HV OpAmp

HV OpAmp...

Can be built to arbitrary HV supplies, just change Q1 and Q2 to higher voltage devices. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson
Loading thread data ...

You might need some loop compensation to drive real loads. What's the sinewave output look like with a load? HV things often have to drive capacitive loads, and arcs happen, so I generally have a cap of my own across the output.

At higher voltages, R1 and R2 will fry. If you make them really big, things get weak and slow. Bipolars need base current, so R1 has a bad tradeoff. And buying HV resistors is a nuisance. Putting lots of low voltage resistors in series takes, well, lots of resistors.

Up to +-200 volts, just use a MOC8204, much simpler and no crossover distortion: perfect class AB.

formatting link

It's also interesting to cascode a couple of low voltage optoisolators with a couple of depletion fets.

formatting link

Even better still, do the PV+mosfet thing that I posted. 1500 volt DPAK mosfets are affordable, and can output a lot of current.

formatting link

No HV resistors or associated dissipation. Swings to both rails. Explicit current limits. Zero quiescent current. This can drive a capacitive load with microwatts of power dissipation on the high side... the main power dissipation is in the feedback resistor.

Reply to
John Larkin

On Sun, 11 Oct 2015 11:26:34 -0700, John Larkin Gave us:

HV applications all differ from application to application.

The proper HV supply ONLY supplies what is need to perform the required HV function. Extra storage on the output is typically a bad thing.

HV Arc suppression is properly managed by a series bulk form resistive current limiting element (read carbon composition). It doesn't actually stop the arc, it merely stops the upstream circuit element damage from occurring.

Your cap on the output will not stop all of your multiplier caps and diodes from fusing. Only an arc suppression current limiting series resistor of bulk form media will do that job. They are getting harder and harder to find too.

Reply to
DecadentLinuxUserNumeroUno

All applications differ from application to application!

If I'm already expecting a capacitive load, adding an output cap of my own here has nice effects on loop stability. The dominant pole becomes my cap in parallel with the load. And the cap helps absorb spikes kicked back by arcs. The energy stored in my cap is tiny, not dangerous.

Why would they ever fuse?

If my customer wants to arc across his own capacitive load, let him go for it! But I don't want that to damage my box.

Reply to
John Larkin

On Sun, 11 Oct 2015 11:56:24 -0700, John Larkin Gave us:

HVPS design must ONLY provide the amount of energy the customer requirements calls for. And no, not all HV load circumstances are capacitive.

Yes, arcs at the output can damage both the HV caps and the HV diodes in the multiplier if no arc suppression is implemented. This is assuming that most modern HV supplies are of the C-W multiplier design on the output stage, and in fact, most are.

Reply to
DecadentLinuxUserNumeroUno

I do not remember all of the details of an alternate way; been 30+ years.

Use HV PNP and HV NPN that are powered from the + supply pin of (say) a 741. The output of the op-amp had a load to gnd,and also went to a divided version of the output. Something like that. Maybe HV zeners used? Too long ago; sorry.

Reply to
Robert Baer

On Monday, 12 October 2015 05:40:29 UTC+11, DecadentLinuxUserNumeroUno wro te:

Carbon composition and carbon film resistors aren't a good choice for curre nt limiting. Carbon has a negative coefficient of resistance with temperatu re, which means you can form low-resistance hot channels.

I once saw a 10k carbon film resistor carrying 0.5amp at 0.5V for half an h our through such a channel - it was a contrived set-up, and a big capacitor charged to quite a high voltage had been used to establish the hot channel which was then sustained by a regular lab power supply.

It still measured 10k after the test, though there was dark line on the res istor value colour rings over what had been the path of the hot channel.

Metal oxide resistors are a much better bet, and I'd tend to pick one that was guaranteed to act as a fuse by blowing to open circuit under overload.

Carbon film and carbon composition resistors don't reliably stop that.

Spreading the potential load over a few resistors will help there, but John Larkin doesn't like doing that.

--
Bill Sloman, Sydney
Reply to
Bill Sloman

On Sun, 11 Oct 2015 17:24:37 -0700 (PDT), Bill Sloman Gave us:

Yes they are in HV circuit outputs (low power).

If the circuit arcs too much, they need to get replaced because they begin to fused shorted within the resistance medium.

Reply to
DecadentLinuxUserNumeroUno

On Sun, 11 Oct 2015 17:24:37 -0700 (PDT), Bill Sloman Gave us:

Doesn't matter. The desired mechanism is the fact that they are of bulk medium construction.

Reply to
DecadentLinuxUserNumeroUno

On Sun, 11 Oct 2015 17:24:37 -0700 (PDT), Bill Sloman Gave us:

Not for HV arc suppression They fuse open with the first few arc events.

You really know absolutely nothing about this.

Whereas I engineered such supplies for years for the likes of NOAA and LANL and SAIC and LLNL, etc.

Reply to
DecadentLinuxUserNumeroUno

On Sun, 11 Oct 2015 17:24:37 -0700 (PDT), Bill Sloman Gave us:

Absolutely a stupid statement. HV supplies are potted and non-serviceable and the thing has to survive numerous arcing events and keep on ticking after that licking.

You really know absolutely nothing about this.

Reply to
DecadentLinuxUserNumeroUno

On Sun, 11 Oct 2015 17:24:37 -0700 (PDT), Bill Sloman Gave us:

Yes they do. They damp the impulse created by the arc, which is exactly what their function in the circuit is.

You really know absolutely nothing about HV supplies.

Reply to
DecadentLinuxUserNumeroUno

You know absolutely nothing about negative temperature coefficient resistors. Once a hot channel is formed, it's resistance is too low to damp anything.

--
Bill Sloman, Sydney
Reply to
Bill Sloman

If you could get a HV PNP transistor to level-translate, it can be simplified further. For +/- 150V outputs, the pullup/pulldown NPNs have to take 300V, the PNP only takes 150V.

Op amp output through limit resistor to PNP emitter, base grounded, and the collector drives the pulldown NPN base (suitable base-to-negative and emitter resistors on the HV NPN).

Depending on the op amp, I'd consider allowing for a speedup capacitor across the high-value feedback resistor. It might be worthwhile shielding it,too; big resistances give me an itchy feeling

Reply to
whit3rd

That circuit was just a demonstration that all you need is a low voltage opto-coupler with a cascode.

Having now Spice-modeled Apex' MP118 (+/-100V) and PA99 (+/-1250V) OpAmps, IRL I'd use current mirrors and a more classic NPN/PNP complementary output stage... makes it easier to compensate.

This one, by Larkin...

simulates just fine, but would smoke in-real-life.

Cascoding avoids the "suicide-bias" issue.

A later Larkin post using depletion mode FET's looks feasible. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

You can also get cutesy with opto-coupled current mirrors...

You can get matched optocouplers, for example Vishay ILD615... I don't have a model, but you get the idea. Controlling the current gain is important to be thermal and AC stability. ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

The Apex amps that I've looked at had horrendous quiescent power dissipations, especially using their recommended app circuits. I'm designing a HV driver right now, an Apex replacement, that has about

0.1% of their dissipation.

Why? I've shipped products with this circuit, no problems.

Yours would fry the resistors at higher voltages. R2 is dissipating

200 mW already.

The opamp quiescent supply current sets the opto class-AB quiescent current. One can add a resistor across the opamp rails if you want more, which one probably doesn't here.

The MOC should have b-e resistors at higher voltages. B-E resistors can actually move it into full class-B if you want. Or do that on the LED side. But I really like the perfect class AB current splitting that comes from putting the LEDs in the opamp rails.

It works fine if you get the numbers right.

I've shipped that, too.

Supertex makes depletion fets up to 700 volts. Their new web site tries to keep that secret. At high voltages or currents, the PV thing with an enhancement mosfet works best.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

When I saw this thread I thought, ah-ah Kilovolts. But, no.

Anyway I thought I must add that I made an amplifier with an 18KV positive rail for driving a penetration phosphor CRT. Colour change with variation o f the final HT, 8 to 16 KV. In the mid/late 60's I used a pair of experimen tal MOV triodes (valves/tubes) in direct-coupled class AB driven from a 741 (what's that)to perform the colour change. We had to modulate the deflecti on system in synchronism to maintain the picture size. It worked satisfacto rily so we thought we had better get rid of the valves/tubes, because it wa s going to have to fly. For the "solid-state" version we ended up using a s tack of TV line output transistors (BUY 69 rings a bell)in the output stage . Interesting times.

Reply to
grahamholloway

I'd still seriously consider tubes if I had to make a 30 KV amplifier.

A 1B3 is a nice slow amplifier if the input is the filament voltage.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Memory on this not what it should be, but I thought drawing max possible i from a filament shortened its life greatly.

NT

Reply to
tabbypurr

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.