Unwanted 600 kHz signal on output!! Amp2.

[ollebula]

I've just finished assembling my amp 2 kit (rev 2) and everything seemed fine, the amp left the mute state just as it should. I grounded the inputs (to AGND) in order to trim the output to 0 V but except for the small expected DC offset I also had a 2 Vpp ~800 kHz signal on both outputs.
How could this signal pass thru the output LP filter?
Anyone who knows how this is possible?
What else could have gone wrong?

I use +/- 54 V power supply with ordinary amp 2 setup (no high power configuration). The 5 V supply was actually 4.7 V because of a resistor in series (to measure current... will be removed).

(Only) the two middle FETs were getting warm after a while which is kind of suspicious I think.

Any tip is appreciated.

/Olof

[gmikol]

There may be something wrong...but there might not. Keep in mind that the output filter is 2nd-order with corner frequency of 100kHz. 800kHz is 3 octaves out from the corner frequency, so you would expect the switching residuals to be 36dB down. Before the filter the switching waveform is 108Vpp in your case. -36dB = 0.0158 so the switching residual is 1.7Vpp by this very simple math. I haven't built my AMP2 yet, but my understanding is that this level of switching residual is fairly typical across different Class-D type amps.

Good luck with the rest of your build.

-Greg

[ollebula]

hmm, you have a point there. But if it is as you suggest, almost everyone would have experienced this HF ripple, making it kind of difficult to trim the output DC to less than 10mV. I'll try to load the amplifier and maybe lower the outputfilter corner freq and see if anything happends.

I have another idea for the cause of the problem:
When powering up for the first time (no FETs attached), I accidently missed a loose wire on the negative side of the power supply. The amplifer was only powered with 0 and +54V. As I wrote in the previous message the middle FETs (the ones working on the negative side) got a lot hotter than the other two. Lets say I destroyed the TK2350 in some way, is it possible that the TC2001 still could leave the mute state? Wouldn't the current och voltage sensing be out of spec?

Is it possible to buy just the TK2350 chip? Jan!?

/Olof

[djQUAN]

I also have the same problem. although it may cause some EMI issues, I have found out that it doesn't seem to affect sound quality. my case is a little worse. mine gets to 3Vpp at one channel and 4Vpp at another.

setting offset is done by using a DMM. most rejects AC signals when set to DC so you can clearly see a DC voltage.

[ollebula]

It worked. I plugged in my MD and beautiful sound there was (not ... using crapy old guitar loudspeaker). But I'm a little bit concerned because the negative side FETs are still a lot hotter than the right side ones (trimed to ~3mV offset). Maybe it won't be a problem when a heat sink is attached. We'll see... I'm very excited.

I'll just have to accept 2V (0.5W won't destroy dome tweeters) HF-ripple as a normal output from an amplifier.

/Olof

[djQUAN]

600kHz will be too high for most dynamic tweeters. with piezos, I'm not sure. the voice coil impedance will be too high at that frequency to cause problems. well, that's what I think is happening.

as I have mentioned before, other than EMI, I don't think it will be an issue. :)

[ollebula]

Ok, the amp is almost finished. Sounds great. Output from the FET:s looks just like I can expect they should and I have not tangled with the output filter yet so the HF switch freq is still present.

My question is, if it's common to have a clearly noticable high frequency rippel on the output, how do you measure THD and CMRR and other quality measures? I might have access to a audio precision (AP)instrument and it would be fun to see if the amp performes according to specifications, but I think the AP will give the amp very low quality ratings... Or are there people out there who has an amp2 working, or any tripath-amp, without any HF visible on the output?