Transistor matching

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

mikeyB

Well-known member
Joined
Apr 12, 2005
Messages
568
Location
Manchester UK
I'm planning on building a transistor matching test jig based around the GM328 tester, many of them on ebay.
Quick question is: is there an equation or correction factor regarding transistor hfe at different temperatures?
The temperature is the only variable as everything else is set by the tester. Obviously, I need a thermometer and need to note this along with the test measurements.

Thanks in advance
 
mikeyB said:
Obviously, I need a thermometer and need to note this along with the test measurements.
Or use a little Peltier plate sandwiched between a heat sink and a block of aluminum with holes for various transistor packages. Then you just rest that on top of the transistor to set it's temperature. Use a thermistor mounted on the aluminum block to servo the DC converter powering the plate.
 
mikeyB said:
I'm planning on building a transistor matching test jig based around the GM328 tester, many of them on ebay.
Quick question is: is there an equation or correction factor regarding transistor hfe at different temperatures?
The temperature is the only variable as everything else is set by the tester. Obviously, I need a thermometer and need to note this along with the test measurements.

Thanks in advance
What is the application that needs transistor's hfe to be matched so closely? Assuming you do the measurement at ambient, that may vary over about 20°C (35°F), the hfe variation would be about +/-5%. I understand that controlling Vbe with accuracy is important in many applications, but hfe?
 
abbey road d enfer said:
What is the application that needs transistor's hfe to be matched so closely? Assuming you do the measurement at ambient, that may vary over about 20°C (35°F), the hfe variation would be about +/-5%. I understand that controlling Vbe with accuracy is important in many applications, but hfe?
It's for matching for discrete opamps, in particular the differential input pairs
 
mikeyB said:
It's for matching for discrete opamps, in particular the differential input pairs
For LTP (long tail pair) use Vbe matching is far more important than beta matching. Devices from the same production batch should be similar enough beta... note: Vbe changes with temperature so ideally Vbe is measured at similar temperature.

JR
 
I may have read / misread  PRR saying something to the effect that once the thing is fired up and pushing audio in its native environment temperature is going to vary, so don't lose your mind seeking a perfect match.  Can't hurt to get them close?
 
boji said:
I may have read / misread  PRR saying something to the effect that once the thing is fired up and pushing audio in its native environment temperature is going to vary, so don't lose your mind seeking a perfect match.
Indeed, transistors operating inside an enclosed piece of gear will drift significantly from the measurements done at ambient, but they will drift identically, as long as they are close enough to be at the same temperature.

Can't hurt to get them close?
Indeed. I used to superglue transistors in the log amp in my 1979  31-band RTA.
 
You can buy transistor arrays where the transistors are on a single substrate (not unlike ICs) for superior thermal tracking.  Over the decades I used lots of inexpensive transistor arrays (lm3086) with 5 small NPNs.

JR
 
mikeyB said:
Thanks John,
But they are for use in various types of DOA  with the 2520 footprint
That corp sells some decent arrays http://www.thatcorp.com/300-series_Matched_Transistor_Array_ICs.shtml, but it should be easy to source modern off the shelf op amps with superior performance to DOAs.

The last time I designed a DOA was in the 70's (80's?) and I used a LM394 in that one (for a bus summing amp). 

JR
 
The best way I've found to do this is to build the front end of the DOA on a bread board, and then you can swap individual BJT's in to the circuit. 

For example, you can build up this circuit:

transistor-long-tail-pair-circuit-03.gif


Set up the current source to whatever you will be running in your actual implementation, then ground the two bases of T1 and T2.  You can then make a measurement of the DC output between T2 and T4, and swap in different T2 transistors and find a pair that provide the same DC output, then you can place both in the T1 and T2 positions.

The other thing you can do is drive T2 with a precision op-amp, with the non-inverting input grounded and the inverting input tied to the output of the circuit.  You then measure the output of the op-amp (T2's base), and find two transistors that have the same input voltage applied that makes the circuit output ground.  These two transistors should have the same Vbe.
 
I hate to sound like I am supporting this (DOA) endeavor but you can pretty easily make a more precise current mirror using 3 active devices and degeneration resistors to force better current matching.

That crude current mirror will distort your matching attempt.

JR 
 

Latest posts

Back
Top