Unstable Class-D Audio Amplifier (Velodyne Active Subwoofer)

[EHT]

All,
I’m having trouble diagnosing a fault here and would welcome some advice or a schematic, even of a similar amp. I have checked quite a few things.

Type: Velodyne SPL-R 1000 Active Subwoofer, circa 2005.
Symptom: Extreme distortion and low output level

I cannot find a service manual or schematic (asked Velodyne, no answer). I doubt a service manual even exists because the amp clearly was not designed to be taken apart! It is glued all over the place and needed components to be removed in order to open the metal shield. I have managed to set it up such that I can power-up the amp with a sig-gen as audio input and probe it with my scopemeter & insulated probes (the power amp is live and high voltage – no isolation).

It has a pre-amp which includes a DSP chip. This is powered from a small linear PSU of +/-15V or so. The preamp works OK. The remote control functions work and the signal output from the pre-amp is undistorted. The amp PSU is rectified 240V mains into a bank of 6 x 470uF electrolytics, setup as 2 series groups (I refer to the centre of this as a “0V” reference). The voltage is stable at 340VDC as expected.

The amp itself is a Class-D Half-Bridge with a pair of IFRP350A MOSFETs which feed a large toroid inductor, with the speaker connected from this to “0V”. These are driven by an IR2113S chip. There is a further pair of IFRP350A which are connected between the Bridge output and 0V and appear to be driven by two IR2101s chips. I cannot quite workout the schematic of this section; it seems rather unusual - it does not make a standard full bridge for instance.

The PWM is created by a circuit containing an HC11 and HC27 triple AND/NOR gates and HC14 schmitt trigger. It doesn’t seem to the “standard” triangle wave generator and comparator. Aside from these there is just one opamp (TL061). This appears to be setup as an integrating amplifier (capacitor in -ve feedback loop) which I believe is used to integrate the PWM signal to give -ve feedback overall. There is a negative feedback connection which goes back to the bridge output through a decoupling cap and also a 1M resistor which links this into the depths of the logic circuit.

The TL061 is supplied by its own makeshift PSU consisting of 2 resistors and 18V zeners giving a +/-18V supply centered on 0V. The HC logic chips have a 5V supply which is referenced from the -ve PSU rail (not the 0V centre). There are 2 x 78L15 regulators which I have not worked out quite how they fit in..

Here are my observations:
-   The main power, logic supply and op amp supply are all good and stable
-   The whole amplifier is unstable. Under quiescent condition with no input it is NOT oscillating at a stable PWM carrier frequency. I am expecting to see PWM oscillation at a few hundred kHz.
-   It sporadically bursts into oscillation, each time for only a fraction of a second. The frequency of this is not stable, some bursts are 500kHz. The bursts will happen about every second or so, but randomly.
-   The MOSFETs gate drive and the inputs HIN, LIN to the IR2113 are all consistent so none of these parts appear to be malfunctioning.
-   When I look at the TL061 the problem is apparent. The opamp is going into saturation. If I apply a low frequency audio signal, the output of the TL061 is clipped to max voltage. Without an input it typically drifts to hit the rails but the sporadic oscillation will bring it back. This is consistent with not seeing a stable PWM signal on the input or output (gate drive) of the IR2113.
-   The TL061 is clearly going into saturation due to the feedback signal from the power amp.

I can attach a few sketches of the circuit that I have traced out. It is the same amp as this chap shows:
I have not replaced the caps he mentions, but they all read at least 33uF in circuit. Perhaps I should just try that. One did seem to have a bit of residue on the top. Unfortunately he doesn’t even say what problem he thinks this fixes.

So, I’m wondering if I now need to try to trace the circuit for the logic section as well as the input opamp. It is very hard going as it is tricky to access the components and many tracks are obscured.

I’m also wondering if it will tell me anything if I break the -ve feedback loop from the power amp back to the input integrating amp? On the other hand I’m also wondering if the fault it because this -ve feedback isn’t working properly.
Overall, I don’t understand how the PWM oscillation is meant to work in this circuit, and I don’t understand what the 2nd pair of MOSFETs which are not in a bridge are doing.

Thanks for reading!

[duak]

I repaired a Bose automotive amp that used a "Self Oscillating PWM" circuit.  I found a few examples of the concept with circuits on the 'net so with any luck you can get the flavour of how these are supposed to work.

There was no triangle wave oscillator or comparator per se and there was no switching (zero output) with zero input.  As soon as the input signal went away from zero, the appropriate plus or minus FET was activated until the output corresponded with the input and the FET was disabled.  I suppose the high gain and time delays in the circuit were worked out to ensure the the FETs were switched solidly.  I recall having to replace the FETs and at least one electrolytic.

Best o' luck,

[EHT]

Thanks very much Duak.
The tip to search for “self oscillating PWM..” was very helpful. It certainly bears similarities with what I have found so far, in that there is a negative feedback loop to an integrating op-amp, but no comparator or carrier oscillator visible. I found this from IR who make the driver chips for instance:
http://www.irf.com/product-info/audio/classdtutorial2.pdf
On p34 it points out that the frequency changes with the modulation, and I am observing the carrier frequency different at different times.

One thing I am wondering is whether I am making it different/worse buy running it with no load. What I’m looking at on the scope appears to be similar to what I was hearing when the speaker was connected. I suppose I should connect a resistive load or wire the speaker back in. Anyway, it definitely looks wrong because it is unstable with no signal input. I will trace out more of the circuit and maybe replace those small electrolytics as a matter of course (they do look cheap and not highly rated).

This is reminding me of when I fixed an SMPS in an LCD monitor which was unstable – it was oscillating sporadically at a low frequency and it was very hard to track down because the instability was visible through the whole circuit, again due to feedback (for constant-current drive). The fault in that was a shorted output on a driver chip...

[EHT]

Ha. I replaced one of the 33uF 35V electrolytics with a new 47uF 50V i had lying around and the amp has come back to life! The removed cap reads 29uF on the DMM.. High ESR? I haven't located the function of this cap in the circuit.

The signals in the circuit are about 100kHz and stable. I'll replace all the others and see if all the distortion is now gone.

Can anyone recommend a suitable glue & glue gun to secure the components again? (it was used extensively, as it is subject to lot of vibration inside the sub cabinet).

Thanks!

[SoundTech-LG]

Having been Velodyne's World Factory Service Tech for several years, I can tell you that these boards are inherently unstable. The traces on the inner layers of the board short out after arcing, and carbon trails.
Be careful with this no transformer module the 350V un-isolated from the mains is a killer. Bad design from the get go (that's why they  :--moved on to vehicle laser systems). Move on to something more conventional. They did...

[EHT]

Thanks for that SoundTech

Yes I realise it is very dangerous. I have been testing it with a battery powered scope/meter and have it screwed down on a wooden board so it can’t slip off the bench. Actually, the thing seems pretty well made but absolutely not designed to be serviced.  (Out of interest the PSU and Power Amp boards seems to be only double sided – I’ve traced quite a lot of the circuit out).

Latest in the saga for me is… I’ve replaced the 5 x 33uF 35V caps in the power amp with decent low-ESR 63V ones. The amp now seems to work. Unfortunately, after I put it all back together and started listening to it for a while, it cut-out after about 15 min and then I smelled the dreaded smoke. Took it apart. The MOSFETs, the output inductor and the metal case were red hot. The plastic holder for the inductor’s bolt has burnt/melted which is what the smoke came from.

I think what happened is that the metal box enclosing the power-amp has shorted the bolt which secures the output inductor to the top of the case (it is really cramped). This will have been like a 1-turn short turning the inductor into a transformer. Luckily the amp has a thermal cutout (I can see the thermistor on the MOSFETs). I’m hoping it has survived OK. I hope the enamelled wires in the inductor are fine and haven’t shorted but rather hard to tell.

This brings me to reflect on the quality of build and safety of these subwoofer amps. I have another one I’m looking at made by a different brand. That one is much worse. It clearly has no thermal protection as the PCB is partially incinerated. Again the output inductor look like it has been heated up, probably well over 100C. Multiple electrolytics have burst and the output relay is melted. This thing is  a death trap. Unusually it has a 100% metal outer case, but this is not earthed (2-wire mains plug) even though it does have an isolating SMPS, so potentially could be. I will see if I can change that. There will be some float between the speaker –ve and speaker’s chassis but I think that is just the potential between Audio Ground and Mains Earth (unlike the Velodyne in which speaker –ve is un-isolated. I’ll start another thread on this once I start working on it.

Another thing I’ve been ruminating on is why we don’t see a bit more discussion of the quality of the circuits and components for Audio gear in some reviews. Most audiophile reviews are like astrology. Actually the 2nd subwoofer I refer to is an “audiophile” brand but the circuit is just standard stuff, the parts low quality and the safety compromised. I’ve just repaired an Arcam amp which I must say is very well built and has extensive protection circuitry in it. British made (perhaps ?  )

[EHT]

OK - I'd appreciate any view on what is "normal" please!

I have reassembled the amp & sub without the top cover of the metal box which could cause the short. I've run it for at least 30 mins. The original fault is gone and it is not getting smoking hot as it did yesterday. However, I still think its running too hot so I'd appreciate any view on that.

I'm playing music at low/medium volume (although I should say my test is "Alpha & Omega" - about the heaviest bass you can get!). The output from the sub sounds OK - not distorted. The main speakers are through a class A/B amp which is not showing any sign of even getting warm at this level.

In the Class-D Sub, MOSFETs are medium-hot - just too hot to comfortably touch, so i guess 50C. However, the output toroid inductor is really quite hot but not burning / melting like before. Probably 60-70C. I can't imagine this is normal for a Class D which is meant to be very efficient. Does this mean the inductor is damaged from when I overloaded it before??

This unit can apparently output 1000W (RMS?? really?) so it doesn't seem right that it is already so hot at low output. It is even getting hot at low volume. Can't be right

thanks!

[duak]

Most magnet wire these days is good for quite a bit more than 100 C.  If it turns a darker brown or the yellow tape melts then it probably got too hot.  If the core is ferrite then it could have got too hot and cracked but otherwise I don't think it could be damaged.

Are you able to measure how much current is actually flowing in it when it's putting out a reasonable amount of power?  The low frequencies will cause heating in the wire due to I^2R.  Switching frequencies cause greater heating in the wire due to skin effect plus also cause heating of the core due to core magnetic losses.  If the core is powdered iron and is conductive an induced current also causes heating.

You should be able to view the inductor current waveforms by running a few loops of wire thru the core then terminating with a 50 ohm resistor to make a current transformer.

Cheers,

[EHT]

Thanks again for your help Duak.

The inductor looks undamaged other than the remains of the melted plastic plug. No sign of cracks in the core. I have observed that the heating seems to be regular all around the toroid, whereas I guess if there were shorts, this may not be the case. I think my hypothesis about the bolt shorting it out must have been correct because although I’m saying it is still hot now, it was far hotter that time.

The heat I’m observing now is not enough to cause damage, but I’m concerned this is from use at a relatively low volume. This amp is about 10 years old and I don’t see any sign of discolouration around this toroid, which I would have thought there might be if it usually ran hot and is old.

What I will do first is:
-   Run it for > 10 min with no signal and check it stays cold
-   Run it for > 10 min with very low signal and check it stays cool
-   Measure temps using IR thermometer
-   See if I can squeeze in an extra winding on it as you suggested.

BTW, the output inductor core on the other subwoofer I referred to IS cracked, so I presume I need to replace that, not just glue it? I might see if it can be replaced with a standard EI shaped core that can be panel mounted for better heat dissipation. These subwoofer amps have no airflow due to being sealed inside the speaker cabinets which are also packed with wadding. Would be much better if they were external to the speaker!

Regards

[SoundTech-LG]

"This unit can apparently output 1000W (RMS?? really?) so it doesn't seem right that it is already so hot at low output. It is even getting hot at low volume. Can't be right "


Yes, but only in your dreams. I have measured these 9 ways to infinity. 175W RMS max is it. They stretched the spec for you.
Again, best to move on from this P.O.S.

 

[EHT]

I'm sure the 1,000W claim is nonsense but it did manage to put a lot of current through the short circuit!! I get your point but I'm not for quitting, so not giving up on it yet.

Here's where I am: I have fixed the original problem by replacing the 33uFs but i have definitely damaged it by shorting the inductor.

Previously i had run the circuit with no load and no heatsinks on the MOSFETs. They didnt even get warm to the touch. Now, I have tried it with no input signal and after 10 mins it is getting pretty hot. MOSFETs are about 35C with heatsink attached. Inductor measured 57C. Ambient temp 21C.

I have noticed the output capacitor is now showing some burn marks so must be ruined. It is a 4u7 100V large ceramic. I think the inductor has survived OK.

Next i will check the waveforms/voltages on the bridged MOSFETs. I guess there is some imbalance or DC present. The MOSFETs and output cap don't measure as shorted with DMM. Maybe one or more of the MOSFETs are damaged... they got very hot and presumably were handling a current well above their rated max.

[EHT]

OK.. I have checked again with the scope, with no load (speaker) attached. With no input the circuit is sometimes not oscillating and sometimes it oscillates at the switching frequency (~100kHz).

- In the case it does not oscillate the bridge output (input to the inductor) gradually rises from 5 to 20V DC over the course of about 30s. Clearly it shouldn't do as this is passing on DC to the speaker output. However, I suppose this could be somewhat due to no-load (?)
- In the case it does oscillate I see a driving waveform of above 5V on the Vgs of the MOSFETs. The output of the bridge has about 300V p-p squarewave at the switching frequency. So it looks to me that the MOSFETs are functional and I noted they are rated to 150C.

I'm going to replace the main output cap which is toasted, then see. I might order spare MOSFETs at the same time as they are only £2.30 each.

For anyone following this saga, I am officially a MUPPET . In the process of trying to check the speaker was still OK after perhaps having DC through it, I shorted out my lovely working amplifier and totally wiped out one channel  I can't believe i was so careless and stupid. Flames even came out of the vents!! So that it another one to fix. Lesson: a hi-fi amplifier is not a piece of test equipment!

I've got 2 amps I have burnt out and another one which someone else has already done for me. The one i've just damaged uses Sanken SAP15N/P which (used to) contain emitter resistors. I've found that these now need to be replaced with STD03N/P plus external resistors. Unfortunately this blaze took out a whole bunch of other components 

[abraxa]

In the process of trying to check the speaker was still OK after perhaps having DC through it, I shorted out my lovely working amplifier and totally wiped out one channel

But... how? Was the speaker damaged?

[EHT]

No, i think/hope the subwoofer's speaker is OK. What i had done was wire up one channel of a stereo amp to the subwoofer speaker. It seems no where near as sensitive as the mid-range speakers so i turned the amp up quite high. I carelessly had a loose lead on the other channel of the stereo amp, and the loose ends caused a dead short  Took me a little while to realize that the channel which was destroyed was not the one with the subwoofer's speaker connected.

I've ordered a lot of spare parts, including an 8 Ohm 50W resistor to use as a dummy load while testing the sub amp...

One thing - I couldn't find an ideal match for the output capacitor in the sub. It is a 4u7 100V ceramic, over 20mm long.

The best i could find on Digikey or Farnell is this (which i ordered):
https://product.tdk.com/info/en/documents/chara_sheet/FG20X7S2A475KRT06.pdf

Which is an X7S type. I was ideally looking to upgrade the voltage rating and/or temp coefficient type, but that's all they had. It is much smaller - only 5mm or so. I ordered the best quality "COG" type for the other sub I have, but that only is only 100nF.

.. and here's an interesting article on ceramic caps showing how they de-rate very markedly with temp, DC bias and frequency. In fact i think they are only the specified capacitance when on test. Did VW design them?
https://www.maximintegrated.com/en/app-notes/index.mvp/id/5527

[SoundTech-LG]

Not sure why I bothered to dig this up, but here you go...

[duak]

I looked at the output stage on the schematic and see a unusual configuration with inductor L8 acting perhaps as some sort of transformer.  What it does isn't obvious though.  I wouldn't want to have to work on this thing if I could avoid it.

Best o' luck!

[EHT]

Hi Sound-Tech.

Thanks very much for digging that out. That looks like it. Mine is a version which has only 1 pair in the bridge, i.e. 1 only have one out of QP2, QP5 and one of QP6, QP1. I haven't spent any more time on it though.

I've literally just managed to fix the Arcam which i had set on fire. A lot of components were needed for that, but touch wood, seems to be OK now!

I get your point but I've already got the thing and I bought it 2nd hand, broken. I'll post some pics or start another thread on the other one I have as this makes this Velodyne one look like a work of art.

L1 was the inductor I shorted. I have a replacement for C3 which i will try soon. It is *way* smaller than the original one but seems to be same spec. 

So, I'm still not clear what the section around QP3, 4 is doing but hopefully I dont need to!

[SoundTech-LG]

The module is simply a big F'n Oscillator modulated by the audio input. L1 reduces the oscillations going out to the speaker, along with C3. When you shorted L1, C3 got the full F'n power of the oscillator. 

We replaced tons (50 gallon drums anyway) of those L1's. The cores were cheaply made, so then a bit more was spent improving it to try to reduce overheating failures.

[EHT]

Yeah, I see! Thanks for your help S-T, and Happy New Year!

Could you remember what sort of temp L1 would be when the amp is idle or on low output? I don't see any discolouration of the PCB or components around it ...

Anyway, I've made a bit of progress. I measured the inductance of L1 by removing one of its legs and measuring the resonant freq in a tank circuit. It measures exactly 150uH as per the diagram so I suppose it must be OK, not shorted. It still stinks of the burnt plastic plug though, yuk.

I replaced C3 and C1. However, I've now realised the cap I put in C3 is not right. I used a tiny ceramic. Now ordered a big fat Polypropylene film one:http://uk.farnell.com/webapp/wcs/stores/servlet/ProductDisplay?catalogId=10001&langId=44&urlRequestType=Base&partNumber=1854900&storeId=10151

Current status is that it does run and it can put some decent power through the load. However:
a) the output is distorted when the input signal is at low volume but then seems OK when turned up
b) the temp of the L1 inductor is still about 50-60C even at idle (MOSFETS 30-35C). This is measured with the case open.

I am thinking (a) is due to C3 not being right and / or other caps being heat-damaged. I'm not sure about (b). I'm going to replace C3, and then some of the others depending what result I get. I imagine that any parasitic oscillations are going to result in distortion of the audio modulation, so this might (?) explain what I've observed.

There is certainly some over-imaginitive design in here. For instance L3 is wound on L1 and it seems to just serve the purpose of supplying power to the two 15V regs. There are two or three signal feedback paths - R19, C17 and maybe R8...  Too clever for its own good i think.

[SoundTech-LG]

That L1 is always going to be warm/hot... remember it is reacting to the huge non-stop oscillation 24/7, with power on. You probably have the old style in there. The newer ones have slightly better cores, run slightly cooler, and able to survive longer. The secondary winding is fine unless L1 gets too hot, then the insulation on the secondary winding burns. Wonderful little time bomb. The main problem with these modules though is the arcing on the inner layers of the board. Once a carbon trail gets established inside, it's done.

[EHT]

OK, I've made some progress. I have replaced the burnt main cap (C3) with a polypropylene film one (see pic, old & new) and likewise replaced C2. I have also reset the settings in the DSP module via the remote control. Seems like the distortion i was hearing before was partly due to running the EQ calibration routine while the inductor was shorted, presumably resulting in some extreme settings.

With the correct type of cap in C3 it is running cooler now. I've run it for about 15 min on medium volume. It is perhaps working OK now. MOSFETs are only 30C, L1 inductor is about 60C. Perhaps this is normal? I'm still not convinced it sounds clean but that is partly due to physical vibration. Next I will:
- look at the quality of the output into resistive load with sig-gen input.
- check for AC at the switching freq on the main rails and around the regulators U14, U1 to verify this is not still a problem, i.e. whether caps C8, C5, C38, C39 may need replacing due to heat damage.
- order a new mounting kit for L1 so it can be boxed up again. I have found the standard type of indented washer here: http://www.conrad-electronic.co.uk/ce/en/product/734432/Sedlbauer-531707-Mounting-Set-For-1530VA-Toroidal-Transformers-1530-VA . It doesn't specify temp rating for the rubber washer. I will put Kapton tape under the metal washer as a precaution.

Having seen the circuit I suspect the original fault was the "bootstrap" capacitor C15 as I think that is working hard and it is supplying power to two of the driver ICs (U11 AND U5).

Pics showing the damage caused by shorting L1 are attached.

[SoundTech-LG]

I'd be perfectly happy with a nickel for every time I saw that same damage.

[EHT]

OK, I'm pivoting on this. Just letting writing up my views in case anyone else is thinking of fixing an active subwoofer.

So, I replaced all the electrolytic caps in the preamp as well, in order to get rid of the distortion which was audible at low volume. It seemed to do this, but now I can still hear some such distortion. This must be very high THD. Coupled with the fact that it reaches 70C idle when it is in the box I am giving up on the amp like SoundTech advised. Piece of junk. To think this sold for $1k originally; unbelievable.

I have another one which is even worse - a Gallo TR3. This thing also was originally sold for big bucks as a high-end piece of kit. It looks flash but when you open it up... cheap parts and not even any space for it to breathe with the wadding. It clearly has no over-current or over-temp circuit and has a metal case which is not earthed!! Should have a product recall IMO as it is simply dangerous.

In both cases i'm ripping out the amps and going to use as passive speakers. I just purchased reliable old Class B amps which have protection built in. So they have a big lump of iron and a big lump of aluminium in, but I trust they will be low THD and not a fire risk!

.. and we'll see if the Velodyne can handle the power. The amp I have is 600W RMS, not "peak music power" or any such rubbish. Real Watts.

Just working on setting this up now...

[dudaindc]

I'd be perfectly happy with a nickel for every time I saw that same damage.

SoundTech-LG,

Have you encountered a problem with the SPL Series where everything works, but there is a constant "low-level white noise" in the output?
The problem is not "perceivable" when the music source signal is above a certain level but clearly heard during quiet moments or between tracks.

To give you some more background on this issue I am struggling to solve, all electrolytic caps were changed to high-quality ones, both for the pre-amp and power supply/power output module.
I tested the power supply board with two pre-amp boards and eliminated the problem as being on the pre-amp side.

I then tested the power output board separately (using a jig I put together) with two power supplies (where the power output module is mounted on) and the problem remained the same.

Under the assumption (or logical conclusion) that the problem is with the power output board, I replaced the TL061 op-amp with a new one and no change in the "white noise".

Obviously, I tested all transistors and diodes on the board.
The board itself has no signs of arcing or any other typical damages.

I do have the schematics I got from Velodyne, but short from the shotgun approach (i.e. replacing all the IC's), I have no clue on how to fix this one.
Any ideas will be greatly appreciated.

Thanks in advance!

Cheers,

D

[EHT]

Mine did something similar in the end, once i'd fixed the other problems (33uF caps failed, causing severe distortion, burnt output cap after i shorted it). It made a quiet sort of grumbling noise and also the signal was distorted when the signal level was low. Like yours, this was not noticeable once the bass was louder. However, this made it hopeless for my general use. I wasn't sure if the problem may have been in the pre-amp or the power amp. I replaced all the electrolytics in the pre-amp (same brand as the failed ones)... but still that didn't resolve it. By that time i wished i'd listened to ST earlier.

My view is this is just a bad design, and I expect many other active subs are the same:
The combination of high frequency (100kHz), high current and high temperature (no ventilation) mean that many of the components are pushed to their limits shortening the life. I think this also means there is not a "binary" failure where a component fails and then the unit can be brought back to its original standard. Its like multiple components are degrading. Also this one is SO hard to work on..

So, i've gutted it and use it as a passive speaker. I've got a 2nd hand 550W Class AB amp which doesn't sweat driving it loud (barely gets warm) and has no sound quality issues. Of course the amp has a huge transformer and a huge heatsink. Must weigh 10kg or so. But it works!!

I'm doing the same with the Anthony Gallo sub. That thing is a death trap - no over current / over temperature protection and metal case with no earth.  It had a "double insulated" sign on it which literally just meant they put two insulator pads on the mosfets rather than one! I'm amazed that passed UK regulations.

Good luck.