PA88 based high voltage and (hopefully) precision amplifier

[gmac34]

Hi, I’m starting this thread to document the project that I have done to make a (hopefully) quite precise high voltage amplifier.


Since my HP 3245A Universal source does not have the HV option (a 10x amplifier based on the Apex PB58), I have been exploring ways to replicate its functionalities, without spending too much…

I have settled on using a PA88 HV op-amp, not as powerful or fast as the PB58 but can work at higher voltage.

More info here https://www.apexanalog.com/products/pa88.html

As the PA88 supports s 450V supply it allows me to do +-200DC if configured as a 20x amplifier (using the HP 3245A as a voltage source for +-10V). AC performance is limited by the slew rate; I will make some “plots” in the next few days, but I can say that it is flat to a few kHz at 100V RMS  and can do audio frequencies at 20V RMS.
Noise Is quite low, and dominated by the noise of the HP 3245A ADC.

The amplifier board is designed following the datasheet with an attempt to add DC offset correction.

The gain is set using 5ppm/K° resistors 2x50K and 5.1k, trimmed with a regular resistor and a multitourn trimmer 160ohm +10ohm.

It is in general stable but drifts at high voltage. This is quite pronounced above +-100V, probably related to self-heating of the Feedback resistors (2x50Kohm). Initially, I was planning to use much higher resistors 2x470K but that led to oscillation, probably something in between would be best such as 2x100K.

The DC offset of these op-amps is in itself quite respectable; on my particular sample, I was getting 2/3mV of offset on the output without any correction and a short on the input.
With the correction circuit, I can lower it about 15 times, but it is not quite stable enough. And will still hover around +-0.10mV.

To summarise, at low voltage the performance is limited by the DC offsets
At high voltage it is limited by the gain drifting because of self heating
In the frequency domain it is limited by the 8 V/µs slew rate

I think it makes a decent extension to the  HP 3245A  or a PDVS 2 mini and can be used to calibrate 4.5 Digits multimeters.

The power supply section is based on a design by https://www.changpuak.ch/electronics/VB408.php It uses 2 A41-25-230L Transformers; these transformers can be configured both for 115V and 220V input. It should be able to provide +- 212 V in excess of 100 mA but is currently configured to limit at around 60mA. The supply is stable and low noise.

Amplifier board visible here:
https://kicanvas.org/?github=https%3A%2F%2Fgithub.com%2FpignoniG%2FPA88-based-HV-amp-200v%2Ftree%2Fmain%2FAmplifier
Power supply board visible here:
https://kicanvas.org/?github=https%3A%2F%2Fgithub.com%2FpignoniG%2FPA88-based-HV-amp-200v%2Ftree%2Fmain%2FPowersupply

And source here:
https://github.com/pignoniG/PA88-based-HV-amp-200v

I currently have populated an older version of the PCB than what is published on GitHub; this version did not include offset compensation so it comes with many bodges. I will receive the new versions in a few days and I will do some characterisation then.


In the mean time I hope this can be interesting and any comment/suggestion to improve things is welcome

[mawyatt]

Nice, like the case!!

Awhile back we did something based upon the APEX PA443 as an HV AWG dual output buffer, altho not for precision use, and can only swing about 300VPP. Works from a 12DC Input.

https://www.eevblog.com/forum/testgear/hv-buffer-amp-for-awg/

The PA58 and PA88 are expensive ($200~$300), the PA443 isn't cheap either at ~$50. 

Best

[gmac34]

The prices for a new PA88 from a reputable distributor are a bit crazy.
I found my PA88 for about 60$ (including shipping and taxes form the US) for 2 with heatsinks on eBay. I believe the same seller has more, but being used parts YMMV.

[coppercone2]

Shame not to put a heat sink on that expensive part.

I also made my own TO-3-8 heat sinks by drilling holes, filling with epoxy, and re-drilling the epoxy so it works with that chip

[gmac34]

It does have a heatsink, just not a very big one...
It came with the ICs I bought, so I went with it. TBD how much current I can pull form this amplifier before it really starts to warm up.

[mawyatt]

These HV op-amps get hot quick, that's why we put a heatsink directly on top of the PA443, then added a small fan directly on top of the heatsink (see link we posted above for details).

Used surplus for the PA88 is a good option. The PA443 price has increased since we got them, think they were ~$25 back then!!

BTW at one point we were considering creating a HV and HC version based upon the PA443. This would have a current buffer on the output to increase the current level to ~1A while still maintaining over 300VPP output. The idea for power protection was to use the SMPS as the limiting factor, where the PS would purposely droop and collapse both the + and - voltages if current levels were continuously demanded which created too much power dissipation in the amp or buffer transistors. Traditional peak current limiting would be employed for the NPN Darlington and Sziklai Darlington output devices.

Anyway, still thinking about this

Best

[TimFox]

Before retirement, I used APEX hybrid TO-3 amps at different voltage levels (different parts) with good results.
For the feedback resistor, it may help to use a longer string of appropriate resistors, both to reduce self-heating and possible voltage co-efficient of resistance problems.
Did you include a small capacitor in parallel with the feedback resistor?

[gmac34]

Ok, I have updated the board to double the gain resistors, I did not want to make another digikey order for this project but it seems I have enough precision 1206 to do 4*50k and 2+5.1k plus trimming.
This should lower the high voltage drift significantly.

There is no capacitor in parallel with the feedback resistor, I do not think it is needed as there is already a compensation capacitor C1 as per the data sheet but I can experiment if I see any oscillation.

[mawyatt]

If having stability issues, it's usually a good idea to include the feedback cap. If not required then just don't populate.

Parasitic capacitance to the negative input "sees" the entire loop gain of the op-amp and can have disastrous effects to waveform fidelity and stability, negative feedback capacitance can help mitigate these effects.

Of course the best medicine is a good design and layout which inherently minimizes these effects.

Best

[TimFox]

The devil is in the details, but it is often (not always) required to have a small capacitor across the feedback resistor to compensate for the input capacitance, either across the other resistor (non-inverting) or at the amplifier input (inverting).

[1audio]

I recently built a similar amp using the PA85 in a compount setup. Attached are some pictures. Not as nice mechanically but its a one off for a specific application- driving electrostatic actuators for calibrating measurement microphones. I can share the details if anyone is interested. I have extra PCB's. The PA85 can be had on eBay for as little as $40. Getting stability is not easy on these.

[Stray Electron]

I can share the details if anyone is interested. I have extra PCB's.

   I would like to see the details.  I've picked up a pile of Mil surplus circuit boards that have PA88s on them and I've been holding onto them waiting for a good project to came along. 

[1audio]

Being lazy and cheap I used EasyEDA to create the layout and JLCPCB to make it. Project here: https://oshwlab.com/1audio/hv-amp  The simulation file is attached and the spice model for the pA85 (change the extension to .lib). Getting it stable needs lots of fiddling and its quite load sensitive. The opamp I ended up using is an HA3-2625. The slower ones wanted to oscillate.

I have extra PCB's. I hacked an eBay switching supply for tube preamps to make the +/- HV supply: https://www.ebay.com/itm/401235834136  and the opamp supply is another cheap ebay 12V to +/- supply: https://www.ebay.com/itm/124193492824?var=425041936662 

[David Hess]

I have settled on using a PA88 HV op-amp, not as powerful or fast as the PB58 but can work at higher voltage.

Did you consider a discrete design for higher performance than the PA88 will provide?

[coppercone2]

lol he is right, I would rather crack open a amplifier textbook then pay for another APEX part, and from my experiance in doing something, it might not be all that much easier to pay for the apex part. 

twitchy suicidal bastard op amp

[Stray Electron]

twitchy suicidal bastard op amp

  The engineers at Martin Marietta (now Lockheed Martin) must have liked them, I used to have about 150 of their circuit boards that used that part. I was involved in field testing the system that used them and I can tell you that it worked extremely well! When they decided to abandon that system I had free access to mountains of hardware and I grabbed everything that I could find that had one of them on it. At one time I even had all of the schematics, parts lists, test requirements, etc but regretfully I never saved any of it. Hell, I even wrote the Theory of Operation Manuals but don't ask me today for the details!

[coppercone2]

I am sure they are great for their cost unconstrained tight space applications but with the price of a new one, and the fact that they are not super trivial to use, there is alot of merit into making your own amplifier if you can

[gmac34]

I was considering doing a compound version to lower the offset, with the circuit above, I can compensate it only at one temperature, as soon as temperature changes I get a few mV on the output...
But I'm afraid of oscillations.
This schematics using a LT1012 seems to be well behaved in ltspice, but reality may differ...

https://kicanvas.org/?github=https%3A%2F%2Fgithub.com%2FpignoniG%2FPA88-based-HV-amp-200v%2Ftree%2Fa74f7cb2e25ef267a25a3961789bf31f31c924ed%2FAmplifier%2520Compound

[David Hess]

I was considering doing a compound version to lower the offset, with the circuit above, I can compensate it only at one temperature, as soon as temperature changes I get a few mV on the output...
But I'm afraid of oscillations.
This schematics using a LT1012 seems to be well behaved in ltspice, but reality may differ...

There are several composite configuration which control input offset with a precision operational amplifier.  The LT1012 is a good choice for this because of its precision and low input bias current, and the LT1012 example circuit is practically foolproof once the time constant and attenuation are adjusted, but there are plenty of parts which will work.

I assume you are using an inverting configuration to easily stay within the differential input voltage range of the PA88.  The lack of offset null capability in the PA88 excludes the more interesting composite configurations.

If you do this, be sure to clamp the inverting input of the PA88 so that failure will not destroy stuff.  Precision will be improved by using several resistors in series for the feedback resistor to keep the voltage across each one down.  If the input and feedback resistor strings are of the same type and value, thermal and voltage coefficient of resistance errors will cancel to a large extent.  I would want something to protect against the feedback resistor string going open because of lowered reliability from so many parts in series, but I am not sure what form that would take.  The feedback capacitance should be just enough to cancel stray capacitance at the input.

Instead of using a ready made high voltage operational amplifier, I would start by extending Walter Jung's high voltage example circuit to higher voltage, but I enjoy the challenge of discrete design.  Based on my experience with a similar circuit at lower voltage, I think performance above 10 MHz would be possible; I achieved 600 kHz bandwidth with 2 MHz output transistors.  Current limiting can be added to the output by clamping the current mirrors, and I am not sure why Jung did not show that in his high voltage version.

[gmac34]

Sorry for the silence, I have had the pcbs sitting on my desk for a while but never the time to work on them.
I'm testing now the second (maybe third..) revision of this amp. visible here: https://kicanvas.org/?github=https%3A%2F%2Fgithub.com%2FpignoniG%2FPA88-based-HV-amp-200v%2Ftree%2Fmain%2FAmplifier%2520Compound downloadable here: https://github.com/pignoniG/PA88-based-HV-amp-200v/tree/main/Amplifier%20Compound

I decided to stick with the non inverting configuration to keep the input impedance high, avoid any trouble with dc offset at the input, and avoid the mental gymnastic of remembering to set my 3245A to -10 when I want +200v out. in exchange I had to battle with the stability of the composite configuration.
I had to "slow down" the LT1012 quite bit, and in the end the frequency response took a severe hit, so now my 100kHz tag on the frond would not even pass as a marketing tagline.
I'm still able to get an accurate 1kHz 100v RMS output, I can live with it. I'll make some frequency response plots at various levels soon.

Gain stability and DC offset are very good now, I cannot see the effect of self heating at high voltage.

I might also try the DC stabilised Fast amplifier that David mentions, I have most of the components already to make it.

On the matter of “protect against the feedback resistor string going open”
I think a reasonable approach would be to use a series of parallel resistors, in my case that would total to 8 100k resistors
4 x 100k+100k in parallel, I think I can live with the risk given the use I plan but it’s easy to implement even on the same board, staking two SMD resistors on top of each other.

I'm also not particularly happy of my "zener" power supply, it works well actually, but those 4 resistors are not particularly elegant...