Broken MOSFET power amp- MOSFET questions

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
JohnRoberts said:
Yup, it looks like they at least did design the bias pot correctly, when it fails (trimpots typically go open-circuit) the bias voltage drops to very low level.
Funny how often commercial designs overlook that detail. Not to mention some old (bipol.) designs that used diode chain + pot.
 
Whoops, it's actually an SVT4-PRO. Quite a different bit of stuff going on in my amp. Sorry for the confusion.
www.bicyclerecordings.com/various/428xxh0_.pdf

So, on my amp, the B channel was initially blown. R218 and Q202, and presumably the whole output transistor compliment. I didn't get around to checking this all out til later. I actually played with the amp a number of times on the functional channel. While I was testing I removed the heatsinks to try to get in there to trace stuff around better (one of them covers a few of the components) and about 10 sec after I turned the amp on I got smoke from the same part of the circuit of channel A.

Since people are looking at the schematic I'd love to have a little bit explained to me. What specific function do the zeners across the power amp input and output (D114, 115, etc) serve?
 
I've been checking around on the net for the 9240 devices and someone on another forum suggested eBay. There are a lot available there, all coming from China. Any experience/comments/warnings about buying power transistors on eBay from China?
 
If you look around the Vishay links I posted you can check stock levels.
 
I would check the electrolytic caps numbered C 111,113,114,112,211,213,212,214.  They look to be part of the RC timing of a protection circuit.

I would also check every power amp part after the 5332 driver with a meter.  Measure all the resistors, caps, transistors and diodes not that many parts.

I would buy maybe x2 the amount of transistors and do a match, or at the least match the ones you buy as best you can.
 
Matched parts would be nice, but perhaps not necessary, if they (each PN) come from single production batches. For drive symmetry you'd want to match P and N which seems unlikely. Matching Ps to Ps and Ns to Ns will improve current sharing.

The fairly large (.47 ohm) source degeneration resistors will help current sharing at higher current. Also the temperature coefficient of the devices (at least the 9240 data sheet I looked at) showed a roughly 25% resistance increase from 25C to 50C (eyeballed). So if one part turns on sooner and heats up more than others it will conduct less helping sharing with other devices. For drive symmetry there is a lot of voltage gain in the drive circuit so P to N symmetry should not be a major issues (and it will be biased into class A, as part of the class AB topology anyhow).

If matched parts are used there is usually a notation on the schematic, further the bias advice says to trim class A current for the "average" voltage drop across all the source resistors (per channel), suggesting that they don't expect them to be close enough to only measure one. You may want to replace any extreme outliers, you see during biasing, but the circuit looks like it will tolerate some degree of mismatch.

I am still not a fan of vertical MOSFETs for linear applications, but this should work as long as you have adequate compensation to keep it stable.. (C115, C116, C204). 

  JR

 
I think matching or at least checking for gross mismatch might be important.  As I posted before I matched power mosfets for a LED multiplexed driver circuit.  What I did not post was that the power mosfets were from the same batch and tube and unmatched there was a problem with the brightness differences between LED columns.

I can understand why people might think the same batch might be close but the batch I used there were differences and this was under 100ma.
 
Maybe Ampeg matched the devices and did not mark it on the schematic.
 
Gus said:
I think matching or at least checking for gross mismatch might be important.  As I posted before I matched power mosfets for a LED multiplexed driver circuit.  What I did not post was that the power mosfets were from the same batch and tube and unmatched there was a problem with the brightness differences between LED columns.

I can understand why people might think the same batch might be close but the batch I used there were differences and this was under 100ma.
 
Maybe Ampeg matched the devices and did not mark it on the schematic.

IMO they clearly don't expect the parts to be the same. I would not be surprised if they replace any really out of whack offenders that show up in production testing, and i would advise the same, but I wouldn't invest too much time or money into matching.

If they are a large enough customer they could pay the manufacturer to preselect or grade the parts.

=====

I happen to multiplex LEDs with MOSFETS and see no possible benefit from matching, but I use the MOSFETs as saturated switches, and the LED driver chip is current limited. I actually multiplex a 16 led driver IC 3x with high side MOSFET switches to control 3x12 LEDs with capability for 12 more.

Do you use current limit resistors in series with your LEDs? Are you running the MOSFETs from very low voltage and/or have marginal gate drive voltage? If you use MOSFETS that saturate with your available drive voltage and you manage the current with fixed resistors in series this should be pretty straightforward. About the only issue I encountered with multiplexing LEDs with MOSFETs was turning them off fast enough to eliminate ghosting in my display, due to the time/current required to discharge the gate capacitance.


JR
 
JohnRoberts said:
The fairly large (.47 ohm) source degeneration resistors will help current sharing at higher current. Also the temperature coefficient of the devices (at least the 9240 data sheet I looked at) showed a roughly 25% resistance increase from 25C to 50C (eyeballed). So if one part turns on sooner and heats up more than others it will conduct less helping sharing with other devices. .  
What the IRFP9240 data sheet doesn't show is how threshold voltage changes with temperature. That usually has a negative temp coefficient up to around a third or half of the maximum rated drain current, where it passes through zero and turns positive above that. That means the device could start to conduct more as it heats up if the bias voltage stays constant. The source resistors need to have a high enough resistance to prevent thermal runaway because of this.

This information seems to have disappeared from most MOSFET data sheets in recent years for some reason, but it's very important IMO.

Edit: My 1984 edition of the Siliconix MOSPOWER Applications handbook has a whole chapter on temperature compensated biasing of MOSFETs in linear amplifiers, and covers this stuff in some detail. A quick Google search shows that used copies of this excellent book are still readily available.
 
bobschwenkler said:
I've been checking around on the net for the 9240 devices and someone on another forum suggested eBay. There are a lot available there, all coming from China. Any experience/comments/warnings about buying power transistors on eBay from China?
Personally I'd be wary of buying them. There are a lot of counterfeits around these days. Any particular reason for not buying from known US distributors like Digikey or Mouser?
 
Thanks for the book suggestion. Lots of cheap copies available used. I'm going to get one.

And I'll ask again, the post may have been overlooked, but I'm can't quite figure out what specific function do the zeners across the power amp input and output (D114, 115, etc) serve? Protection of some sort it seems, but from a specific user error or failure mode?
 
johnR said:
bobschwenkler said:
I've been checking around on the net for the 9240 devices and someone on another forum suggested eBay. There are a lot available there, all coming from China. Any experience/comments/warnings about buying power transistors on eBay from China?
Personally I'd be wary of buying them. There are a lot of counterfeits around these days. Any particular reason for not buying from known US distributors like Digikey or Mouser?

The 9240s are out of stock from both. Digikey lists a shipment coming in this month, so maybe I'll just order from them and wait it out.

Edit: I take it back. Mouser has them both. They must have just got in the 9240, it was out last time I checked...
 
bobschwenkler said:
And I'll ask again, the post may have been overlooked, but I'm can't quite figure out what specific function do the zeners across the power amp input and output (D114, 115, etc) serve? Protection of some sort it seems, but from a specific user error or failure mode?
They protect the MOSFET gates from excessive drive voltage. Lateral MOSFETS would have them built in.
 
The zeners are to protect the power MOSFET gates that fail above +/-20v.. While in general the current limiting should become active before that much gate voltage is present, the zeners become an extra level of protection and perhaps some crude instantaneous peak current limiting, since clamping the gates at 10V limits peak current, and there is some lag in the current limiting circuit. 

====
Thanx.JohnR .That actually sounds familiar but recently I've been clearing my bookshelf of old data books. These days you can find stuff so easy with google and the WWW.

I'm not very experienced with using vertical MOSFETs as linear devices. I recall a pretty decent characterization from Hitachi about the lateral devices... IIRC from that there was also a sign change in the tempco wrt drain-source current but at lower current.

The typical transfer curve on the IRF data sheet shows curves for 25'C and 150'C and they appear to cross. I suspect this intersection of the two curves is the zero tempco point, but eyeballing again, up at a few amps so to high to be useful in class AB biasing. So yes these too could suffer thermal runaway at lower current levels if not degenerated and adequately heat sunk.

Do you have a opinion to break the tie about the need to match these power devices?

JR


PS: Coincidentally I am in the middle of converting a MOSFET high side switch into a simple pass regulator to scrub off some excess, no-load (fresh) battery voltage (or trying to). 
 
JohnRoberts said:
Do you have a opinion to break the tie about the need to match these power devices?
It might be nice to have matched threshold voltages for better current sharing, but I'm not sure how much variation there is with these devices. As you pointed out, trying to match P to N for better signal symmetry would be futile, as they will be quite different (more so than with lateral MOSFETs). I'm not a big fan of vertical MOSFETs in linear amplifiers either, and I'm pretty sure they only get used as a cost-cutting measure.
 
Back
Top