and has (at 25 degrees C) a typical Ib of 15pA and a Vos of only 5uV. I don't have a SPICE model for this part, so I just approximated it to a adequate degree (the part of the schematic enclosed inside the dotted rectangle) Here is the low frequency square-wave performance:


   
And the "high" frequency square-wave performance just for good measure 

[GK]

From one of my AC books, here is a typical example of an old school, chopper-stabilized tube amplifier, with the same basic topology.

Has anyone here ever measured the closing contact bounce duration of a typical small signal reed relay? I've got some tube amplifiers I'd like to stabilize and using active semiconductor circuitry would be besides the point. When essentially switching uA even the cheapo reed relays have MTBF's in the order of >10e9 contact closures.

If I half wave rectify the AC from a mains transformer secondary, and R-Zener clamp the voltage at, say, 1/5th the peak AC amplitude, I'd have a nice approximation of a 50Hz square wave generator that can directly drive the coil of a single SPST reed relay. The other chopper relay for demodulation can be driven in the same manner, the required 180 degrees out of phase, by rectifying the same secondary in the opposite polarity.   

 

[David Hess]

I might try a mercury wetted reed relay but a better option would be the way HP did it with a neon bulb astable oscillator and some CdS photocells.

The old Linear Technology application notes show a number of different composite amplifier configurations for achieving the same thing.

[GK]

Yes, photoelectric choppers are covered in my texts too. That would be an alternative, though I think I might go out the back and measure a reed relay..............

[David Hess]

What book is that?  It looks pretty interesting.