UK Vintage Radio Repair and Restoration Powered By Google Custom Search Vintage Radio Service Data

Go Back   UK Vintage Radio Repair and Restoration Discussion Forum > General Vintage Technology > Components and Circuits

Notices

Components and Circuits For discussions about component types, alternatives and availability, circuit configurations and modifications etc. Discussions here should be of a general nature and not about specific sets.

Closed Thread
 
Thread Tools
Old 9th Dec 2017, 12:52 pm   #21
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

The whole matched transistor thing got me wondering, how well matched were transistors that were sold in pairs ? Luckily I had some to test , the results are interesting. I tested two lots of OC72's, that were matched by Mullard (see attached photo).

I set up a circuit to bias them into class A with 5mA emitter current.

( I'll describe the circuit in case anyone wants to try it, all resistors 2%)

The circuit was powered by 12V battery, the transistors had 1k2 collector resistors and 200R emitter resistors, bypassed by a 10uF electro. The base to ground (positive) resistor was 6k8, and the base to -12V bias resistor 51k. In series with each base was a 1k2 resistor and a 0.47uF coupling cap. This connected to a 1kHz signal generator, the ground of which was connected to +Ve.

This established an emitter current of just under 5mA

With 160mV pp from the generator's terminals the collector voltage swing was 6V pp.

When I tried the first "matched pair", with the same batch code, from Mullard's packet, the emitter voltages and currents were within about 1 to 2% of each other, And the output amplitude of the 1kHz sinewave practically identical within a few % at 6Vpp. Very difficult to tell the transistors apart, which I guess was the aim of the matching.

However, the supposed matched transistors from another white packet, which interestingly had different batch codes, where quite different at least 15% for the emitter voltages(currents) and dynamic gain. Made me wonder if this packet had been tampered with, as it was opened. And probably its less likely that transistors from different batches would match, but I guess they could.

A then tested a couple of random AC188's, these were over 15% different on both gain and emitter current, and the gain was about 20% higher than the OC72's.

So if at least one of these supposed matched transistor packets is anything to go by, Mullard looks like they did actually try to match them.

I do not have any more factory matched ones to test.
Attached Thumbnails
Click image for larger version

Name:	matched.jpg
Views:	43
Size:	59.6 KB
ID:	153652  
Argus25 is offline  
Old 9th Dec 2017, 1:38 pm   #22
mhennessy
Dekatron
 
Join Date: Jul 2007
Location: Evesham, Worcestershire, UK.
Posts: 3,354
Default Re: Matched Transistors

That seems like a complicated way to test for hFE

The input impedance of each transistor is hFE/gm.

If fed from a source with no or very low source impedance, then the variations in overall gain would have been rather less significant. But the (unknown) source impedance of the generator added to the 1k2 series resistor has formed an effective potential divider with the hie of the transistor, causing a loading effect with is roughly proportional to hFE.

In theory, the gm of the transistor would have been about 200. So for a hFE of 100, hie will be 500 ohms. But for a hFE of 50, we're talking about 250 ohms.

Given the source voltage of 160mV from 1k2, that's 47mV at the base of a transistor with a hFE of 100, or 28mV for one with a hFE of 50.

At first glance, the gain of the transistor should be 240 (200 times 1.2). 47mV times 240 is 11.3V, and 28mV times 240 is 6.7V. Obviously, there's a bit more to it that just that - not least Zout of your generator - but essentially this shows that while the transconductance of a transistor is not affected by hFE directly (assuming Ic stays the same), the changing input impedance will change the overall voltage gain when there is significant impedance in the base circuit. But this should come as no surprise, of course.

(For brevity, I've ignored other effects here - the above should be treated as "back of the envelope" rather than a rigorous analysis. But if nothing else, 470n at 1kHz is far from a short circuit )

What's more interesting is to consider the impedance that feeds the output pair in a push-pull amplifier. The driver transformer is fed from the collector of the preceding transistor (which usually is a CE amplifier, so has a moderate output impedance). This is obviously reduced by N-squared - what's the typical turns ratio of a driver transformer? Given that, we could have a stab at working out the effect of different hFE in the output pair - that would be interesting
mhennessy is offline  
Old 9th Dec 2017, 2:44 pm   #23
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

There was a bit more to it than you thought.

I deliberately added the 1k2 series resistor to the base circuit (the generator output has a low Z), the reason being, as had been pointed out by Kalee20, the transistor's hfe affects the input impedance. So the series resistor helps form a divider at the input. This is also why in the test circuit, the emitter was bypassed with the electrolytic so the AC degenerative effects were minimized, as they would make one transistor specimen look more like another.

Also the drive voltage resulted in a fairly wide collector voltage swing using a good proportion of the transistors transfer function in that circuit, not just the small signal hfe on one part of the curve. The idea being to show any obvious differences in the two transistor specimens for both the static bias conditions and the dynamic performance, much better than a simple hfe test.
Argus25 is offline  
Old 9th Dec 2017, 2:52 pm   #24
kalee20
Dekatron
 
Join Date: Feb 2007
Location: Lynton, N. Devon, UK.
Posts: 4,619
Default Re: Matched Transistors

Agree with Mark!

Voltage gain is sensibly independent of hfe, when measured actually on the base, but the loading effect of poor hfe on the generator, will result in a lower signal on the base.

What's wrong with just DC measurements - a 4.7k base resistor from a variable voltage source to drive the base, and a current meter in series with say 100 ohms (for protection) from a 12V supply. Twiddle the base supply for a 10mA collector current, and the hfe calculation is simple.

The transistor may need to be kept cool for this, as leakage current increases with temperature. But a quick trial will determine this.

Argus's method is good in that the transistor's operation is stabilised, but it does need some means of determining the actual (AC) base current to work out hfe.
kalee20 is online now  
Old 9th Dec 2017, 3:13 pm   #25
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

Quote:
Originally Posted by kalee20 View Post
Argus's method is good in that the transistor's operation is stabilised, but it does need some means of determining the actual (AC) base current to work out hfe.
There is no requirement at all to identify the actual value of the hfe and it has no value for the application. The circuit was intended to simply illuminate differences in the behavior of two transistor specimens. Even the base bias resistors were chosen to be a little higher range than usual to help indicate possible different leakage properties. The purpose is not to measure hfe, but to see if two different transistors have near identical static & large signal dynamic electrical properties..or not.The collector current was kept at a modest value to avoid any significant thermal effects, none noticed during testing.

Last edited by Argus25; 9th Dec 2017 at 3:23 pm.
Argus25 is offline  
Old 9th Dec 2017, 4:04 pm   #26
Refugee
Dekatron
 
Refugee's Avatar
 
Join Date: Apr 2012
Location: Worksop, Nottinghamshire, UK.
Posts: 3,563
Default Re: Matched Transistors

I will have to keep an eye out for the factory matched pair that is in my collection.
They are marked GETXXX and have factory fitted heat sinks.
If I spot them in the next week or so I will run the test at 10 or 20 ma as they have a higher power rating than the AC types we have tested up to now.
Refugee is offline  
Old 10th Dec 2017, 1:39 am   #27
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

Quote:
Originally Posted by mhennessy View Post
The input impedance of each transistor is hFE/gm.
I still tend to prefer transistor analysis as though they were current amplifiers rather than transconductance devices, since the input impedances are low and the input currents not insignificant, most of the time, but it is personal preference and each has advantages/disadvantages.

If you want a rough calculation of the hfe:

see next post

Last edited by Argus25; 10th Dec 2017 at 1:53 am. Reason: Had to make corrections
Argus25 is offline  
Old 10th Dec 2017, 2:02 am   #28
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

Since 1/re is the gm then in theory the input impedance, at the base alone, is quite high for the test circuit despite the fact the emitter was grounded to AC, assuming say an hfe of at least 50 at 25/5mA x 50 or 250k, so that is swamped by the base bias resistors to around 6k. This suggests, taking the 1k2 series resistor into account and dividing the base pp voltage by re and taking into account the collector dynamic current, the hfe calculates at about 180.

To double check this I will take some more measurements and measure the actual dynamic base current and post the result
Argus25 is offline  
Old 10th Dec 2017, 3:20 am   #29
kalee20
Dekatron
 
Join Date: Feb 2007
Location: Lynton, N. Devon, UK.
Posts: 4,619
Default Re: Matched Transistors

Quote:
Originally Posted by Argus25 View Post
Quote:
Originally Posted by kalee20 View Post
Argus's method is good in that the transistor's operation is stabilised, but it does need some means of determining the actual (AC) base current to work out hfe.
There is no requirement at all to identify the actual value of the hfe and it has no value for the application. The circuit was intended to simply illuminate differences in the behavior of two transistor specimens...The purpose is not to measure hfe, but to see if two different transistors have near identical static & large signal dynamic electrical properties..or not.
Good point. I'd lost focus of the overall aim!
kalee20 is online now  
Old 10th Dec 2017, 3:43 am   #30
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

I re-tested to circuit and also to see how close calculation & measurement might agree if one was interested in the hfe:

Replacing the bias network with a simple 270k resistor to keep the collector current close to 4.8mA, also increasing the 200R emitter bypass cap to 1000uF(the 10uF one I used had a high esr it turned out it was an NP type) also increasing the input coupling cap to 10uF so that it didn't factor in, the generator voltage is 110mV pp for 6vpp at the collector.

The 110mV pp signal is distributed as: The voltage across the series 1k2 is 60mV and the B-E voltage of the OC72 50mV pp.

The current in the series 1k2 to the base is therefore 50uApp, ignoring the current into the 270k, this is nearly all dynamic base current. (the dynamic or signal current in the 270k is 50mV/270k or 0.18uA and so doesn't count).

The dynamic collector current is 6vpp/1k2. or 5mA. This makes the measured hfe 5mA/ 50uA or 100 exactly.

From the DC perspective the base current via the 270k bias resistor is approx 44uA from the 12V supply, and the DC collector current 4.8mA, so that makes the HFE around 109, which agrees with the large signal test.

Since the voltage gain comparing the base and emitter dynamic voltage is 6Vpp/50mVpp or 120, and the collector resistor is a 1k2 then the measured value of re is 1k2/re = 120, making re = 10 ohms in practice. However, the theoretical value of re is 25/4.8 = 5.2 ohms.

Trying to asses the input impedance by using hfe/gm or hfe(re) by calculating re (or gm) from the collector current is not as accurate as measuring it, it would appear.

I've attached an image of the test circuit. Obviously this is not an ideal way to bias for an amplifier , but actually I have decided, for comparing two transistors it is better as it also helps compare the C-B leakage for germanium types.

Please ignore the error in post 28. And I mis labelled the 1000uF electro polarity in the diagram !
Attached Thumbnails
Click image for larger version

Name:	oc72.jpg
Views:	34
Size:	47.9 KB
ID:	153712  

Last edited by Argus25; 10th Dec 2017 at 3:59 am.
Argus25 is offline  
Old 10th Dec 2017, 3:09 pm   #31
mhennessy
Dekatron
 
Join Date: Jul 2007
Location: Evesham, Worcestershire, UK.
Posts: 3,354
Default Re: Matched Transistors

Quote:
Originally Posted by kalee20 View Post
Quote:
Originally Posted by Argus25 View Post
Quote:
Originally Posted by kalee20 View Post
Argus's method is good in that the transistor's operation is stabilised, but it does need some means of determining the actual (AC) base current to work out hfe.
There is no requirement at all to identify the actual value of the hfe and it has no value for the application. The circuit was intended to simply illuminate differences in the behavior of two transistor specimens...The purpose is not to measure hfe, but to see if two different transistors have near identical static & large signal dynamic electrical properties..or not.
Good point. I'd lost focus of the overall aim!
But it really depends on the circuit.

If you build the circuit above, but incorporate a variable resistance in the base so you can adjust for 6V at the collector for every transistor you test, you'll discover two things:

1. A measurement of Vb and Vc will show that the voltage gain of the transistor itself doesn't change. Even if you switch between silicon and (healthy!) germanium.

2. The overall gain goes change, thanks to the changes in hie as a result of changes in hfe.

Neither of these things are a surprise. Of course, a designer will expect them, so will design the circuit around this. For one, you might want to stabilise the bias so the collector current remains fairly constant with hFE. But in practice, it's relatively unusual to expect to get this much gain from a single transistor stage as the results aren't exactly linear and low distortion, so today a designer would use more active devices and negative feedback. Obviously things were different back in the early days of transistors...

Really, the point of all this was to see how transistor hFE affects the performance of the 2-transformer push-pull circuit the OP originally enquired about. My original reply - which goes against the accepted wisdom, but comes from years of experience - perhaps raised a few eyebrows but I'm not in the habit of making claims without good reason. Measurements of a single CE amplifier aren't really sufficient, as we don't know the source impedance "seen" by each transistor in a PP amplifier as yet. But I've remembered that up in the attic I have a scrap Roberts R300 chassis - the audio stages are OK, so if I can remember where that might be, I'll brave the cold later on today or (more likely) next week.

In the meantime, I'm looking at a Hacker amplifier - not the same one I linked to earlier, but a similar one from the earlier RP18. This gives a wholly respectable distortion of <0.5% when no load is connected to the output, but as soon as you connect the 30 ohm load, it jumps to 10%. On the 'scope, the positive-going parts of the sine wave are about 25% lower than the negative-going parts. The hFE of the upper output transistor is fine, so the fault is perhaps in the preceding VAS transistor, or somewhere else entirely - I'll get back to it in a moment. But suffice to say, it actually sounds quite reasonable on speech and music - the asymmetry causes mostly "nice" second-order distortion rather than unmusical 3rd-harmonic.

That set was not brought to me with a complaint about distortion - I'm only investigating it out of completeness and principle (and interest). Which certainly shows that in practice, it's quite possible that even a gross mismatch of output transistors in a transformer-coupled PP amp might go unnoticed by the listener

All the best,

Mark
mhennessy is offline  
Old 10th Dec 2017, 11:14 pm   #32
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

Quote:
Originally Posted by mhennessy View Post
If you build the circuit above, but incorporate a variable resistance in the base so you can adjust for 6V at the collector for every transistor you test, you'll discover two things....
Yes, that is correct, but the idea of this circuit was not to put a transistor into a circuit and adjust its operating point to see how its properties change with different operating points (we know already what to expect there).

The idea was to have the circuit fixed as it is and duplicated twice. Two transistors are plugged in and the collector waveforms compared. In a test jig a scope is not needed to look at the collector waveforms because these simply get subtracted from each other with an OP amp.

The magnitude of the output signal from the OP amp then becomes a difference signal which can be precision rectified and sent to a meter circuit.

With two transistor specimens, identical in every respect, the meter output is zero. The output then becomes dependent on the differences between two transistor specimens. So it is part of a tool find differences between transistor specimens, not to investigate the behavior of individual specimens.

Of course what magnitude of difference between two specimens would be enough to have a practical effect in a transistor radio's output stage would need to be determined, but the question I was interested asking was how closely matched were Mullards' matched pairs, and the answer at least to one pair I had was the specimens are practically indistinguishable, quite different from a few randomly selected ones from my junk box.
Argus25 is offline  
Old 11th Dec 2017, 1:11 pm   #33
Refugee
Dekatron
 
Refugee's Avatar
 
Join Date: Apr 2012
Location: Worksop, Nottinghamshire, UK.
Posts: 3,563
Default Re: Matched Transistors

That sounds like a job for a centre zero meter between the collectors surely?
Refugee is offline  
Old 11th Dec 2017, 1:16 pm   #34
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

Possibly yes if the meter was sensitive enough. It would also have to respond to a ripple voltage at the test frequency, so it would have to be a center zero AC meter. With the OP amp its easy to subtract signals, add gain, precision rectify the ripple(error voltages) and feed a boring insensitive DC meter.
Argus25 is offline  
Old 11th Dec 2017, 1:23 pm   #35
Refugee
Dekatron
 
Refugee's Avatar
 
Join Date: Apr 2012
Location: Worksop, Nottinghamshire, UK.
Posts: 3,563
Default Re: Matched Transistors

Just use a variable bias and no AC signal.
You can then plot the HFE at several values and get an idea of linearity.
Refugee is offline  
Old 11th Dec 2017, 1:37 pm   #36
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

Quote:
Originally Posted by Refugee View Post
Just use a variable bias and no AC signal.
You can then plot the HFE at several values and get an idea of linearity.
But then you have to record values and create a plot at different values, its easier just to look at a single meter reading which averages the errors over a large proportion of the curve.
Argus25 is offline  
Old 11th Dec 2017, 1:50 pm   #37
Refugee
Dekatron
 
Refugee's Avatar
 
Join Date: Apr 2012
Location: Worksop, Nottinghamshire, UK.
Posts: 3,563
Default Re: Matched Transistors

That will average the error over the range.
Perhaps it is time to build a curve tracer?
Refugee is offline  
Old 11th Dec 2017, 2:34 pm   #38
Argus25
Octode
 
Argus25's Avatar
 
Join Date: Oct 2016
Location: Maroochydore, Queensland, Australia.
Posts: 1,717
Default Re: Matched Transistors

Quote:
Originally Posted by Refugee View Post
That will average the error over the range.
Perhaps it is time to build a curve tracer?
I wonder if that is how Mullard actually matched their transistors, since they seemed to make a big deal of it at the time and selling specially packaged matched pairs, one would have assumed they would have had a curve tracer which is a relatively simple device.

Has anyone seen any documentation on exactly how Mullard did it and what parameters they actually matched ? It might just have been a simple hfe test at some bias point & leakage match, or it might have included a full sweep up to its max value for a base and collector current curve match.
Argus25 is offline  
Old 11th Dec 2017, 3:05 pm   #39
Refugee
Dekatron
 
Refugee's Avatar
 
Join Date: Apr 2012
Location: Worksop, Nottinghamshire, UK.
Posts: 3,563
Default Re: Matched Transistors

When I was a teenager I built a transistor tester for all those bags of mixed parts.
It is still in my workshop.
Collector currant was originally set with fixed resistors with 5 values between 2ma and 100ma with fixed VCE.
It was originally accompanied by a set of log tables and later a calculator to divide collector current by the measured base current to get the HFE.
A bench power supply has since been added to get many more collector current settings going as high as 2 amps for big power transistors.
It was effectively a 1970s teenagers curve tracer but it is still useful.
Refugee is offline  
Old 11th Dec 2017, 3:29 pm   #40
kalee20
Dekatron
 
Join Date: Feb 2007
Location: Lynton, N. Devon, UK.
Posts: 4,619
Default Re: Matched Transistors

Quote:
Originally Posted by Argus25 View Post
... so it would have to be a center zero AC meter.
I've never seen a centre-zero AC meter!

It could be made possible if synchronous rectification was used, with a gating signal direct from the master oscillator. Then, the movement drive could swing negative as well as positive. But it's getting rather complicated!

Argus's reports on matching accuracy of Mullard transistors are quite impressive. It's unlikely that Mullard actually matched for more than one characteristic, though still possible. Almost certainly then, the chosen characteristic would be hfe.
kalee20 is online now  
Closed Thread

Thread Tools



All times are GMT +1. The time now is 7:06 pm.


All information and advice on this forum is subject to the WARNING AND DISCLAIMER located at https://www.vintage-radio.net/rules.html.
Failure to heed this warning may result in death or serious injury to yourself and/or others.


Powered by vBulletin®
Copyright ©2000 - 2018, vBulletin Solutions, Inc.
Copyright ©2002 - 2018, Paul Stenning.