HV amplifier/follower for function gen

[Orson Maxwell]

Hello, dear colleagues!

I've been watching Dave's channel and browsing this forum's wisdom for some time, but it's my first post. So hey!

Recently, a need arose to be able to test inductance of various chokes and transformers as well as plot simple B-H curves of refurbished magnetic cores, but I need your savvy in solid state schematic design, folks.

I have entry level test kit (JDS6600 gen and a Tek TBS1032B DSO), so right now I'm able to test magnetisation curves and leakage inductance of power trannies during no load conditions and AC inductance of various chokes using the resonance method.

I mainly do some audio gear repair stuff and designing tube amps, so I'd like to be able to supply significant signal swings and DC magnetization currents in order to be able to test magnetic materials with both DC and AC components at required levels and with variable core gaps.

So basically I need a rig that is able to supply amplify my generatior's 10Vampl signal by at least 32.5 (32.25dB) and then supply it to a low impedance load which is either a test coil, a transformer winding or a choke in the range of 1-200kHz (I guess) at least.
As for the DC component, I'd like to be able to supply at least 500mA into a load of 500Ohms maximum.

Looking at these specs I immediately understand that I'll need some kind of an high voltage amplifier for the test signal and then an emitter or source follower of some kind as a buffer to drive the test coil. Looking at integrated op amp solutions, I wonder if I should go for a discreet amplifier or even a step-up transformer for the gain.

I'm also tempted to use a couple of variacs and hack together the supplies that way, but then I lose the ability to vary the frequency which is very helpful.

I'm also tempted to supply the DC component via a seperate test winding, but it is not really feasible when testing existing devices.

This is where I start to feel uncomfortable as I do not have the schematic lexicon to construct the device in question. I'll be really grateful for any advice.

I can simulate circuits in spice, so I'd like to think that I'll need just a nudge in the right conceptual direction.

Thanks in advance!

[precaud]

Krohn Hite makes nice amplifiers that might fit the bill but hang on to your wallet:

http://www.krohn-hite.com/htm/amps/Amplifiers.htm

[awallin]

So basically I need a rig that is able to supply amplify my generatior's 10Vampl signal by at least 32.5 (32.25dB) and then supply it to a low impedance load which is either a test coil, a transformer winding or a choke in the range of 1-200kHz (I guess) at least.
As for the DC component, I'd like to be able to supply at least 500mA into a load of 500Ohms maximum.

pro high-power RF stuff (lie AR, Kalmus, mini-circuits etc) will cost a lot...

Something like the BK Electronics power-amp modules will drive a few hundred W at 100 kHz (but not much beyond that). possibly with a bit of tweaking of the preamp-stage (smaller compensation-caps, but still large enough to avoid self-oscillation).
http://www.bkelec.com/modules/mf1000.htm
lots of similar mosfet-modules with different power-specs available on e-bay also...

[_Wim_]

Why not consider a PA (high power) audio amp? It will not reach your requested 420V and 200kHz, but when you dc-couple it, it will already make quite a bit of measurements possible. By coupling the output of the amplifier (attenuated version) back to one of the input channels of the sound card, a non flat frequency responce of the amp under load can be compensated (I pressume your want to measure with a sound card?)

[_Wim_]

Another option can be a power supply like a kepco BOP-series. These only go up to about 20kHz, but can be found relatively affordable on Ebay...

http://www.kepcopower.com/bopfeat.htm

[Orson Maxwell]

Hello, dear colleagues!

Thank you for you feedback, everyone! I was actually considered an audio amplifier, however I failed to find the one that meets the HV specs. Some of them go beyond audio band, but the actual performance is usually not specified and needs to be tested.
As a matter of fact, I was thinking about putting up a D class amp myself for this purpose, but haven't researched if there are any pitfalls regarding the schematic design (especially the reconstruction filter).

I might have phrased my OP a bit misleading - I was thinking about assembling my own amplifier instead of buying a piece of kit (hence the projects section) - since I'll definitetly have to get a board for the opamp integrator and stuff like this, so I thought I'd make a nice device with the coil driver integrated.

Anyway, thanks, guys - I have a new direction to investigate now - I'll what it takes to make a high voltage low power D class amp to my spec. I don't really need good THD figures or any particular accuracy - maybe worth a look.

Cheers.

[_Wim_]

You could have a look at "100V line" speaker and amplifiers (aka "installation" speakers). You can maybe re-use the transfo from such a speaker, of modify such an amplifier to suit your needs...

[Mazo]

If you are willing to try and build a say 10x step-up transformer(ferrite core I think is the most probable candidate) you can roll a class G amp from say +-24V and +-48V(easy to get ready made PSU) rails and you are set.The idea being that there aren't that high voltage trannies(BJTs) that are meant to be used in linear region(AFAIK) but there are plenty in the range that I mentioned that can handle the increased current.Bonus-lower voltage higher current output for free

[Orson Maxwell]

Yes, this idea has crossed my mind, but in a slightly different gain staging formula - step the voltage up via a transformer and then put together a high voltage MOSFET unity gain buffer. But an output transformer for a TDA7294 amp might prove to be a viable option - it is capable of delivering 30-40W in the 5-200000Hz range from what I've found in the Net.

One thing I'm trying to investigate now is how to do DC injection, which would affect the choice of the topology.

[Mazo]

I think i've misread your original post.How much AC power do you need exactly?

[Orson Maxwell]

I reckon about 50W in order to supply enough idling current to 100-200W power transformers with a bit of headroom for overvoltage saturation/leakage testing.

[duak]

Could the DC injection be performed with a variable DC supply connected in series with the transformer or transformer/ buffer?  It will require that the AC section be able to pass DC.

The hp 3245A Universal source had a high voltage option with +/- 100 V capability.  Perhaps a look at the design of the output stage might give you some ideas.  I was able to find service info plus schematics on the 'net.

Please note that driving a reactive load such as a capacitor or inductor can be very difficult for an amplifier not designed for it.  This is because the reactance causes the current to be up to 90 deg out of phase with the voltage which requires that an active device have maximum voltage across it while having maximum current passing thru it.  Remember too that capacitive reactance falls as the frequency goes up.  Applying 100 KHz at 100 V to a 100 nF capacitor requires 6.29 A, something non-trivial.

Cheers,

[Berni]

I have needed high voltage at high power and high bandwith recently for doing some sonar stuff.

I ended up designing an amp for it but it needed a fair bit of engineering behind it before it worked. In the end it would produce about 100 Vrms output with 600KHz of bandwidth along with an output power of about 1kW. The output stage was made from 8 HV MOSFETs in half bridge pairs and that output power was only in bursts, the cooling is not up to the task for continuous operation at such powers. The supply for it was +/- 160V.

The thing loved to oscillate in weird ways, including the output transistors themselves, creating a 400MHz radio transmitter. I would never believed that such large MOSFETs can switch so fast. Eventually i tamed all of it down to get it working reliably. My first go at it was using BJT transistors in the output stage but i could never get them to go fast enough (Turns out large BJTs really really don't like turning off in a hurry)

[T3sl4co1l]

Oh yes, modern (Superjunction) power MOSFETs are quite squirrely, I often see them oscillate in the 200-400MHz range under linear conditions (switching edges making tone bursts, etc.).  Typically the offending problem is having too low of a gate-source impedance so it makes a tuned-gate oscillator, and typically all that's needed is a ferrite bead on the gate terminal (assuming you can tolerate this -- a high frequency switcher may overheat the FB, needing another approach instead).

Still been meaning to try a linear RF amp with them, see how much gain-bandwidth vs. impedance I can get.  The main downside is the poor power vs. capacitance ratings, i.e. a typical "6A 600A" type is only good for 30W (fullpak) to 100W (TO-220, TO-247), or a typical 10A 900V TO-247, 150-250W.  Meanwhile, Coss is in the 100pF range for these parts, assuming you keep Vds > 50V or so (the extremely high Coss at low Vds will manifest as ludicrously slow clipping recovery).  All together, these power and voltage limits circumscribe the operating point, load impedance and bandwidth you can get from them.  You can expect, say, a few dozen watts into 50 ohms, with ~30MHz bandwidth.  Not at all impressive, in the grand scheme of things (a similar rated LDMOS will do multiple GHz, in an otherwise very similar circuit), but hugely better than ye olde IRF540 and such (and none of the heater power that a vacuum tube would incur -- which, by the way, can otherwise offer very similar load impedance (low 100s ohms, nothing intractable) and output bandwidth).

Tim

[Berni]

Yes ferrite rings on the drain pin of all the FETs is what ultimately fixed the oscillation issue (that indeed only happened at max supply voltage and just the right output current). I actually had resistors on my gates that went to near by BJT push pull stages, but i was suspecting that the parasitics of my external zenner diode on the gate was forming the oscillating circuit. The thing contained all in all like 30 transistors so chasing the oscillation modes was quite a job. I likely over complicated the design, but hey it worked in the end.

Heating was not a problem since the amplifier operates in low duty cycle pulses of a few miliseconds of operation every 100s of miliseconds. In fact the amplifier is designed in a way, that allows it to go from off (Drawing zero current from the HV rails) to fully operational at full power in about 10us, then back to off again where it floats its output.

I came across some pretty interesting amplifier designs during development of it (LT had Jim Williams make a superb app note for it). There are designs that involved vaccum tubes or even floating SCRs. But it also shown a audio amplifier used as a rather powerful output stage. All in all making this thing work has taken a lot longer than expected.

[T3sl4co1l]

Yes ferrite rings on the drain pin of all the FETs is what ultimately fixed the oscillation issue (that indeed only happened at max supply voltage and just the right output current). I actually had resistors on my gates that went to near by BJT push pull stages, but i was suspecting that the parasitics of my external zenner diode on the gate was forming the oscillating circuit.

Oh yes, zener capacitance is more than enough to close the loop and make an oscillator. I did exactly this before; FB after the zener fixed it.

Heating was not a problem since the amplifier operates in low duty cycle pulses of a few miliseconds of operation every 100s of miliseconds. In fact the amplifier is designed in a way, that allows it to go from off (Drawing zero current from the HV rails) to fully operational at full power in about 10us, then back to off again where it floats its output.

Yeah, switched or variable bias is great for pulsed amps.

Regarding BJTs, they get quite slow at low currents, so you really need more of a class A design.  There's a (patented?) design that does that, maintaining a minimum current in all phases of the signal, while exhibiting overall class AB behavior.  I forget what it's called though.

Tim

[Mazo]

Had some fun with LTspice.Here is a bare bones circuit-ALOT of additional circuitry is needed and tweaking the design(dead time and SC protection is desperately needed) but I hope you like it.
The bandwidth is far from 200kHz but the cirucit is pretty simple and atleast covers the audio band very nice.The feedback loop after the filter is to compensate for nonlinearities in the output filter circuit(but it introduces a whole bunch of problems if the load isn't resistive so beware).If you intend to use the circuit with highly inductive or capacitive loads disconnect that feedback loop and feed the FG output to the base of Q1.

[Berni]

Also screenshot of the schematic please.

[Orson Maxwell]

Yeah, I installed LTSpice just to look at it.



Haven't played with the sim yet however. I'm looking and this Apex app note right now: AN56

[Berni]

I see you had quite the fight against oscillations with all that compensation everywhere.

The follower output stage I had the least trouble designing. That was simply a MOSFET P N pair push pull follower that had each of its gates driven by a diode biased PNP NPN pair follower. The inputs to the two followers ware separated by a voltage reference for setting the bias current of the output MOSFETs. The FETs and BJTs together had a negative temp co so that the bias fell as it got hot. Yes this means a lot of transistors but it worked.

What was more difficult for me was designing the gain stage. The job of it was to take the few Volt input signal and get it to the 300Vpp needed by the follower power output stage. I wanted it to have a reasonable frequency response up to 800KHz and be DC coupled so that it can start up instantly while also being reasonably close to rail to rail (It got within a few volts of the HV supply. High bandwidth and swing needs low value resistors, but that causes large currents trough the stage and that makes things toasty in a hurry due to the large supply voltage.

The gain stage does not rely on any feedback as its linear by design. This makes the final output tolerable of 100nF across its output. But still a opamp was avalable to do closed loop around the whole amp and make its frequency response ruler flat past 500KHz, but that did take some of its rock solid stability away.

The whole thing could be beefed up to provide 1kW continuous output, but being class AB that means a ton of cooling and a giant PSU. All of what I could avoid due to pulsed operation and large caps on the rails.

[Orson Maxwell]

Care to share a schematic? I don't need this kind of power capabilities or bandwidth, but it sounds like I'm looking at a similar topology.
Thanks for the feedback, everyone! I really appreciate it.

[Mazo]

I see you had quite the fight against oscillations with all that compensation everywhere.

If you are talking about the circuit that I posted it's a phase shift oscillator with alot of excess gain to get the HF square wave(hence the 3 RC low pass filters).If OP prefers he can get a clocked solution will be pretty much the same.BTW I would also be very glad to see a schematic of your high voltage high power amp.

Anyway to OP:You should make a decision if you are after a switching solution of after a linear one(also transformer coupler or not).
I will try to list some pros and cons of both(IMO).
Linear:
-may require alot more discrete circuitry and more "design"
-output devices probably are going to be more expensive
-Alot of cooling may be needed(class G may be your friend here)
+layout slightly less forgiving(watchout for the pesky VHF oscillations going behind your back )
+the 200kHz bandwidth needed is easier that way(if your output devices are fast enough)
+lower distortion,no switching noise,possible to produce small amplitudes that don't get drowned in noise.
Switching:
-If you filter it with a 2pole filter atleast 800kHz switching frequency needed(and you only get 24dB reduction of the switching noise )
-If you want to get feedback from the POL and get crossover frequency after the filter breakpoint(essentialy a buck converter),you will need alot of knowledge to compensate it,(reactive loads will mess with your output filter so good luck  ).Feedback before the filter or with a crossover frequency significantly smaller than the filter corner frequency is easy(bandwidth reduction,no control over filter resonance!).
-Layout extremely important(EMI,ringing,stuff blowing up)
+Low power dissipation
+Probably cheaper output devices
+Not much discrete circuitry needed and no big heatsinks so design will be smaller than equivalent linear.
Last but not least is bridge tied load a possibility?

[Orson Maxwell]

Anyway to OP:You should make a decision if you are after a switching solution of after a linear one(also transformer coupler or not).
I will try to list some pros and cons of both(IMO).
Linear:
-may require alot more discrete circuitry and more "design"
-output devices probably are going to be more expensive
-Alot of cooling may be needed(class G may be your friend here)
+layout slightly less forgiving(watchout for the pesky VHF oscillations going behind your back )
+the 200kHz bandwidth needed is easier that way(if your output devices are fast enough)
+lower distortion,no switching noise,possible to produce small amplitudes that don't get drowned in noise.
Switching:
-If you filter it with a 2pole filter atleast 800kHz switching frequency needed(and you only get 24dB reduction of the switching noise )
-If you want to get feedback from the POL and get crossover frequency after the filter breakpoint(essentialy a buck converter),you will need alot of knowledge to compensate it,(reactive loads will mess with your output filter so good luck  ).Feedback before the filter or with a crossover frequency significantly smaller than the filter corner frequency is easy(bandwidth reduction,no control over filter resonance!).
-Layout extremely important(EMI,ringing,stuff blowing up)
+Low power dissipation
+Probably cheaper output devices
+Not much discrete circuitry needed and no big heatsinks so design will be smaller than equivalent linear.
Last but not least is bridge tied load a possibility?

Thanks - this is really helpful. But before that I think I should pass up the idea of any funky DC bias current as it is really beyond my skill at the moment. I may go the good old "huge freaking choke" route of decoupling a seperate DC supply from the AC amplifier (I have just the canditate iron for that).

As for the points above, the idea of a 800kHz (or 400kHz if I cut the BW spec in half) switched mode amplifier into a hugely inductive load kind of intimidates me to be honest. Since my power requirements are not that huge, I might want to stick to the linear side and employ some nice cooling solutions which I'm familiar with from my past power LED (30-100W) experience.

I really don't care much about distortion characteristics because I can compensate for them later if it is not something insane and within say 10% THD.

Also, small signal measurements can be important for determining the initial permeability of materials.

EDIT: Oh, and regarding the BTL option, I will need to measure currents in the inductors and a voltage across an LC contour in case of resonant measurements, so I don't believe anything prohibits a bridged configuration (am I missing the DC injection via a big inductor bit?).

[T3sl4co1l]

Hah, Beige Bag, haven't seen that in years, a coworker liked to use it.

Tim

[sourcecharge]

If you guys are going to use that circuit, don't spread it please.

B2 rocks!