Designing a differential probe for large class-D amplifier

[Twellmann]

Hi all

I am helping a friend with some electronics for part of his Ph.D. studies.

He needs to measure the voltage output of an loudspeaker amplifier in order to do some real-time adjustments on some large PA-systems. To begin with it will be the subwoofers.

The amplifiers are various large class-D amplifiers. For now something like four channels, but possibly many more (24 maybe) in the future as his projects continue.

So, the requirements begin something like this:

• Bridged/floating/similar type output stages with several hundred volts peak-peak.
• Only audio bandwidth.
• Battery-powered.
• Must be able to scale numbers = cheap.
• Low distortion rather than low noise.

This to me sounds like i need some form of high impedance differential probe. I have looked around a little bit at various teardowns of commercial diff probes and I feel that it should be possible to design something that would fit the bill without breaking the bank.

The input stage would look something like this for a 100:1 probe:


For the difference amplifier I first looked at the AD8479. It has very high common mode voltage capabilities and high input impedance, but also somewhat higher noise. I then realised that I don't precisely know what it is that I need.

Does it have to be a monolithic difference amp or could it suffice to use normal opamp and then get precision reistors to form the circuit instead?
Do I really need to be able to withstand very high common mode voltages?
Do I need very high input impedance at this point?
The AD8479 is used for high side current monitoring where it makes sense, but I don't see that it is needed in a diff probe like this.

I realize that this is only part of the design, but this is where I am at the moment. It is also entirely possible that the solution can be simpler yet.

Still to be looked at:

• Proper isolation distances. At least one of the amplifiers will do more than 200Vpp.
• Bandpass filter. It is only to be used with audio and also to filter the class-D switching frequency.
• Low power. I would like for a pair of 9V batteries to last a day at least.
• Be able to drive a very long cable at the output. No telling where the amp/loudspeaker will be in relation to the PC doing the filtering.
• Safety. I don't want this to destroy neither my friend, his equipment or the mega-dollar soundsystem this will be attached to.

Please share any advice or comment on my ideas.

Daniel

[Mechatrommer]

you got it somewhat right on the the input attenuation end, except you'll need trimpot and variable capacitor on either R3 or R4 for input matching. for audio BW i've done it with pretty much good (acceptable) result with TL074 opamp iirc, the key to the success is fet input opamp i think (very high input impedance).

Does it have to be a monolithic difference amp or could it suffice to use normal opamp and then get precision reistors to form the circuit instead?

you can use 3 opamp instrumentation amplifier for matched input impedance if you like... i used this setup with the jellybean TL074

Do I really need to be able to withstand very high common mode voltages?

your attenuation above will withstand high CM ±160V already. your attenuator is the key here..

Do I need very high input impedance at this point?

yes, any slight loading to R3 and R4 will mess up the attenuation ratio, CMRR and everything. your suggested opamp is 500Kohm input impedance and that is bad imho.

The AD8479 is used for high side current monitoring where it makes sense, but I don't see that it is needed in a diff probe like this.

usually high side current monitor is very low impedance nodes (0.1 - 0.01 ish ohm), lower input impedance (Kohms) will negligebly affect those nodes, thats why 500Kohm impedance opamp is ok. and difference opamp like you suggested is not meant to couple with high impedance attenuator like you posted. and you must balance your attenuator to the unbalanced (unmatched) opamp input, iirc.

• Proper isolation distances. At least one of the amplifiers will do more than 200Vpp.

isolate the traces of your attenuator accordingly esp the 3Mohm lines and you should be fine...

• Bandpass filter. It is only to be used with audio and also to filter the class-D switching frequency.

use lowish BW opamp and you have your BW limit taken cared of automatic in the opamp

• Low power. I would like for a pair of 9V batteries to last a day at least.

yes this can be achieved easily with jellybean opamps, probably for more days.

• Be able to drive a very long cable at the output. No telling where the amp/loudspeaker will be in relation to the PC doing the filtering.

design for 50ohm probe output so you can transmit the signal to your scope with long 50 ohm coax and a 50 ohm terminator.

• Safety. I don't want this to destroy neither my friend, his equipment or the mega-dollar soundsystem this will be attached to.

properly made diff probe will unlikely damage anything. you only need to take care of your probe end where it clip somewhere, too much metal expose on the clip may short something.

btw i recently trouble shoot class D audio amp with only a DSO alone. monitoring high and low side mosfet gate, 10V riding on ±80V, switching between DC and AC coupled probe. i even got shocked by the rail capacitors, fun project why you really need diff probe for this? its not like the mosfet/bjt working in linear mode that you have to probe the gate voltage precisely? anyway you can check this long dead and halted "build log" thread to get an idea on how it is done...

fwiw, ymmv. best regards.

edit: oh btw. class D amplifier will probably switch the gate at 10X the audio BW, so expect 100KHz switching, and then another 10X to monitor switching rise/fall time. so anticipate for a MHz diff probe BW. ymmv.

[fourtytwo42]

Sounds (pun) like a lot of work to me, I would just use the two channels of a scope in differential mode. There should not be any substantial carrier frequency present as that should have already been removed by the amplifier output low-pass filter.

[Mechatrommer]

I would just use the two channels of a scope in differential mode.

i tried that on DS1054Z... useless! update rate of math channel is like 1 fps, you only see math function of signals seconds ago.

[fourtytwo42]

I would just use the two channels of a scope in differential mode.

i tried that on DS1054Z... useless! update rate of math channel is like 1 fps, you only see math function of signals seconds ago.

I do apologise I have been the owner of a tek465 for over 20 years and it does things like that with it's eyes shut (figuratevly). I just tried my owon vds1022i and that doesnt seem to have your problem so maybe the settings of your 1054 (that I have no experience of) are awry somewhere or maybe you can get access to a different scope ?

[Mechatrommer]

yes it can work that way with other scope

[ZomBiE80]

With a 'scope, you should NOT need an isolated differential probe. Amplifiers are easy to be probed with an oscilloscope and a normal probe. You might only need to isolate ground.

[mk_]

With a 'scope, you should NOT need an isolated differential probe. Amplifiers are easy to be probed with an oscilloscope and a normal probe. You might only need to isolate ground.

you should read the first post...

you should try to understand what a mess a classD amplifier does with signals when probing at the switching nodes using normal probes to other inputsignals on a scope.

and you should also try to understand what a diffprobe does

isolating ground (what ever you want to say with this words) is an absolute nogo whe working with high voltage/high frequency (more then 50V, more then 100kHz) and even simple diffprobes are a cheap and secure way to measure.

Michael

[ZomBiE80]

What the crap are you talking about? I've measured many class-D amps  with ordinary probes. Maybe understanding signals beneath the noise would help? Also, I've measured PSU's mains with isolated scope ground, or with an isolation transformer when neccessary. And he was measuring OUTPUTs.

[ZomBiE80]

I would actually build a load that somewhat simulates a load that simulates a speaker load. A coil with iron core and a resistor with decent power handling capability would do. Measuring output of a D-class amp without a suitable load would indeed induce noise that you wouldn't want to see.

[mk_]

What the crap are you talking about? I've measured many class-D amps  with ordinary probes. Maybe understanding signals beneath the noise would help? Also, I've measured PSU's mains with isolated scope ground, or with an isolation transformer when neccessary. And he was measuring OUTPUTs.

If someone in my company does measurements as you descibed with groundlift (= no protected earth on an oszilloscope which is not buildt for this)... this person will receive a written (and harsh) warning, some kind of intense teaching and will not be allowed to do such measurements without a tutor in the near future...

And yes, we do really fast stuff with several 100V and several 100A. Primary fuses are 125A at the 400V AC-Side to get an idea about the powerlevels at the oputput.

Michael

[Twellmann]

Hi again, thanks for the input and discussion so far.

I have looked a little bit at your suggestions and have modified the input stage. I still need to look at the Bandpass filter and line driver (if needed).

The measurements will purely be done post LC filter in the amplifiers and only in the audio range. So I don't feel like I need BW to look at any switching flanks, but I might be mistaken. The probe will never be used for design or repair of the amplifiers. The output from the probe will be fed into some form of data acquisition system, likely a NI USB-4431 or something similar.

I have looked at the AD8429, but something less exotic might just as well fit the bill. I prefer a complete IN-amp as opposed to the 3 op-amp 'discrete' version. It is a bit cleaner and takes up less space, it might be a bit more expensive, but...