Hifi amp output MOSFET matching (or lack thereof)

[klausman]

So I am about to replace a blown output MOSFET in a Sony TA-FB930R stereo amplifier. Specifically, the Right channel N-channel MOSFET is blown and needs replacement.

Here's part of the circuit diagram, with red dots next to the four MOSFETs (Q511 and Q512 on the left channel, Q612 and Q611 on the right one). x11 are the P-channel 2SK1529-Y ones, x12 are the N-channel 2SJ200-Y ones.



The blow MOSFET is the right-channel PNP 2SK1529-Y. I've used a PeakElec DCA75 to get some measurements of the surviving ones:



As you can see the existing, working MOSFETs have wildly different characteristics (or so I think). Does this mean that the replacement I am looking for can be equally "wild" (while still being within manufacturer specs)?

I know if there were MOSFETs in parallel (e.g. two 2J200s on a single channel), matching was important (usually Vgs and gm), but now I am wondering about this particular circuit and if I am overthinking the replacement of Q611.

[CaptDon]

I find in many of the MOSFET musical instrument amplifiers they are not matched at all. Unlike parallel bipolar devices parallel MOSFET's seem to survive nicely in parallel without careful matching. Even the old bipolar amplifiers with as many as eight devices in parallel didn't require more than a moderate match since the emitter swamping resistors helped equalize the load sharing. Most of the older MOSFET amplifiers, Ampeg in particular are really poor designs that can't control thermal runaway properly. I have resorted to installing fans on the heatsink, setting the bias so low that there is 'zero crossing' distortion until the amp actually warms up a bit. My personal experience has found most of the early design MOSFET power amplifiers are a smoking pile of burned resistors and shorted MOSFET's waiting to happen. (Unless they are class D switch mode amps)

[madires]

There's feedback from the MOSFET output stage to the differential amp stage at the input creating a feedback loop which deals with parameter variations of the transistors (within the limits of the feedback loop), i.e. the MOSFETs dont need to be closely matched (n and p-channel pairs). However, some audiophiles prefer matched transistor pairs. When paralleling multiple MOSFETs you would add source resistors (equivalent to emitter resistors of a BJT output stage) for compensating parameter variations.

[Le_Bassiste]

I know if there were MOSFETs in parallel (e.g. two 2J200s on a single channel), matching was important (usually Vgs and gm), but now I am wondering about this particular circuit and if I am overthinking the replacement of Q611.

in general, due to the high amount of negative feedback employed in such amplifiers, variations on gm, RDSon, VGSth and even Ciss aren't as much of an issue as one might expect. one really important property of the MOSFET that you want to replace is DC-SOA. make sure that the replacement has a DC-SOA at least as wide as the original one. unfortunately, this requirement will rule out many of today's MOSFETs that are optimized for switching- rather than linear-mode operation. a MOSFET that doesn't have a dedicated DC-SOA plotted in the datasheet is no option for linear musical amplifiers.

[klausman]

I know if there were MOSFETs in parallel (e.g. two 2J200s on a single channel), matching was important (usually Vgs and gm), but now I am wondering about this particular circuit and if I am overthinking the replacement of Q611.

in general, due to the high amount of negative feedback employed in such amplifiers, variations on gm, RDSon, VGSth and even Ciss aren't as much of an issue as one might expect. one really important property of the MOSFET that you want to replace is DC-SOA. make sure that the replacement has a DC-SOA at least as wide as the original one. unfortunately, this requirement will rule out many of today's MOSFETs that are optimized for switching- rather than linear-mode operation. a MOSFET that doesn't have a dedicated DC-SOA plotted in the datasheet is no option for linear musical amplifiers.

Fortunately, there is still NOS of these parts (though not as much as I'd hoped). I've just ordered one and hope is that it being an original* part, it's SOA values are the same as the blown one. Not that I could test them easily.

* as original as ebay parts can be. At least it's not super expensive. The wait time/delay would be more annoying if it's a dud.

[CaptDon]

I had gotten some darlingtons from Ebay shipped from china of course MJ10023, a Motorola designed part. When they arrived they seemed to be obvious fakes and the logo rubbed off with my fingers. I installed a pair that instantly blew up on power up. Much to my amazement I accidently ran across a tray of over 100 real original Motorola parts at my work place. They were ordered for a project that never went to market. Lucky me!!

[madires]

Also a lot of fake MOSFETs and JFETs on AliExpress, eBay and so on. Measuring C_GS helps quite often to identify fakes.

[klausman]

Also a lot of fake MOSFETs and JFETs on AliExpress, eBay and so on. Measuring C_GS helps quite often to identify fakes.

I will definitely test the device I get before letting it near any of the other components in the amp. Wouldn't want to create collateral damage!

[Kleinstein]

I don't think one would need matched transistor in this circuit. One may want to combine them to have the sum of the thresholds not to high or too low. The sum of the thresholds effects where the trimmer for the bias current should be. After changing a FET set the trimmer to minmal current (maximal resistance) and than adjust the the correct corrent.

So one may want a FET with pretty low threshold.

[Jeff eelcr]

Sony tends to manufacture around the parameters of the semiconductor they use, on occasion this creates problems with subsitution.
Japanise transistors are marked with a code for matching parts other makers do not do this but keep same runs close.
I have found that the better the matching the better the DC offset is to 0.00VDC.
Jeff

[Kleinstein]

The output transistor matching is not effecting the DC offset. The DC offset is set by the input stage.
If at all it effects the cross over distortion. This would here be more Gm matching, not threshold matching. It mainly gets important if multiple transistors are used in parallel to get power sharing right.

[klausman]

Ok, an update on this one. I finally got the replacement 2SK1529-Y, and it measures almost identical to existing one on the other channel:



(the green box is the values of the new MOSFET)

Since I never figured out quite _why_ it blew up, I am now irrationally troubled by how much the P-channel MOSFETs differ. Is there a chance the right channel 2SJ200-Y (bottom right) caused this? Or is damaged as well? I've not turned on the amp since the original Pop, for fear of making things worse. I checked almost everything downstream of both main amp circuits, and found no shorts and the star ground seems to be all well connected everywhere it should be. Thanks to tooki, I will be able to power this up using a Variac. When doing that I will of course keep a close eye on temperatures and voltages on the B+/B- rails. Anything else I should check?

[klausman]

So here's an update. The thing works again!



Now I just need to figure out the matter of not being able to set the right channel bias current high enough...