Poor performance from TI's TPA3251D2EVM

[matt09]

Questions on performance from TI's TPA3251D2EVM

I have a TPA3251D2EVM board that I'm sorting for a friend. I upgraded the op-amps to lme49720's and started testing with resistive loads for max power before clip LED illumination. The results somewhat lower than I anticipated for 1kHz @ 36V:

51W into 8 Ohm
95W into 4 ohm
123.5W into 2 ohm

I get 72W and 138W into 8 ohm and 4 ohm respectively before flat top clipping which is in line with the 1% THD+N figures in the datasheet. The waveforms however look far worse than 1% THD+N with a lot of switching noise, see attached.

There is also 1.2V pk-pk at ~570kHz imposed on the output signal at all times, irrespective of power output.

Does any of this seem familiar to anyone?

Thanks

 

[rayma]

Does any of this seem familiar to anyone?

Did you test the board before changing the op amp? The output filter also may not be good enough.
It's possible the 570kHz is common mode pickup. Is it still there when you short the scope probe tip
to the scope probe ground lead, and then touch both to the pcb ground? If so, it's common mode pickup.

 

[29285]

Did you test the board before changing the op amp? Your output filter also may not be good enough.
It's possible the 570kHz is common mode pickup.

What you see on the scope is NOT what you hear.
This is a switched amplifier, and so there is a residual clock ripple on the output. And this is not synchronous with your sine-wave, so it might look like some superimposed noise.
If you want to measure what you hear, a low pass-filter, ideally a 20khz brickwall filter, should be inserted between amp-output and scope input.
Otherwise there may be things you can hear, but not see on the scope. Small amounts of distortion, for instance.

All in all, take any scope plots with a grain of salt. It is not easy to
to conclude the audible impacts from what you see.

 

[matt09]

I'm not getting any common mode noise at the Gnd point, I'm seeing noise everywhere at all the supplies including the op amps and main supply terminals, but it's definitely not from my regulated linear PSU.
The superimposed sine wave was easier to measure with no input and is 600Khz, almost perfect sine wave with some crossover distortion. I don't know a huge amount about class D amps but I'm sure this is linked to PWM switching.

The mess at the top of the peaks in the image is at the 1% THD+N output, this isn't present at lower levels of output so I assumed it was distortion.

 

[DUG]

...

The mess at the top of the peaks in the image is at the 1% THD+N output, this isn't present at lower levels of output so I assumed it was distortion.

It could be the inductors operating into saturation...reducing their effective inductance...reducing effective filtering.

Try to relate current output with filtering effectiveness.

Increase loading (lower R) and check 570KHz signal level vs current to the load.

 

[doctormord]

I can't remember when i lastly heard 570kHz tones.

 

[Adam Reed]

Go look at TI's website. There is a very good white paper on how to build a simple filter to make it easier to look at your output signal on a scope. Pretty impossible to see anything without. I've used the same network with mu UCD amps as well.

 

[matt09]

Thanks I will build the scope filter, didn't realise this was necessary as assumed it would already be taken care of on the board.
This doesn't explain the early clip LED illumination though, unless it's similarly inaccurate?

 

[multisync]

as per TI data sheet switching freq is 600kHz. You will see the 600khz everywhere you probe with your scope.

7.5 Electrical Characteristics
PVDD_X = 36 V, GVDD_X = 12 V, VDD = 12 V, TC (Case temperature) = 75°C, fS = 600 kHz, unless otherwise specified.

when doing power tests are you measuring 36 volts as close to the TPA3251 as possible. There was a person here wondering why his voltage regulator wasnt working. Turned out his test leads were no good and couldnt supply the required current.

 

[matt09]

Indeed, I just didn't expect to see it so prominently on things like op-amp power rails and outputs. Yes voltages were taken with a calibrated DMM at the output posts.

 

[doctormord]

How did you connected the scopes probe?

Hint:

An externally hosted image should be here but it was not working when we last tested it.

Source: Understanding, Measuring, and Reducing Output Voltage Ripple - SIMPLE SWITCHER(R) FAQ - SIMPLE SWITCHER(R) - TI E2E Community

 

[29285]

Indeed, I just didn't expect to see it so prominently on things like op-amp power rails and outputs. Yes voltages were taken with a calibrated DMM at the output posts.

I experienced that the clipping LED of the original TI-EVM-board triggers way below real clipping. So you should not bother much about this.😉

 

[doctormord]

Yes, it roughly starts to lights up at -3dBfs.

 

[matt09]

Thanks for confirming what I suspected about the clip LED.

So the ~1.2V pk-pk at switching frequency into speakers at all times is also normal and to be ignored? I know it's not audible or going to damage anything but that's nearly 1/5 W into 4r.

 

[29285]

The presence of spikes everywhere is more often than not explained by a strong magnetic field, that emerges from the loop circuit consisting of input capacitor and chip supply terminals.
To check this there is a quite simple test: Take your scope probe, short the input with its ground clamp and sniff with this loop antenna close to your circuitry.
You will find these spikes without connecting the probe or scope either.

 

[29285]

Thanks for confirming what I suspected about the clip LED.

So the ~1.2V pk-pk at switching frequency into speakers at all times is also normal and to be ignored? I know it's not audible or going to damage anything but that's nearly 1/5 W into 4r.

Yes, it is to be expected!

 

[doctormord]

Again, how did you connected your probes?

 

[matt09]

Again, how did you connected your probes?

That is with the probe used directly across the output terminals, the clip and gnd wire are not used.

 

[Khron]

Your math's a bit off there - with the mH of inductance that a woofer has (or even the hundreds of uH of a tweeter), the impedance at Fsw is well into the hundreds of ohms, so that power damn near vanishes 😉

In the words of AvE over on youtube - "f**k-all in a big ship" 😀

So the ~1.2V pk-pk at switching frequency into speakers at all times is also normal and to be ignored? I know it's not audible or going to damage anything but that's nearly 1/5 W into 4r.

 

[SanjeevM]

The recommended voltage for 3251 if 36 volts. If a linear psu 10A rating is giving out 39v in open circuit, is it safe to connect it to the 3251 board?