High Voltage Sin/Square wave amplifier

[KBJ62]

I am looking for an amplifier circuit which can amplify sine/square waves to +/- 250 Vpeak-peak. I'm currently using Ckt-1 for sine waves and Ckt-2 Square wave. 

[mawyatt]

Think you mean 250VPP, not +-250VPP since you only show 300V power supply. Take a look at APEX Semiconductor PA441 for a HV integrated type Op-Amp. We have evaluated these for a precision +-150V peak output HV amplifier, they are a nice Op-Amp.

Your circuits show many flaws, too many to discuss, please spend more time studying and refine your circuits if you decide to use a discrete design approach. One example is there is no low impedance return path for HV output current from the 300V supply, another is the op-amp output is only 2 Vbe's above the -VEE rail, another is the op amp output can't go towards the VEE rail.

Caution, don't build this circuit unless you want to burn up a bunch of components!!!

Best,

[David Hess]

Q1 in the first circuit is backwards.

Use local feedback around the discrete level shifter and output stage to better control gain within the feedback loop.

[Zero999]

It seems like the original poster has just kept plugging circuits into the simulator, until the desired result is achieved, which is never a good way to design something.

How much current is required? Here's a basic op-amp voltage booster circuit, which adds a couple of external transistors, in cascode with the op-amp's internal output stage. The op-amp's power supply floats, so the other op-amp in a dual package, can't be used for anything else. In this case. U1a is the master and U1b is the slave op-amp, with their outputs connected together with current sharing resistors: R8 & R9.

If the power supply isn't very stable, then replace R4 and R5 with 15V zener diodes.

It will need frequency compensation, unless a really slow op-amp, like the LM358 is used, otherwise it's likely to oscillate.

By the way, many of the SPICE models don't simulate the op-amp supply current accurately, so just plugging it into the simulator, doesn't always word and is why I used a generic op-amp model.

Q1 and Q2 need to be rated to at least 300V and have sufficient power rating.

The maximum output current is a little less than the rating of both op-amps together. Additional transistors will be necessary, if more current is required.

EDIT: Bad circuit attached. It exceeds the common mode range.

[TimFox]

You need to specify your requirement past the peak-peak voltage output, especially when dealing with square waves.
An important specification is the slew rate, maximum rate of change dV/dt, which will be an important limit on the rise and fall times of your square wave.  This is a non-linear problem, but can be simulated with a proper transient analysis in Spice.  The bandwidth is normally defined for relatively small amplitude sine waves, but due to slew-rate limitation the "power bandwidth" at high voltage may be much smaller.

[capt bullshot]

Here's a basic op-amp voltage booster circuit, which adds a couple of external transistors, in cascode with the op-amp's internal output stage. The op-amp's power supply floats ...

By the way, many of the SPICE models don't simulate the op-amp supply current accurately, so just plugging it into the simulator, doesn't always word and is why I used a generic op-amp model.

This circuit will exceed the common mode input voltage of the OpAmp due to the floating supply ...

The attached one works (kind of ...), I've built this amplifier long time ago as a proof of concept.

[Zero999]

Here's a basic op-amp voltage booster circuit, which adds a couple of external transistors, in cascode with the op-amp's internal output stage. The op-amp's power supply floats ...

By the way, many of the SPICE models don't simulate the op-amp supply current accurately, so just plugging it into the simulator, doesn't always word and is why I used a generic op-amp model.

This circuit will exceed the common mode input voltage of the OpAmp due to the floating supply ...

The attached one works (kind of ...), I've built this amplifier long time ago as a proof of concept.

I can't believe I missed that. You're right. The gain is far too high for the floating supply trick to be effective.

[Zero999]

This circuit is a little better.

[Atomillo]

This article might be useful:
https://www.edn.com/bootstrapping-your-op-amp-yields-wide-voltage-swings/

[Zero999]

This article might be useful:
https://www.edn.com/bootstrapping-your-op-amp-yields-wide-voltage-swings/

That's how the bad circuit I posted earlier works, or to be more procise, doesn't work.

Unfortunately it won't help.

Look at the schematic in the article


The AD820 has a maximum total voltage rating of 36V, so we could select values of R1, R2, R3 & R4, so V_CO - V_EO = 30V.

Now suppose, when V_OUT = 125V:

V_CO = 130V
V_EO = 100V

V_CM will have to be > 100V, otherwise the common mode range of the AD820 will be exceeded, so the input voltage must swing between -100V and +100V, for an output voltage swing of -125 to 125V, which means the gain is only 1.25V.
 
Now we could shave a few volts off, by increasing the supply voltage to the op-amp, but it doesn't gain us much. It's much better to use a more complicated cirucuit, which will also give more output current.

[David Hess]

Positive feedback has to be added to increase the gain and common mode range as shown in the second example from that article.

[Zero999]

Positive feedback has to be added to increase the gain and common mode range as shown in the second example from that article.

Silly me. I should have read the whole thing. I'm not doing very well in this thread, am I.

Positive feedback does make a bit more extremely sensitive to component value variation though. Try changing the component values a little and see what happens to the gain. I think it's impractical. Unless anyone has any ideas to make it better.

[Atomillo]

I'm also in the process of digesting the document so no worries (in my application I need a fast around 1us or less fall time -100V square wave for quenching a Geiger tube).
I expected bootstrapping to be "easier" than discrete component level design, might be not so much...
Another thing to worry not yet specified is Slew Rate for the demanded square wave. In this the design you mention should be better due to the higher current capability but one has also to avoid possible latch ups.
This is why in the document there always is a low pass filter in the input.

[capt bullshot]

I'm also in the process of digesting the document so no worries (in my application I need a fast around 1us or less fall time -100V square wave for quenching a Geiger tube).

If you want a square wave with "fast" edges, IMO using MOSFETs +as switches would be the easiest way to achieve this.

[David Hess]

I'm also in the process of digesting the document so no worries (in my application I need a fast around 1us or less fall time -100V square wave for quenching a Geiger tube).

If you want a square wave with "fast" edges, IMO using MOSFETs +as switches would be the easiest way to achieve this.

Or bipolar transistors.  Variable amplitude fast and clean edges can be made using emitter/source switched differential pairs to drive a parallel termination.

[SiliconWizard]

I'm also in the process of digesting the document so no worries (in my application I need a fast around 1us or less fall time -100V square wave for quenching a Geiger tube).

If you want a square wave with "fast" edges, IMO using MOSFETs +as switches would be the easiest way to achieve this.

Yep.
To the point that if you want to amplify both square waves and sine waves, two separate circuits would be much simpler than trying to make a general-purpose amplifier with a very high slew rate.

Of course, maybe with a clever design, you could reuse the same MOSFETS for the sine wave amplifier.

[Zero999]

I'm also in the process of digesting the document so no worries (in my application I need a fast around 1us or less fall time -100V square wave for quenching a Geiger tube).

If you want a square wave with "fast" edges, IMO using MOSFETs +as switches would be the easiest way to achieve this.

Yep.
To the point that if you want to amplify both square waves and sine waves, two separate circuits would be much simpler than trying to make a general-purpose amplifier with a very high slew rate.

Of course, maybe with a clever design, you could reuse the same MOSFETS for the sine wave amplifier.

The original poster did attach a separate circuit for amplifing squarewaves which would work, at low frequencies, at a low power level, but people responded more to the sine wave circuit, because it's more of a challange. It could easilly be improved, but it depends on the requirements.

[SiliconWizard]

I'm also in the process of digesting the document so no worries (in my application I need a fast around 1us or less fall time -100V square wave for quenching a Geiger tube).

If you want a square wave with "fast" edges, IMO using MOSFETs +as switches would be the easiest way to achieve this.

Yep.
To the point that if you want to amplify both square waves and sine waves, two separate circuits would be much simpler than trying to make a general-purpose amplifier with a very high slew rate.

Of course, maybe with a clever design, you could reuse the same MOSFETS for the sine wave amplifier.

The original poster did attach a separate circuit for amplifing squarewaves which would work, at low frequencies, at a low power level, but people responded more to the sine wave circuit, because it's more of a challange. It could easilly be improved, but it depends on the requirements.

Yes, of course. What I was suggesting though is a design that would work for square waves at higher frequencies, with a high slew-rate, and focus on a design for amplifying sine waves that would favor linearity over slew rate.

I was also hinting that a clever approach could actually merge both amplifiers in one.

[Neomys Sapiens]

Also take a look at AN-18, AN-21 and AN-87 from LT, AN-272 from NatSemi and the circuits in the datasheet of the OPA454

[Vovk_Z]

I just left it here. This circuit isn't very fast, but can be quite powerful (we can have 0.1-0.2 A at its output).
The original circuit is here. Works well with a fast, cheap and precise OP37.

[Atomillo]

Thanks for the app notes recommendations. I will be printing and reading carefully all of them!

And yes, a MOSFET switch indeed seems to be the way to go for my application, and is how it's been done before. I had some vague idea of changing the polarity of the voltage to accelerate quenching (make ions created after a discharge travel to the wire and not the wall) but of course a closer look at the literature reveals this has been done before with not much success so conventional quenching it is.

[KBJ62]

I would like to thank everyone for giving me their solutions. I did make a mistake of not telling the requirement properly. I am trying to drive a capacitive load which requires  high voltage of 300Vpp. The frequency will be from 10Hz - 400Hz.

[trobbins]

I'm just preparing something similar, but need to try and get to circa 390Vrms so have had to use two parafeed triode output stages, with balanced ss drive, and lots of feedback.  But the ss opamp options in this thread are quite interesting, although the convenience of using a high powered stereo ss PA amp has minimised prep effort.

[KBJ62]

Found this circuit in digikey's website with transistors for high gain.
https://www.digikey.com/en/articles/control-amplify-high-voltages-effectively-high-voltage-op-amp

[capt bullshot]

This is one of the circuits from the Linear Technology / Jim Willams app notes mentioned somewhere above.