# High voltage gain block

Schematic 1 shows a discrete operational amplifier which is used as a gain block to make up to 450V out of 12V. Q3 and Q4 are used to generate a current which is used by the current mirror. This is basically a simple op amp with a impedance transformation at the end (Q11 and Q1). The gain is set with R1 and R2. The other resistors are just for current measurement in ngspice which is use to simulate.

The dc transfer characteristic is shwon in figure 2. The source V3 is used as the dc control input.

Do you guys have any ideas how to make the output more linear starting from near zero and get rid of the plateau at the end?

• By "at the end" do you mean when the output is near 0 V or when it is near 500 V? – The Photon Jan 4 at 19:09
• I mean when the input is near 0V the output should be near 0V. – Gustavo Jan 4 at 20:27
• I think you aren't entirely understanding the diff-amp and its mirror load well enough. I can even see where this was headed. But it's not "there." Are you looking for the least possible modification? Or do you want to understand what's wrong, first? – jonk Jan 4 at 20:55
• If you want to keep this basic circuit design, you'll probably need to provide a negative voltage, -1 or -2 V, where R6, R7, R10 and Q12 currently connect to ground. Or somehow offset the voltage at the base of Q8 & Q9 above the input voltage. – The Photon Jan 4 at 20:56
• I want to understand what is wrong with it and the design does not need to stay as is. – Gustavo Jan 4 at 21:14

You might wish to explore some standard audio power amplifier output stage topologies, rather then doing a class A design with a current sink at the bottom like that.

A complimentary pair biased with a Vbe multiplier and a PNP current source seems to me to be likely to let you get pretty much all the way down (to within a few hundred mV of the negative rail), and within a few volts of the positive rail.

Do not expect a particulary good high frequency response, Miller capacitance will be a nightmare with that rail voltage, and watch the SOA and second breakdown issues.

There are useful ideas in Doug Selfs book on audio power amplifiers that look applicable to me (In particular some of the ideas on protection of the output stage are likely relevant).

Lets first discuss that Plateau at the end, that is, where the SPICE source has pushed the diffpair to an imbalanced condition, because the voltage divider (250K and 5K, or 51:1 division) no longer can move the right-hand base of the diff-pair any higher even as the SPICE source continues to move up.

Cure for this is to reduce the SPICE source maximum voltage to (500 volts - 2*Vbe) / 51 or (500 - 1.2v)/51 or approximately 499/51 or about 9.8 volts.

Now lets talk about that lower end behavior.

The diffpair (used nicely as an error amplifier, or servo-amplifer, or regulator amplifier) compares the input voltage from your SPICE source, to the output of the 250K/5K voltage divider.

To perform this comparison, the diffpair must be used properly, where

1) suitable current flows into the emitters from that single NPN below the emitters

and

2) the emitter-collector voltage needs to be several volts (examine the datasheet curves)

Lets assume the base voltage into the right transistor of the diffpair must be at least 2 Vbe (2 diode drops, which vary slightly with temperature and with current and with the transistor PARTNUMBER and probably whether the transistor is rated for 6 volts or 60 volts or 600 volts (as in this case) across the emitter-collector pins.

Recall that divider has a 51:1 ratio. You need about 2Vbe, or about 1.2 volts, so the bottom NPN (current source) outputs a steady current.

Thus the minimum voltage coming into the 51:1 divider is going to be

51 * 1.2 or about 60 volts.

To get this down close to ZERO volts Vout, you need to take the advice of someone here, and grab all those Ground nodes (except the 51:1 divider) and run those Ground nodes at about -1 or -2 volts.