Help repair a D class amplifier NAD D3020 V1

[robca]

Some time ago I got a NAD D3020 V1 amplifier. It's a decent class D amplifier, and very compact.

Like many D3020 V1, mine had dim LEDs that made the display basically useless (the more a LED was used, the dimmer... so for example the never used inputs still had some luminosity. I found that Inolux IN-S63AT5UW is a replacement for the original WR-PB1608UW-M58 (which is not available), so after ordering 30 of these https://www.mouser.com/ProductDetail/Inolux/IN-S63AT5UW?qs=fAHHVMwC%252BbjOUaZe%2F%2FIGRQ%3D%3D and some SMD soldering, the display was as new.

But there is still a very annoying problem on power on, a high frequency audio tone that increases in frequency until it's inaudible. That seems to become worse the more I use the amplifier. If I stop using it for a few days, the problem is less obvious. Sometimes it goes away completely, sometimes it returns after a while

I tried measuring the output signal, and noticed that with no input and only the speakers connected, both channels have a 522-533 KHz sinusoidal tone at around 480 mV (I used a 1M Ohm scope probe, 1x)

I have only another D class amplifier, but that one has no high frequency tone on the oscilloscope. I know that D class amplifiers might have high frequency components well above the audio range, but an almost 500 mV tone on the speaker output seems really wrong.

The amplifier is based on a UcD102 module from Hypex, for which no schematics is available, as far as I can tell. I included the amplifier service manual with schematics and the only document I have about the UcD102.

The headphone output is fine, but that is not using the Hypex module. So I suspect a problem with the SMPS, but I wouldn't know what to check either

I know nothing about class D circuits apart from a general understanding of the basic principles, so I have no idea where to start troubleshooting. I'm hoping some kind soul here can provide suggestions on what to check, otherwise I'm stuck with just an headphone amplifier

[robca]

Adding the UcD102 document (too long for the first post)

[Circlotron]

That “522-533 KHz sinusoidal tone at around 480 mV”, is it identical on both channels? That is to say, could they be different and you are getting an audible beat note between the two tones?

[robca]

That's an interesting hypothesis, but the noise is mostly at the beginning (like a sweep when the UcD102 receives power initially, then goes away). In any case, the two channels are exactly 180 degrees apart.

I captured both channels, and you can see the waveform in the first two pictures below. It's a ~530KHz sine with additional noise overimposed.

I finally managed to get a good contact point in the SMPS and got readings on the +35V and -35V power rails (AC coupled), and there is significant ripple there, too (almost 800mV) I'm enclosing a screenshot for the +35V and -35V

I can't find any specs on the ripple of the Hypex SMPS, so I don't know if that is considered acceptable or not, nor I'm sure if such a high frequency ripple (~20x max audio frequency) is really the problem

If I can't get any pointers here, probably will end up selling this "for parts only". It mostly works, and it's a nice amplifier, but I doubt I'll be able to address the intermittent noise problem

[EHT]

The 500kHz is the switching frequency (note it says so in the manual) so there is always going to be some residual component of this frequency.
I suggest you move your scope to look at AF timebase to see if you can observe what you can hear on pin 7, 30, 17, 19, 21 of this module. If you can't see it on these when it is making the noise then I suppose it is a component in the module. I thought maybe the filter caps C403, 406 but if you're hearing in both channels then seems unlikely they'd have degraded in both channels.

[robca]

Thanks for chiming in.

There is no ripple on the audio inputs (which would be present all the time and in the headphone, too). Also, the ripple frequency is not influenced by the volume setting, for what is worth

Power supply (-35V, +35V) on 17/18 and 19/20 has the ripple in my previous message for the power supply (~700mV). I understand that 500 KHz is the switching frequency, but 700mV of ripple out of 36V seems high (2%) for an SMPS as well as >300mV of ripple high for an amplifier, but maybe not as much a concern for a class D one (as I said, I have no class D experience to draw on)

The amp runs very hot, after an hour of use the plastic case is almost uncomfortably hot (you can hold your hand, but only a few degrees from where you would want to remove it after a few tens of seconds), so cap failure could be a possibility, but I would expect electrolytics to degrade, not a 10nF chip capacitor (and not both the same way)

[TheMG]

If the preamp is ok and the intermittent issue happens simultaneously in both channels, I would first suspect power supply problems. 800mV ripple on the output of an SMPS seems rather excessive. I would start by checking capacitance and ESR of all electrolytic capacitors in the power supply. Especially since this device runs rather warm, that makes caps degrade quicker.

I don't have experience with the UcD102 but have an amplifier with UcD400 modules. They do run warm at idle, these modules are only efficient at high output levels the idle power dissipation is pretty high, but doesn't go up much as you crank the volume.

[robca]

Thanks for the additional info on heat, I wasn't sure if my UcD102 was hot because faulty or just normal. Good to know it's almost surely normal

I think your suggestion is my best option, even if I'm not looking forward to it: in order to be able to remove the capacitors to measure them, I have to completely disassemble the amplifier. And it looks like a nightmare of interlocking pieces that are hard to remove. The worst part is that's impossible to test the amplifier while disassembled without special harnesses to connect all the modules in the open configuration

I guess that if I plan to work on the SMPS, I might as well buy all replacement electrolytic capacitors ahead of time: once a capacitor is removed from the circuit to measure it (with the risk of overheating it), might as well replace it with a new one.

[TheMG]

One of the wonderful things about an ESR meter is that you can generally measure the ESR in-circuit. Bad capacitors that are going to cause issues, are usually going to be waaaaaaaay high on ESR.

I also would check the electrolytics on the UcD modules themselves as well. With the heat, and also being subjected to high ripple currents, possibly made worst by the SMPS having a high ripple output, chances are they could be going bad too.

Looking at pictures... this amplifier seems to be in a plastic case with virtually no room for airflow? The datasheet for the UcD102 specifies idle losses at 6-7W. While that doesn't sound like much, add to that the thermal dissipation of the power supply and other devices, and with the fact that the heat has practically nowhere to go, things can get pretty toasty in there! Obviously not enough to cause short term problems, but electrolytic capacitors really don't like heat in the long term.

Oh and I measured the ripple on the speaker output of my UcD400 amplifier, both show around 900mVpp at around 440kHz. So that seems rather normal as far as ripple on the speaker output. Can't measure the SMPS as mine has an old fashioned 10kg chunk of steel for a power supply (not an SMPS).

[robca]

Yes, there is almost no airflow in the case, even if there are slits protected by a perforated metal. The SMPS is at the bottom, the UcD102 above it, tightly packed. One side of the device is basically a big aluminum surface which is on top if the amplifier is used horizontally, or on one side when used vertically. In the picture attached, the SMPS is on the right side (which is the bottom when vertical). The other side is one big PCB with most of the non-UcD102 electronics. The UcD102 is just under the aluminum heat sink, and connected to the main PCB with a weird  sandwich of 2.54mm long headers.

In order to take it apart, you need to dismantle everything, including undoing a lo of hex screw/nuts on power components. And the plastic on the connectors is pretty brittle due to the heat (even if it's only 6 years old), including the thin flex cables connecting a lot of sub-PCB and modules

I don't really have an ESR meter, the best I have is either the $20 tester https://www.eevblog.com/forum/testgear/$20-lcr-esr-transistor-checker-project/ or use a square wave generator ad the oscilloscope, but I'm not sure either will work in circuit