Class D repair help please, 3525, IR2110, IGBT

[check12]

R237 (10R) only has 440mV across it
R248 (2R21) only has 80mV across it .... so not a lot of watts ......

Going back quickly to post #112, I should've added that those figures were no load & volumes at minimum, just in case that was important.....

 

[newvirus2008]

In the worst-case scenario, the case is at junction temperature i.e. RthJC is same as RthJA.

A small correction:

In the worst-case scenario, the ambient is at case temperature i.e. RthJC is the same as RthJA. Thanks.

 

[Rick PA Stadel]

That is an easier one to try (Z101, Z102, Z201, Z202) ... I have 1W through hole zeners in 5.1 & 5.6v here.

So you think that could significantly reduce the 2092's temperature without any other problems?

Glad you asked . . After a further think, that might not be the best idea. The specs say the Comp pin (which Peavey un-helpfully labels 'CLIP_x' (*1)) is allowed almost full excursion across ±~5V (positive V_AA to negative V_SS) supplies, so it might lower the 'actual' clipping threshold (without changing the behavior of the detection/gain-reduction circuitry).

!! NO! Still more hard-thinking needed:

OK, assuming I haven't butchered the arithmetic -- the detection comparators' thresholds are ~±2,55V, so there's a solid ~6dB to spare.

I guess, since you wouldn't need to remove the original zeners to tack in the lower-V parts across them, it might be worth trying. After all, it might be as much as ¾ of the ~40°C rise; and you DO have an accurate way to measure temp. If it was me I'd probably give it a shot. But don't let me talk you into it, especially since the Absolute Maximum Ratings (PDF page 6) show 20mA each for pins 1 and 6.

Or you could just check the voltages on those pins, if they're not too hard to get to, then make a reasonable guesstimate as to the added dissipation. One other thing: Those 1W zeners will probably not see the current needed to meet the rated voltage. There's only ~125mW at stake. And the heavier leads could be a real nightmare (for something temporary 'just to try' ). Then there're the facts that it is a working amp at this point(!), and Infineon's guy said 65°C was OK.

Best Regards

*1) then as they leave the board heading for the LM339's doing clip detection, 'CLIP_A/B' mysteriously become 'COMP_A/B' --

 

[Rick PA Stadel]

There's something else that I've been fighting against the urge to add, since it's not really essential to the repair. So ignore this if you want. Besides, I might be wrong . .

Earlier we were busy patting ourselves on the back over the reliability improvement contributed by replacing the IGBTs in the half-bridge main power supply with MOSFETs. Generally that may be true. But something to keep in mind -- the conduction of IGBTs increases with increasing temperature, whereas MOSFETs decreases. That means, as the power demands increase, the IGBT's dissipation increases somewhat less than proportionally to the output power (of the power supply, not the amp), but the MOSFETs run MORE than proportionally hotter . . Even though they run much cooler at low powers. And those IGBTs are rated to 175°C!

Considering the minimal heat sinking available to a D²PAK, I'm not sure that the cooler Idle/Low-Power operation, is a good trade for hotter MOSFETs under sustained high-power conditions. YMMV . .

Cheers

 

[newvirus2008]

If the MOSFETs were chosen correctly, they would match the IGBT drop, even at double the RDSon and double the current.

For example, if the amplifier draws 2000W power at 240V, the average input current would be around 9A. Let's make it 10A to make the calculations simpler. The chosen MOSFET has an RDSon less than 100mOhms, which may be assumed to increase by 100% to 200mOhms at high junction temperatures. The conduction losses in the MOSFET (that now dominate the switching losses) may then be calculated as 10*10*0.2 = 20W, which is similar to that of a typical IGBT with a drop of 2V at regular temperature, reducing to maybe 1.5V at Tj=125*C.

Now, coming back to reality, the amplifier isn't going to draw 2000W or 10A, and the RDSon isn't going to increase by 100% either. However, I still agree with your point regarding the D2PAK not having much scope for heat removal. Nevertheless, that would be equally applicable to the IGBT as well.

 

[newvirus2008]

Another difference is that the MOSFET uses its channel(s) to conduct the freewheeling (reverse) current which, in an IGBT, would flow through the accompanying diode, giving high conduction losses in the opposite device of the half-bridge.

 

[check12]

Been busy with other stuff the last few days ... the fets that replaced the IGBT's are running pretty cool here is thermal image after 40mins (8 ohm load, both volumes full clockwise, audio signal hitting the limiters). I think I already posted this but here it is again ... replacement fets outlined in blue (the tab of the top fet is the coldest area on screen @ 28.2c):

Still considering the alternative zeners but haven't had time yet ....

 

[check12]

Glad you asked . . After a further think, that might not be the best idea. The specs say the Comp pin (which Peavey un-helpfully labels 'CLIP_x' (*1)) is allowed almost full excursion across ±~5V (positive VAA to negative VSS) supplies, so it might lower the 'actual' clipping threshold (without changing the behavior of the detection/gain-reduction circuitry).

!! NO! Still more hard-thinking needed:

OK, assuming I haven't butchered the arithmetic -- the detection comparators' thresholds are ~±2,55V, so there's a solid ~6dB to spare.

I guess, since you wouldn't need to remove the original zeners to tack in the lower-V parts across them, it might be worth trying. After all, it might be as much as ¾ of the ~40°C rise; and you DO have an accurate way to measure temp. If it was me I'd probably give it a shot. But don't let me talk you into it, especially since the Absolute Maximum Ratings (PDF page 6) show 20mA each for pins 1 and 6.

Or you could just check the voltages on those pins, if they're not too hard to get to, then make a reasonable guesstimate as to the added dissipation. One other thing: Those 1W zeners will probably not see the current needed to meet the rated voltage. There's only ~125mW at stake. And the heavier leads could be a real nightmare (for something temporary 'just to try' ). Then there're the facts that it is a working amp at this point(!), and Infineon's guy said 65°C was OK.

Hmmm difficult to fathom the advice here ... ok the zeners are quite easy to install ... what are the potential cons to reducing the zener voltage?

Actually I'll check the voltages first later on today sometime ...

 

[Rick PA Stadel]

OK, good -- that's the best place to start.

Sorry about the crappy reasoning/writing. Feel free to ignore it . . won't hurt my feelings a bit !

The temps look great at 8 ohm loads. But by my experience, if an amp is rated into 2 ohm loads, sooner or later it's gonna see them!

Now that it's working, seems like some testing at heavier loads might be worthwhile.

Cheers

 

[check12]

Ok voltages are:
Pin 1 = 6.16V, Pin 6 = 6.22v ... (no load, pots full anti-clockwise)

The temps look great at 8 ohm loads. But by my experience, if an amp is rated into 2 ohm loads, sooner or later it's gonna see them!

That is what the original owner killed it with 1st time around (partially shorted voice coil impedance <2ohm).

Now that it's working, seems like some testing at heavier loads might be worthwhile.

I don't have two heavier loads but I could combine the 8's for a 4ohm load for 1x side.

 

[Rick PA Stadel]

OK, good. That's decent evidence that the zeners in the '2092 aren't adding too much dissipation.

Might want to consider acquiring some more loads, if you plan to continue servicing Class-D.

Also don't want to neglect to mention -- I hope it was just a 'typo', '(no-load, pots . . )'. Class-D should not be powered up without some load - - since the switching continues, with or without a signal applied. The filter components store a lot of energy that demands to go somewhere.

Other smarter folks on here can better suggest what a minimum (highest resistance) reasonable load might be, but 20 or 25 ohms might be in the neighborhood. Somebody correct me if I'm wrong . .

Best Regards

 

[check12]

Might want to consider acquiring some more loads, if you plan to continue servicing Class-D.

Yeah I agree ... I've only started to see class D a lot recently here, that's why I'm here trying to learn more about them.

Also don't want to neglect to mention -- I hope it was just a 'typo', '(no-load, pots . . )'. Class-D should not be powered up without some load - - since the switching continues, with or without a signal applied. The filter components store a lot of energy that demands to go somewhere.

Other smarter folks on here can better suggest what a minimum (highest resistance) reasonable load might be, but 20 or 25 ohms might be in the neighborhood. Somebody correct me if I'm wrong . .

I am aware that some class D IC's e.g. TPA3116, 3130 etc. require a load to be attached at all times as the filter is designed with the load as part of it ... but I have asked a few times about these pro-audio power amps and been told to treat them just like other solid state.

If that's incorrect and they do need a load attached all the time .... now would be a good time for me to find that out LOL ...

Certainly there is nothing mentioned in the manual and the specifications give power rating for 1x speaker ... see attached manual page 22.

 

[Rick PA Stadel]

I concede, there's nothing in the manual (didn't read every word, but read plenty!), but it seems like a reasonable requirement -- just the parasitic elements of the output filter alone ought to be enough to demand a dissipative component here or there.

There's bound to be someone on here willing to straighten me out . .

Cheers

 

[check12]

I concede, there's nothing in the manual (didn't read every word, but read plenty!), but it seems like a reasonable requirement -- just the parasitic elements of the output filter alone ought to be enough to demand a dissipative component here or there.

There's bound to be someone on here willing to straighten me out . .

Cheers

I want to know too ... who here knows about pro-audio Class D????

 

[Rick PA Stadel]

A minute ago as I was still studying the giant schematic, I noticed each output has a pair of 3k32's at the output PCB. That seems kinda high to provide the service we were talking about; but these filter components could be somewhat higher quality than most.

I'll keep studying. Gotta admit, so far most of my time has been spent on the myriad power supplies and protection circuits; far less on the actual amp stages.

Best Regards

 

[newvirus2008]

The amplifier in question doesn't have post-filter feedback and is therefore not capable of directly (accurately) controlling the voltage across the filter capacitor. It's therefore a good idea to have a load connected to it, as the same would enable damping of the resonance in the output filter.

 

[check12]

OK load will be used from now on ... is there any potential benefit in trying the lower voltage zener's ?

 

[Rick PA Stadel]

I don't think so at this point. The measured voltages were within 1% tolerance, and a later section of the PDF showed somewhat higher voltages than I was expecting for the 'internal Zener clamp voltage', considering the 'typical application circuits'. I think we can assume the external Zeners are handling the ~9mA (of the 19mA total the 620R's provide) that the '2092 isn't using.

 

[check12]

Ok ... I'll just leave it at that .

I've run out of free time anyway ... lot's of other stuff to do.

Thank you very much to all who commented! ... especially newvirus2008 who was on almost everyday for literally weeks ...

 

[Dyan]

Hi guys, hope u all are doing great. I recently sent my peavey IPR2 7500 for IR2110 chip replacement and the guy lost a silver capacitor (C1060 in schematics) near the chip, in the case of IPR 1600 it shows as C66 in schematics, i was wondering if any of you knows the value or where to get a replacement

I was also wondering if fets IRFB4227pbf can be used in IPR2 7500, both models share many similar components.

I really appreciate ur inputs.


Thanks in advance for your valuable help!
Regards!