400 Thermal meaurements of output stages with a WO driver board

[Bikeron]

I recently rebuilt a 400 for my daughter. As part of the rebuild, I matched the output devices. Surprisingly, the old RCA (66546) driver transistors were well in spec (min beta of 30 at 1A 5Vce), so I kept those. The Old Delco devices were also measured for beta and leakage, and the beta tests found them lacking as there was quite a beta variation (low of 25, high of 59). I don't know that the factory matched these Delco units, but in the "old days," beta did vary a lot, so I would be amazed if they weren't matched initially.
I bought 14 MJ21196G units from ON Semi and sorted them into four groups of three based on beta.

The resulting temperatures that occurred at no load after bias adjustment:
White Oak driver:


Left Channel Output Stage (as viewed from the back of the 400):

Right Channel (as viewed from the back of the 400):

 

[J!m]

Counterpoint gear runs with no base resistors so they would heat the devices and measure at temperature.

Needless to say you can’t really repair them if the output fails- you need hundreds of devices to make up matched quads (SA 120) or octets (SA220).

 

[wattsabundant]

I don't believe matching is necessary. The emitter resistors force load sharing. The test procedure has you check the voltage drop across the emitter resistors to confirm they share.

 

[Gepetto]

I don't believe matching is necessary. The emitter resistors force load sharing. The test procedure has you check the voltage drop across the emitter resistors to confirm they share.

Don is correct.
Matching is unnecessary and does nothing. The local negative feedback provided by the 0.33 ohm emitter resistors forces sharing across all the devices and prevents thermal runaway that results if you do not have these resistors.

If you purchase devices from White Oak Audio, your order will be all of the same date code so that matching is a natural byproduct.

All my lab amps run random date codes on purpose as proof that it does not matter.

It does no harm to try and match devices but also provides no benefit. While any matching procedure is in process, the device thermal environment is changing during the course of the measurement forcing change in the parameters you are attempting to measure.

 

[Bikeron]

Don is correct.
Matching is unnecessary and does nothing. The local negative feedback provided by the 0.33 ohm emitter resistors forces sharing across all the devices and prevents thermal runaway that results if you do not have these resistors.

If you purchase devices from White Oak Audio, your order will be all of the same date code so that matching is a natural byproduct.

All my lab amps run random date codes on purpose as proof that it does not matter.

It does no harm to try and match devices but also provides no benefit. While any matching procedure is in process, the device thermal environment is changing during the course of the measurement forcing change in the parameters you are attempting to measure.

Yes, the beta will change as the die temperature changes. I did think of putting them on a cold plate at 25C but instead, I timed the measurement at 10 seconds to compensate.
I agree that the emitter resistors will balance the currents and make beta less critical; the beta is so high in the ON transistors that the resistors dominate entirely; if the beta were too low, perhaps not. The original Delco devices were suspect as their range of betas was large.

This experiment showed how the devices spread heat, all other parameters being equal. I will add a view of a single channel at 157W, showing the heat spread at higher loss levels below.

The idea that all transistors in the same shipment all have the same date codes and therefore they all match? There is no guarantee, as the assembly may not use die that comes from the same wafer. The date code shows only the assembly date. I did not see that White Oak guarantees all date codes match.

I was also considering that a better way to adjust bias is thermally. Is the bias in the optimal spot if the devices' temperatures are all in the same region and the THD is as low as possible for a specific bias setting? What is the trade-off between device temperature and THD?
I didn't get a chance to run that experiment...

Here are temperatures at 157W out into 8 ohms after 45 min at 24C ambient, no fan (impressive, actually):

 

[Bikeron]

Counterpoint gear runs with no base resistors so they would heat the devices and measure at temperature.

Needless to say you can’t really repair them if the output fails- you need hundreds of devices to make up matched quads (SA 120) or octets (SA220).

Great argument for emitter resistors...

 

[Bikeron]

I don't believe matching is necessary. The emitter resistors force load sharing. The test procedure has you check the voltage drop across the emitter resistors to confirm they share.

I agree that it is not required for routine repair. At the point you have even to measure the emitter current, the beta has probably degraded so far that there would be other problems.

 

[J!m]

Great argument for emitter resistors...

Mike eliminated them as it sounded better to him that way.

 

[Gepetto]

Mike eliminated them as it sounded better to him that way.

Probably MOSFET tho...

 

[J!m]

I think so yes.

 

[Gepetto]

I think so yes.

Totally different Jim. MOSFET have a positive tempco D-S unlike bipolars that have a negative tempco B-E. You can directly parallel MOSFETs and they will self balance as long as you have gate damping resistors to eliminate parasitic oscillations between the devices sharing the load.

 

[wattsabundant]

Worst case condition would be at about 25 volts (78 watts ) at 8 ohms. The output stage dissipates the most heat at 1/3 power. When I work on an amp I check device temps with one of the cheapie infrared guns. I've also used a FLIR camera with good results. The absolute value is not as critical as long as they are consistent. I don't let anything go above 90C.

 

[J!m]

Totally different Jim. MOSFET have a positive tempco D-S unlike bipolars that have a negative tempco B-E. You can directly parallel MOSFETs and they will self balance as long as you have gate damping resistors to eliminate parasitic oscillations between the devices sharing the load.

I guess self balancing in simulation is easy whereas in reality no so easy… all I know is that it is unconventional and a big pita if you need to repair it. Maybe new boards are the next WO project…

 

[mlucitt]

I did not see that White Oak guarantees all date codes match.

I buy NPN and PNP Output Transistors from WOA in ONSEMI factory trays of 100 at a time (all 100 had the same date code). I used to test each device for Gain and Leakage with my Sencore TF-151A Transistor Tester. When I got consistent variations of less than 1%, I stopped testing.