High Voltage Discrete Opamp

[Matador]

Does anyone see anything obviously wrong with this topology? It is adapted from one of Sam Groener's designs (which itself seems to have come from the 990 opamp). It is essentially modified to:

a) Use high-voltage current references rather than BJT's that are diode biased
b) Be capable of operating from 400V rails
c) Uses BJT's that have 600V to 800V VCE ratings

I don't expect/need to be able to drive significant output currents, so there isn't a beefy output pair, as it can be difficult to find PNP transistors rated over 1A that can withstand high collector/emitter voltages. In this case, there are two applications: one, to use as a comparator in a high-voltage MOSFET power supply (meaning all that will be driven is the gate of a MOSFET), and the other is to increase the output of a signal generator that can output 10 Vpp max, up to something around 70 or 80 Vpp for testing tube output stages (ones that may even need to partially operate in class AB2). I expect 50mA or less in the worst case.

Here is the schematic:



Is there any reason that the dominant pole compensation (C1) would need to be adjusted from moving to higher voltage rails? Sam's original was rated for 80V supplies, whereas I expect to be operating this one between 150V and 400V.

 

[JohnRoberts]

If you build one you will find out soon enough.

Have you considered using an audio step up transformer? Constant voltage (background music) systems routinely step up low level 4 Ohm amplifier outputs to 70/100VAC.

JR

 

[Matador]

Have you considered using an audio step up transformer?

That would work in only one of my use cases, but yes, for audio testing a 1:10 would probably work, especially as I don't need to test especially high frequencies. A transformer would also be nice as the secondaries could be level shifted up to whatever operating point is needed (for example, level shifting down to -50V or so to directly drive the grid of a 6L6 tube, for example).

 

[volker]

Is there any reason that the dominant pole compensation (C1) would need to be adjusted from moving to higher voltage rails? Sam's original was rated for 80V supplies, whereas I expect to be operating this one between 150V and 400V.

The compensation depends on the input stage transconductance and your desired closed-loop bandwidth and gain. So using different transistors alone changes this relationship from the original circuit.
I either simulate or build and adjust.

 

[ruffrecords]

The main problem with high voltage BJTs is low hfe which limits the op amp open loop gain. Having said that, I have toyed with the idea of building a pre using a 48V supply

Cheers

Ian

 

[Matador]

There is a pair from Diotec semiconductor. Interestingly enough, most of the high voltage BJT's (e.g. VCEmax of 400V or greater) are only available in SMD packages.

But in any case, the MMBTA44 and MMBTA94 NPN/PNP pair are rated between 400V to 500V, yet have 'typical' hfe's between 50 and 200 when run at 10mA or lower.

 

[Matador]

Another possibility: FCX458, NPN, 400V, hfe 100-300 : FCX558, PNP, 400V, hfe 100-300

 

[Admiral-dk]

Thinking out Loud ....

Diodes as Input Protection - you don't want much more than 10-12 volt reverse voltage on the Base-Emitter of the Input Transistors or the will be very noisy afterwards ...!... and this is very likely in a Comperator Application + possible as an Audio Amplifier in some senarios ....

Old school Bipolar OpAmps often used Darlington Transistors in the Input - or if you don't need a very high Input Voltage - use Input with Low Voltager High Beta Transistors and Cascode them with High Voltage Transistors ....

Per

 

[user 133392]

"Class AB Amp Uses Two Floating Cells," Joseph Ting, EDN, March 19, 2009.

May not be what you're looking for but clever nevertheless...

https://proaudiodesignforum.com/ima...es_Two_Floating_Cells_Ting_EDN_March_2009.pdf

 

[offkey_99]

I think the input impedance "looking into" Q5 will be quite low (r_pi + HFE*68 Ohms) --> likely to severely limit your "loop gain".

It might still be fine though, but a simple remedy would be using one more NPN in-between as an emitter follower (basically creating something similar to a Darlington pair); might want to bias with resistor from emitter to ground with value in kOhms. This is a textbook option though, nothing crazy.

EDIT:I just noticed that Admiral-dk said essentially the same thing a few posts up.

​