Class D amplifier output oscillating? (Mackie Thump18s Subwoofer)

[Yamin]

Hey guys,
I’ve got a Mackie thump18s subwoofer amp module in for repair. The initial problem was the protection mode is engaged with no sound. Someone else was in before me and managed to fix the problem. But it failed on the first gig, same problem.
I had a look and it looks like components around and probably the IC NE556 (U5) were replaced.
To start with I was probing around that area too. After an hour of troubleshooting couldn’t find any obvious issue. During this time I was turning the amp on and off so it was not on the whole time.  After a break, while I had the amp on I heard a very faint spark sort of noise (the sort of noise you might hear when a component burns and gives up) the amp came out of protection mode and the HV was turned on. I figured a shorted component which was causing the protection to turn on finally gave up. But as the HV was turned on, in couple of seconds I started to see smoke. Quickly turned off the amp.
The cause of the smoke was the glue on R174, and that resistor gets very hot the moment the HV is turned on. I probed that area and I measured ridiculous level of oscillation, I have attached a screenshot. I probed the same point on Channel 2 and it also did have a similar waveform but was very small in amplitude. I have two 8 ohm resistive dummy loads on each output, removing them cause the amplitude to go even high.  I had to do all the measurements with in minute of the HV turning on as the R174 gets very hot.
Any idea what could be causing this? What should my next troubleshooting point be? What is the purpose of R174? it forms a a filter with C120?
The amp is set at bridge mode and I noticed R175 was not populated on Channel 2.
On a second note, when the power is turned on, there is a time delay before the protection mode is off and the time varies each time. That is not the usual behaviour for these as far as I know, and I noticed when I probe with a multimeter the output of the NE556 which is on the measuring side for the current limit, protection turns on in a second and turns back off again. Happens every time when I touch the probe to the output pin. What could the reason for this be?
Thanks in advance.
P.S- for future reference- is it ok to work on class D amps with no loads connected? or like tube amps do they need a load?

[MathWizard]

Yeah R174 and C120 are some sort of filter or snubber, if they were around a SMPS I'd know more on what they are doing. That cap would prefer no voltage across it too, IDK what the DC voltages are on either side of the transformers, but it would help stabilize them too.

Have you measured their values out of circuit? It's rated 2W so that's pretty big. Maybe it's drifted. Or somethings somethings dumping too much power into it.

IDK what the expected operating temp should be, but in the name of cost savings, it could be pretty high too, rather than using a 3 or 5W resistor.

[Vicus]

L9 should have an impedance high enought to not let the most part of the carrier to pass over it so I bet L9. In this case the other symptoms are not related.

[temperance]

@Vicus

-you probably mean L7 instead of L9.

The switching waveform will appear at the output If C118  is open with no load attached. With capacitors being more fragile than inductors I would rather place my bet on C118.

[nigelwright7557]

With no load the output capacitor in the output filter can get an over voltage due to ringing.
Sometimes class d amps can have a little residual oscillation on the output.

[temperance]

With no load the output capacitor in the output filter can get an over voltage due to ringing.
Sometimes class d amps can have a little residual oscillation on the output.

That doesn't seem to be the case here because there is 50 Vpp at the output. See original post.

[CaptDon]

The 'self oscillation' frequency is probably different from the fixed class D PWM frequency. The resistor / capacitor series circuit on the speaker output is a stabilization circuit and yes, they do burn up if an amplifier is in an uncontrolled self feedback mode. Usually it will kill the output devices also. Some leakage of the PWM frequency is always present and is simply accepted as long as it isn't radiating 'outside the box'. One thing that seems to happen on these class D units is the output inductor cracking with a huge shift in inductance. Everything is up for grabs at that point. Check the PWM clock frequency vs. the self oscillation frequency. I expect the clock to be around 250KHz to 500KHz and the self oscillation frequency to be at 50KHz to 100KHz. Be very careful about 'Alias' if measuring with a digital sampling scope!! For audio work I refuse to use a digital scope, especially for class D troubleshooting. Grab the analog dinosaur scope!! My analog goto is a Tek SC504 80MHz although the SC502 15MHz is also perfect for analog / PWM. My digital buddies are a TEK TDS644B or my portable TDS2004B. One tip also on the Thump units, There is a TO-92 size transistor that is known to fail. I forget what it does but they fail in a weird way. They begin breaking down and causing huge transients in the output which trips the protect circuit.

[Yamin]

Thanks guys for your inputs. I came across some text about the resistor in the zobel network burning up due to oscillation. In that particular case the feedback resistors for the IC was way out of spec. I also had thought and had tested them in circuit and it appears to be same for both channels.
My next thought was to check if there was anything being fed in through the inputs to the switching IC. I figured I'd attach a jumper to the amp mute so that the HV does not turn on.

[Yamin]

The 'self oscillation' frequency is probably different from the fixed class D PWM frequency. The resistor / capacitor series circuit on the speaker output is a stabilization circuit and yes, they do burn up if an amplifier is in an uncontrolled self feedback mode. Usually it will kill the output devices also. Some leakage of the PWM frequency is always present and is simply accepted as long as it isn't radiating 'outside the box'. One thing that seems to happen on these class D units is the output inductor cracking with a huge shift in inductance. Everything is up for grabs at that point. Check the PWM clock frequency vs. the self oscillation frequency. I expect the clock to be around 250KHz to 500KHz and the self oscillation frequency to be at 50KHz to 100KHz. Be very careful about 'Alias' if measuring with a digital sampling scope!! For audio work I refuse to use a digital scope, especially for class D troubleshooting. Grab the analog dinosaur scope!! My analog goto is a Tek SC504 80MHz although the SC502 15MHz is also perfect for analog / PWM. My digital buddies are a TEK TDS644B or my portable TDS2004B. One tip also on the Thump units, There is a TO-92 size transistor that is known to fail. I forget what it does but they fail in a weird way. They begin breaking down and causing huge transients in the output which trips the protect circuit.

Thanks for the tips, fingers crossed that its not the inductor which has failed. Would it be a good idea to take it out circuit to measure its inductance? there is so much glue on it though.
I was hoping that you could explain what you have suggested about PWM clock freq vs self oscillation freq what could be deduced from the results?
And thanks for the tip about the TO-92 package component, there is one and its a thyristor used to shut down when there is a DC detect fault.

[CaptDon]

Class D is typically a PWM squarewave. When the duty cycle is 50% the output voltage is zero. When you shift say toward 60/40 or 40/60 duty cycle the output shifts toward either the + or - power supply as a function of the duty cycle. The frequency of the PWM remains constant at typically 10 times the highest audio frequency to be passed so 250KHz to 500KHz. When things go wrong and the amplifier self oscillates it often occurs at 50KHz to 100KHz. That is why I am suggesting you look closely at what the waveform looks like and the frequency. The Mackie Thump and its off brand relatives (there are many) were not over engineered, they were at best 'cost effective' for mobile D.J.'s and such. They really are not very robust if you push them to the limit and they may have some kind of built in limiter circuit because they really start to sound bad at high volume.

[temperance]

There is picture in the OP first post where you can see 50Vpp triangle wave with a switching freq. in the neighborhood of 400 KHz on the output.

[CaptDon]

50vpp at 400khz, That looks like an L7 inductor problem. One shorted turn? Cracked core? That thing is passing the PWM frequency right out the tail pipe!! No wonder the phase stabilizer circuit fried!!

[temperance]

The filter doesn't work because of either the inductor or the capacitor. My bet is still on the capacitor because a 0.47µF capacitor has an impedance <1R at 400 KHz.

Broken inductor: the amp would drive the 0.47 µF output capacitor with 50Vpp. That will not last very long as the peak current would be around 19A. (0.47µF * 50V / 1.2 µs)
Broken capacitor: the switching waveform appears on the output and the damping network (RC across the output) will go up in smoke. 

[CaptDon]

I took some time to investigate the circuit this morning and I have to agree with Temperance, the .47 capacitor must be open circuit or bad solder connection! The triangle waveform is the perfect example of a good inductor doing what is is supposed to do, create a current ramp from a squarewave. If the .47 was open circuited the snubber network would surely burn. Have a look at the solder connections around the .47 capacitor and the possibility of a failed capacitor. Cheers mate, I think you will have this thing fixed in short order. Sorry I didn't dive into the schematic details sooner.

[EHT]

BTW, it's not any old junkbox capacitor. You need one which doesn't lose too much capacitance with high temperature and high frequency. Caps are graded with a code for temp range and capacitance change. I think 'C0G' Class 1 is the common highest stability type for ceramic, but not sure this is the right thing here. Maybe someone can advise on the best type and rating of cap for this application before you buy one. I think this type will be considerably more expensive than a bog standard one..

[temperance]

BTW, it's not any old junkbox capacitor. You need one which doesn't lose too much capacitance with high temperature and high frequency. Caps are graded with a code for temp range and capacitance change. I think 'C0G' Class 1 is the common highest stability type for ceramic, but not sure this is the right thing here. Maybe someone can advise on the best type and rating of cap for this application before you buy one. I think this type will be considerably more expensive than a bog standard one..

0.47µF/100V or something in C0G? They do exist but they are rare and probably expensive.

A regular MKT, metalized polyester type capacitor will do fine.

Perhaps the OP can post a picture of the cap in question.

[Yamin]

0.47µF/100V or something in C0G? They do exist but they are rare and probably expensive.

A regular MKT, metalized polyester type capacitor will do fine.

Perhaps the OP can post a picture of the cap in question.

Thanks guys, it's a generic metalized polyester type cap (brown-reddish colour). I haven't gotten time to go back into the shop since my last post. Will be back in the shop hopefully after the weekend.

I also did wonder what would happen if another type  of cap was used in this application.

[temperance]

A MKT, metalized polyester capacitor will do.

[Yamin]

Thanks guys you were right it was the 0.47uF cap being way out of spec. I replaced it and the amp is working fine.

I measured a sine wave at the same point from post#1, on both the channels. You guys mentioned about this before but was hoping that you could elaborate as to why the sine wave exist here, I assume its normal.

The amp takes 7ish seconds to come out of protection on initial power up. I'm not sure if this is normal for these. I noticed that, if I turn off the power and turn it back on before what I assume is the main capacitor discharges its the protection is not engaged on power up.

Thanks again for your help.   

[temperance]

Of course I'm right 

The signal you see on the output, across the 470nF capacitor is commonly called the switching residual. With no input signal applied you will see a see a 50% duty cycle square wave on the left of L7. The square goes trough a low pass filter made up by L7 and the 470nF capacitor.

Let us calculate the switching residual to see if what we see is normal:
The LC filter (25µH with 470nF) is a second order filter and as such the roll of is 40dB/decade. The cut of frequency for the LC filter is 46kHz. One decade up in frequency is 460kHz and the attenuation should be 40dB or a factor 100 at 460kHz. With the supply being +/-60 V, you will a little over 0.6V of the original 400kHz carrier signal which seems to agree with the oscilloscope screenshot shown.


Protection: this might be normal. (although the circuit seems to have a design flaw in that the protection has not been properly reset when briefly interrupted)

Reference describing filters:
https://web.engr.oregonstate.edu/~webbky/ENGR202_files/SECTION%203%20Second%20Order%20Filters.pdf

[Yamin]

Of course I'm right 

The signal you see on the output, across the 470nF capacitor is commonly called the switching residual. With no input signal applied you will see a see a 50% duty cycle square wave on the left of L7. The square goes trough a low pass filter made up by L7 and the 470nF capacitor.

Let us calculate the switching residual to see if what we see is normal:
The LC filter (25µH with 470nF) is a second order filter and as such the roll of is 40dB/decade. The cut of frequency for the LC filter is 46kHz. One decade up in frequency is 460kHz and the attenuation should be 40dB or a factor 100 at 460kHz. With the supply being +/-60 V, you will a little over 0.6V of the original 400kHz carrier signal which seems to agree with the oscilloscope screenshot shown.


Protection: this might be normal. (although the circuit seems to have a design flaw in that the protection has not been properly reset when briefly interrupted)

Reference describing filters:
https://web.engr.oregonstate.edu/~webbky/ENGR202_files/SECTION%203%20Second%20Order%20Filters.pdf

Thanks man for the detail explanation. While we are on the subject, I wanted to ask if there is anything to keep in mind when scoping around Class D-amps especially at the outputs. Some texts suggest using differential probes etc..

Also could you please elaborate on the design flaw you mentioned about the protection circuitry. I've been trying to wrap my head around this particular circuit.

Thanks again for the help.

[temperance]

Thanks man for the detail explanation. While we are on the subject, I wanted to ask if there is anything to keep in mind when scoping around Class D-amps especially at the outputs. Some texts suggest using differential probes etc..

For a class D amp running at +/-50 V or so you don't need to use differential probes. At least I never do that. But you must inspect the probes you use to find out how much voltage can be applied versus frequency. Not all probes can measure such voltages at high frequencies. Here is an example:
https://www.pmk.de/files/downloads/PML%20Series%20Datasheet.pdf

The plots for frequency/voltage are shown on page three.

However, there is probably some switch mode power supply in this amplifier and probing around in those can be very dangerous if you don't know what you're doing. Here is a video explaining the how and why made by Dave himself:

Also could you please elaborate on the design flaw you mentioned about the protection circuitry. I've been trying to wrap my head around this particular circuit.

The "timer" in this protection circuit doesn't reset properly because if it would, briefly interruption mains by disconnecting the plug should reset the timer instead of generating a loud blob trough the speakers when the mains reappears.

In case you need this, some class D amplifier theory:

[Yamin]

Thanks man for the detail explanation. While we are on the subject, I wanted to ask if there is anything to keep in mind when scoping around Class D-amps especially at the outputs. Some texts suggest using differential probes etc..

For a class D amp running at +/-50 V or so you don't need to use differential probes. At least I never do that. But you must inspect the probes you use to find out how much voltage can be applied versus frequency. Not all probes can measure such voltages at high frequencies. Here is an example:
https://www.pmk.de/files/downloads/PML%20Series%20Datasheet.pdf

The plots for frequency/voltage are shown on page three.

However, there is probably some switch mode power supply in this amplifier and probing around in those can be very dangerous if you don't know what you're doing. Here is a video explaining the how and why made by Dave himself:

Also could you please elaborate on the design flaw you mentioned about the protection circuitry. I've been trying to wrap my head around this particular circuit.

The "timer" in this protection circuit doesn't reset properly because if it would, briefly interruption mains by disconnecting the plug should reset the timer instead of generating a loud blob trough the speakers when the mains reappears.

In case you need this, some class D amplifier theory:

Thanks so much