Bad SNR with WONDOM T-AMP HV 2 x 900Watt Class D Amplifier - how to improve

Hi folks,
I have a Stereo amplifier Board from Wondow - offering a whopping 2x 900W @6ohms and 10% THD.
AB32551 WONDOM T-AMP HV 2 x 900Watt Class D Audio Amplifier Board

The datasheet promises a SNR or 99dB (A-wheighted).
The datasheet notes a 450µV A-wheighted noise rms voltage - which would correlate to 45Vrms output voltage. The spec is with 120Vdc so the amp deliver arrount 84Vrms).

But when connected to a power supply (I tried various SMPS and laboratory supply) the output noise on a connected tweeter horn is substantial.
Even at 48V it is much higher than on comparable Wondom products (AB35511 WONDOM T-AMP 3 X 500Watt Class D Audio Amplifier Board).
I tried shorting the input jacks - without much of an improvement.

A request to wondom did not deliver any feedback.
Any idea how to improve or what to test?

1. How can I test the RMS noise with DIY equippment?
2. How can I simulate the A-weighted characteristic?
3. Is there any means of grounding required?

Help is much appreciated!
 
Hi!

Do you have an oscilloscope?

It would be interesting to know what is in the amp output terminals when there is no signal and with the input shorted.
Maybe there is an oscillation or something else.

You can try to put this board over a metallic plate, with an isolation material in between, and ground the plate (it can be an aluminium foil).
 
Hi Ron, I have an oscilloscope - but signals are small and the probes are single-ended.
Do you have any recommendations on how to measure differential output with potential shift of >30V?
Shall I use capacitors for level shifting?
 
Hi,

6kHz is something strange in a class D.

Let me think a bit. I have bridged class D amp and single ended scope probes like yours and I'll do some tests here.
I've got curious to find a way of measuring this.
 
My trueRMS Multimeter shows 6mV AC when AMP is actively switching.
But on oscilloscope I can see a single ended ripple (Output vs GND) of 400kHz with 1,7Vrms per output.
When measuring two outputs and substracting with oscilloscope - they are in phase and almost completely cancel each other out (less than 100mVp).
 
Hello!

I was playing with my class D (TPA3255 from FOSI Audio) and it is really difficult to measure noise at the output without filtering and converting the bridge signal to a single ended signal. I tried some passive hacks, but some other 60Hz noise comes in.
In addition, if you don't filter above 20kHz, noise from the power supply also shows up.
It's not easy 🙂

So I've built a quick differential pre-amp in a protoboard limiting the frequency from 5Hz to 20kHz (aproximate values).
Then things got more stable. And I noticed that you need to connect a load - I used a 8.2ohm resistor.
I'll put this in a box cause it's handy.

This is the schematic of the circuit I've built and tested.
I use my scope always without the cord ground disconnected - I know what I'm doing in the workbench and it is only for my own use.
I think this way we play safe and get a reasonable result to check noise. Input impedance is 3.6kohms so it will not interfer with the amp.

First the signal from the amp is decoupled removing the DC and divided by 10, so as to not overvoltage the opamp.
Then signal goes to the differential unitary gain pre-amp.
The 220pF capacitors do the low pass filter, limiting the bandwidth so as to remove the high frequency switching signal.

We can improve the low pass filter to be second or third order so as to cut sharply above 20kHz.
We also can put more gain in the differential amplifier if needed - just play with the opamp 4 resistors.

1738699430422.png
 
Hey cool! Thanks for the quick answer!!
I will have to see as I don't have DIP components here (only SMD) - so no quick answer.
Last week I already tried to measure with REW and an analogue front end - but if BW is not limited enough I am afraid to measure alias frequencies and won't be able to see the true noise. So a LP will be neccessary, I guess.

One thing that I observed and what does not make any sense to me - the noise spektrum arround 6kHz is moving all the time by 1-2 khz up and down.
So each snapshot the oscilloscope shows the spectrum at a different spectral position.
Can it be a mixing artefact as the AMP has a stereo output - that the noise is actuall higher frequency noise and mixed down by the two independant PWM frequencies?
 
Hello,

Yes, it might be an intermodulation between the high frequency switching frequencies.
Something like one is at 500kHz and the other at 506kHz.

Most chips have a feature to synchronize the clock in a master/slave topology so they go together.
Is it possible to somehow deactivate one of the amps just to check if this ghost frequency goes away?
Some kind of mute? If it stops, that's it.

Do you have the chip specs for your board? I couldn't find it. It just mention Sure T-AMP

Regarding measurements, TI datasheet, for example, mentions that a low pass filter is in fact necessary to measure class D amps.
It seams they have a special pre-analyzer filters when they take measurements.
See below:

1738751153261.png
 
I did it the easy way.
My TrueRMS multimeter in mV has a very good CMRR. When shorting the inputs connected on one AMP output terminal - it measures 0.0mV
I added a LP Filter (2x60R and 100nF) across the output terminals and was measuring a 2,6mV rms ripple across the capacitor.
So this abut 5x the specified noise figure.

Wondom sent pictures how they measure the SNR - under load with ca. 450W output power.
I wonder if SNR is better (dampening of oscillations etc) under high power load.
 
Hi,

Ok, so you've built a first order filter with around fc=13kHz [Fc=1/(2*3,14*120*100*10^-9)] so the 2.6mV is probably within the 20-20kHz range, which is in fact an audible noise.

Considering output voltage of 45Vrms, we would have 20*LOG(0.0026/45) = -84.7dB.
It's good at high power although more noisy than promissed.

But if we think about 1W at 4ohm, we would have 2Vrms, then 20*LOG(0.0026/2) = -57dB, which is just ok.

Good if you could measure with an oscilloscope so as to send to the vendor and try to troubleshoot it.
 
  • Like
Reactions: Aoxomox
But on oscilloscope I can see a single ended ripple (Output vs GND) of 400kHz with 1,7Vrms per output.
I saw that the claimed switching frequency is about 550kHz, but your measurement shows 400kHz, which could be the issue. However, the architecture of its Class D modulator is unclear. If it is self-oscillating, try increasing the voltage and check if the problem is resolved.
 
  • Like
Reactions: Aoxomox
I saw that the claimed switching frequency is about 550kHz, but your measurement shows 400kHz, which could be the issue. However, the architecture of its Class D modulator is unclear. If it is self-oscillating, try increasing the voltage and check if the problem is resolved.
Oh, I thought a higher Supply voltage woud increase the problem. I will try and report back.