Gain Matching Output device ???'s

[Zero Cool]

How do i determine what specs are important for gain matching bipolar output transistors???

Lets say i have some MJ15023/24 output devices in the output section of a power amp. and i have a batch of new Onsemi transistors to select from and i have a curve tracer capable of supplying 10 amps max. at what level do i want to be looking at?

what is important to ensure a proper match between output devices?

From everything i can find. the simplest answer seems to be to testthe device at the current level you will be using the device at.

Well how do you do that in an audio power amplifier? Should i be looking at the curve at BIAS level? or should i be looking at the curve of what the max power the amp can supply by diving that by the output power by the number of output devices (400w @ 2 ohms / 4 devices = 100watts each with a 70V rail voltage) sort of a thing?

When looking at curves in general, is there a relation to the curve at low levels and at high? So in otherwords. if i look at the curve at 1 amp, would that then relate to the curve at 10 amps? so that if i matched devices at 1 amp, they would all then be the same at 10 amps?


I need to understand better the sorts of things i should be looking for to understand how this works and i have not found the answer by searching so far.

I have read N.P. articles on matching Mosfets and that seems easy enough to setup a test system and do. Bipolars seem more complicated and i am trying to find a easy to understand system to setup for testing output devices.


Zc

 

[Zero Cool]

I re searched the forum using a different key word and method and found this post from another thread. this seems to answer my questions,

Match all the NPN's you need, do the same for the PNP's. If you can get the NPN group's gain close to the PNP group, then great. That's what you do. This will reduce distortion. Use the same emitter resistor values for all regardless.

Match the gains around your idle current, that is where it's most important. The gains will tend to remain in the ballpark. At higher currents, the emitter resistors will enforce sharing. Your per device bias current for bipolars on a consumer amp normally runs 10~25 mA, sometimes higher but not typically over 50 mA.

One blows in an output stage, replace the lot. You can play with the "good" ones. I'd replace the driver transistors too and possibly the bias control transistor (VBE multiplier).

Make sure the heatsink surface is flat and clean, same for the transistors. Use new mica and fresh grease aplied evenly. Do not overtighten the transistor case. Work clean, clean, clean.

-Chris

This is a good thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=84737&highlight=

 

[AndrewT]

Hi,
I think this is an important parameter for matching

Match the gains around your idle current, that is where it's most important. The gains will tend to remain in the ballpark. At higher currents, the emitter resistors will enforce sharing. Your per device bias current for bipolars on a consumer amp normally runs 10~25 mA, sometimes higher but not typically over 50 mA.

do you have emitter resistors?
I am unsure whether Vbe at the bias current is as important or more important.

I tend towards Vbe being more important than gain AT THE BIAS CURRENT.

my reasoning being that at the same bias current the emitter resistors each have the same current and so the same voltage drop. The VOLTAGE coming from the driver is fed into the base. IF the emitter currents match then the Vbe + Vre must also match and this leaves Vbe as the parameter needing matching.

As output current starts to flow then Vre begins to dominate and reduces the effect of slightly unmatched Vbe at these higher currents.

Now consider what happens to a device with lower gain at the higher current. It will pass less collector current and so have less Vre than it's partners. This forces a higher Vbe after subtracting Vre and so the higher Vbe pushes more current than it was expecting. So it appears that the emitter resistor (at least partially) compensates for variations in Vbe and gain as the output current rises. Nice coincidence.

 

[Zero Cool]

Yes the design i am working on (Classe DR-15) has emitter resistors. .27 ohm 5 watt.

According to the service information, they are setting BIAS at 74ma per device. measuring 20-22mv across the .27 ohm resistors +/- 6mv
(one section mentions 20mv another handwritten note mentions 22mv).


Zc

 

[Zero Cool]

Does the voltage matter when matching device's or just the current through the device?

Reading RoD Elliots website there are some interesting articles on building your own test rig ALA Pass. His set ups use alow voltage transformer. But i was wondering if i could simply set up a test rig to run right at the same rail voltage with the same Emitter resistor and Bias current?

In this article :
http://sound.westhost.com/transistor-matching.htm
There is a simple test setup drawn. Basically a resistor from the + rail to the Collector of the device under test with the emitter grounded. a pot in series with a resistor to the + rail and ground with the wiper feeding the base of the DUT.

With this rig you set it up once and then compare that first device with the other devices in the lot to see which come closest to matching. Sounds logical.

So, why couldnt i design this same rig but have it run right at the same rail voltage as the amplifer so that it is running as close to the real deal as possible? Again, sounds logical.

Thoughts?


Zc

 

[AndrewT]

Temperature!!

With Vrail on the device the power will ramp up the junction temperature rapidly.

You will find it very difficult to measure at the same temperature for each device and combined with varying Vbe and gain, your numbers will be all over the place.

 

[Zero Cool]

ahhhh I see. good point. So meauring with a lower voltage at the same current is better? wouldnt the device heat up just as fast?

What if i were to do a long term measurement. mount the transitor to a heatsink. let it warm up and become stable etc?

Or would i be better to pulse test them. I could use a MCU to give a short pulse of whatever duration.


Zc