UK Vintage Radio Repair and Restoration Discussion Forum

UK Vintage Radio Repair and Restoration Discussion Forum (https://www.vintage-radio.net/forum/index.php)
-   Components and Circuits (https://www.vintage-radio.net/forum/forumdisplay.php?f=40)
-   -   Matched Transistors (https://www.vintage-radio.net/forum/showthread.php?t=141966)

Neil Purling 6th Dec 2017 9:48 am

Matched Transistors
 
How close a match must two transistors in a p-p radio o/p stage be?
I can measure the hfe figure.
Suppose I had measured hfe of 200 & 202 for an o/p pair.
Is the measurement linear, so that would be much closer than if the scale were logarithmic.

mhennessy 6th Dec 2017 10:33 am

Re: Matched Transistors
 
hFE varies wildly with temperature, collector current, planetary alignment, etc. Any circuit designer that relies on a particular value of hFE is committing commercial suicide - such a circuit wouldn't have made it past a design review.

Certainly, worrying about a 1% difference is pointless. I'd be happy with 50%, frankly. The key thing is making sure that neither transistor has a hFE that is lower than the minimum spec for that device, because that implies a transistor that is in some way faulty, albeit not quite dead yet.

The acid test is to see how the circuit performs. Check the output waveform with a 'scope - make sure it's a symmetrical sine wave with no premature clipping or crossover distortion. Job done :thumbsup:

MrBungle 6th Dec 2017 12:24 pm

Re: Matched Transistors
 
Indeed.

I think the only time I've seen this being a problem is in discrete differential amplifiers, typically where I am being too cheap to buy a CA3028 for a circuit. Keeping them at the same temperature helps there, enough to be useable. I pick two (2n3904's) with an hFe within 20 or so and then thermally bonded them together with heatsink compound and stick some heatshrink over the pair. Worked very well.

paulsherwin 6th Dec 2017 12:32 pm

Re: Matched Transistors
 
It certainly isn't a critical issue in a typical domestic radio. I always hand select transistors with reasonably close hfe readings but it isn't really necessary. You might get a bit more crossover distortion if they are seriously mismatched but they will still work.

Nuvistor 6th Dec 2017 12:57 pm

Re: Matched Transistors
 
1 Attachment(s)
Attached advert from WW, a Mullard article on how they match AC127 and OC81 transistors.
They match the gain on the transistors at a collector current of 50ma.

Herald1360 6th Dec 2017 1:23 pm

Re: Matched Transistors
 
I believe it depends on the output p-p circuit, whether the match is important or not. Complementary circuits where the output devices are used as emitter followers don't worry too much about matching but the transformer driven transformer output configuration is more fussy about reasonable matching.

Not sure where the transformer driven totem pole types fall.

kalee20 6th Dec 2017 2:21 pm

Re: Matched Transistors
 
The input impedance of the stage, whether emitter follower or common-emitter, is proportional to hfe. So, if the hfe's of the two transistors in a push-pull amplifier are widely different, there will be different loads on the driver stage on the two half-cycles. This could give rise to asymmetric distortion.

So, while agreeing with mhenessey that a circuit which relies on a particular value of hfe is a bad design, it is still good practice to match them if you want best performance.

Being mounted on the same board or chassis, they'll be at the same temperature, see the same planetary alignment, have the same collector currents on symmetric waveforms, everything will track!

mhennessy 6th Dec 2017 3:19 pm

Re: Matched Transistors
 
I think the final paragraph of my first post covered it really. If the amplifier gives good results, then that's all that matters.

I've had failure modes where the positive half-cycle was clearly a different amplitude than the negative half-cycle - in each case this was caused by a faulty output transistor that had a hFE of around 10 or 20 when measured rather than the expected 100+. Clearly very unwell, and the low hFE would have been but one symptom of its illness.

Yes, in this case, it would have been loading down the preceding VAS transistor because it would have presented a low input impedance. But lower by an order of magnitude, not a few percent, or the 20% quoted in the marketing material in post #5, or even the 50% that I suggested.

By way of example, take the amplifier used by Hacker (schematic available on this page. T4 is the VAS, and has a standing current of around 9mA - this is the most base current that can be supplied to the PNP output transistor on the negative peaks, thanks to bootstrapping (more can be supplied to T5 by T4 on the positive peaks). The 15 ohm loudspeaker requires around 440mA peak (1.5W), so T6 (an AC128) must have a minimum hFE of around 50. However, providing this is met, it really doesn't matter what the hFE is of both output transistors. This circuit uses negative feedback to ensure the output voltage is a multiple of the input voltage, and the only times I've had distortion problems with this circuit is - as I've described above - an output transistor has a hFE of around 10 or 20 rather than the usual 100+ (I've seen over 150 with good AC128s).

OK, compared to what the OP asked about, this is a more sophisticated circuit with lots of negative feedback. If you take a simple circuit with a pair of transformers with two transistors in the middle, then yes, you'd perhaps expect the hFE matching to be more critical. But I can only say that this is not my experience. Providing they are both healthy (100+), it really doesn't matter. Any distortion caused by hFE mismatch is likely to be swamped by other non-linearities, such as from the transformers and the limitations of the basic topology of such a circuit. Some of these might produce 10% THD even when in the best of health!

I've repaired hundreds - perhaps thousands - of transistor radios in the last decade, and I take the time to measure each one carefully (including distortion). It's that experience - plus my experience designing "proper" amplifiers - which gets me here. But as always with "rules of thumb", there will always be "outliers", so I wouldn't be surprised if someone comes back with an example of where 50% was not good enough (if so, full data required, please - I'd like to measure it for myself ;))

Differential amplifiers were mentioned earlier. With those, if anything ought to be matched, it's Vbe rather than hFE (although it is possible to design poor circuits which rely on it, naturally). And of course, Vbe varies with temperature (-2.1mV/C), so that's why it's good to thermally bond them in critical circuits. A pair sharing the same silicon will have inherently good matching and thermal tracking. Such precautions are never necessary in a well designed audio amplifier though...

Cheers,

Mark

Argus25 7th Dec 2017 1:59 am

Re: Matched Transistors
 
Quote:

Originally Posted by kalee20 (Post 997378)
The input impedance of the stage, whether emitter follower or common-emitter, is proportional to hfe. So, if the hfe's of the two transistors in a push-pull amplifier are widely different, there will be different loads on the driver stage on the two half-cycles.

I agree with this entirely and I think for transformer coupled push pull output stages its best to have matched transistors.

One issue is that the transistor's small signal current gain value alters depending on the static or DC bias conditions. So there is static and dynamic matching to consider.

Transistors may look matched for hfe on a tester but then they are sometimes not at the same emitter current in two identical bias circuits. So the best move is to set the transistors up in a similar bias condition as in use
and drive them both with the same signal from a 1k source impedance. Connect the two outputs into the + and minus inputs of a voltage differencing op amp to subtract the signals, when matched the output will be close to zero. Also under the same bias circuitry conditions the emitter currents can be compared.

Many years ago I built a jig like this for matching OC72's and OC74's for push pull amplifiers. I think it is worth using matched pairs, as Mullard once supplied them, but perhaps the subtleties of it might be wasted on domestic radio repairs , but for my own radios I think its worth it.

I wonder if there is anyone who knows what the matching protocol once was in Mullard's factory ?

Argus25 7th Dec 2017 5:54 am

Re: Matched Transistors
 
Quote:

Originally Posted by mhennessy (Post 997399)
Any distortion caused by hFE mismatch is likely to be swamped by other non-linearities, such as from the transformers and the limitations of the basic topology of such a circuit. Some of these might produce 10% THD even when in the best of health!

Some output and driver transformers in transistor radios unfortunately were not bifilar wound, as they should have been and there is a not insignificant DC resistance difference from the center tap to either winding end which is not ideal. This puts asymmetry into the output, no matter how well the transistors are matched. Also the distortion and frequency response at the low frequency end is dominated by the core size and often the transformers, for weight & space considerations, were smaller than they could have been.
Interesting though the transformers were often a substantial size in pre 1960's early transistor radios such as the Sony TR72 or the NZ made Pacemaker brand radio that used 3 watt sized cores.

I also agree that the kind of transformer-less audio stages Hacker were fond of do not require any special selection or matching of transistors most of the time, it is really only any advantage to match transistors for the prior generation audio circuits which are transformer coupled class AB push pull amplifiers.

For the most part if the amplifier's performance is checked into a dummy load with the scope and a 1kHz test signal, its pretty easy to see if one of the output transistors is sick, or if there are other issues like crossover distortion.

Neil Purling 7th Dec 2017 7:49 am

Re: Matched Transistors
 
I was thinking about badly matched transistors in one of those Chinese pocket radio kits.
The audio stage transformers are probably rubbish anyway. Specially matching o/p transistors will probably be un-noticeable.
Thanks anyway for the discussion.

Argus25 7th Dec 2017 9:06 am

Re: Matched Transistors
 
Quote:

Originally Posted by Neil Purling (Post 997611)
The audio stage transformers are probably rubbish anyway.
Thanks anyway for the discussion.

I think you are right there, especially the very small pcb mount transformers, might only be good for 35mW output if that. This is one thing I admire about very vintage transistor radios, they often had very impressive transformers in the audio.

Have a look at the transformers in the Pacemaker radio in figure 6 in this article, you can see the size compared to the 6V lantern battery. The driver transformer is the same size as the output transformer:

http://worldphaco.com/uploads/THE_PA..._FROM_1957.pdf

On page 7 of this Sony TR-72 article you can see the substantial transformers they used:

http://worldphaco.com/uploads/THE_SO...STOR_RADIO.pdf

Naturally these radios sound wonderful, both had rear ventilated cabinets, like many Hacker radios.

kalee20 7th Dec 2017 9:56 am

Re: Matched Transistors
 
Quote:

Originally Posted by mhennessy (Post 997399)
I think the final paragraph of my first post covered it really... I've repaired hundreds - perhaps thousands - of transistor radios in the last decade, and I take the time to measure each one carefully (including distortion). It's that experience - plus my experience designing "proper" amplifiers - which gets me here... Such precautions are never necessary in a well designed audio amplifier though...

I think Mark sums this up well!

I stand by my earlier post, that matching is good, to present equal loads on the driver for each half-cycle. But a good driver will be tolerant of different loads on it, hopefully giving a only few percent difference in outputs for widely different loads. That would be a well-designed driver stage.

Some of the cheaper transistor radios would have cut things to the bone at the design stage, though, on the basis that it might be cheaper to spend a few moments matching transistors than designing the thing better (doubtless using more components). Made in volume, with thousands of output transistors to choose from, sorting them into hfe 'groups' would be simple.

Wrapping negative-feedback around the whole thing as a cure-all-ills solution largely works, of course, though some decry it as it increases higher-order harmonics.

Whether necessary or not, it won't do any harm to match the transistors. But if it's a cheap Far East radio, it's not going to be worth buying more than 3 transistors and using the 2 which match most closely.

Herald1360 7th Dec 2017 1:48 pm

Re: Matched Transistors
 
Unless you use silicon ones at a penny or so each......

Boater Sam 7th Dec 2017 2:48 pm

Re: Matched Transistors
 
A penny EACH! Too dear, they get 100 for that.

mhennessy 7th Dec 2017 2:55 pm

Re: Matched Transistors
 
We all know the theory, but what this boils down to is context.

10 years ago, before I started messing about seriously with old radios, I would have instinctively agreed about matching the hFE to 20% or better. And with a "quality" hi-fi amplifier using this sort of topology - such as the Rogers Ravensbourne, to pick a random example - then yes, attention to detail can make a difference. But that is because such an amplifier was designed to give as good a performance as possible, given the constraints of the topology.

But in the context of a transistor radio, even a primitive one using a pair of same-polarity transistors sat between a pair of transformers - and perhaps even with no overall NFB to help - hFE matching is a non-issue in my experience, providing both transistors are healthy.

And quite apart from the amplifier, how about the speakers found in these sorts of radios? Again, quite different to hi-fi speakers. Some radio loudspeakers have terribly high levels of harmonic distortion, along with a very uneven frequency response - again, I've measured lots of these in recent years (and put some of the results on this forum). The speaker dominates the overall performance, and as such, there's simply no point making a transistor radio amplifier with 0.01% distortion. In practice, you're lucky to get below a few percent with ancient transformer-coupled designs IME.

And remember, most of these sorts of sets are AM only - ever measured the THD of a typical AM receiver? If anything, some "nice" even-order harmonics might usefully brighten up the sound of an AM transmission ;)

Of course, transistors were relatively new back then, and manufacturers lacked experience and knowledge, so it is possible that some believed that hFE matching was important, even in a radio. But of course, people used to believe that transistors are current operated devices ;)

Anyway, sorry to have waffled on. I hope some of it is of interest :thumbsup:

G6Tanuki 7th Dec 2017 7:42 pm

Re: Matched Transistors
 
Back in the days of OC81s, AD161/162 and the like you could buy "matched sets" but I never bothered. As has been mentioned, they weren't amplifying anything (or driving anything) where serious quality would be needed so my test was just to listen to the radio/amplifier with it turned up loud, and if it sounded good after 10 minutes or so I'd put a finger on the transistors/heatsinks expecting them to both be similarly warm.

TonyDuell 7th Dec 2017 7:52 pm

Re: Matched Transistors
 
The EE1004 extension kit for the Philips Eletronic Engineer series contained a matched pair of AC132s (with heatsinks on the little PCBs used for transistors in those kits). If you damaged one you had to buy another pair from Philips.

Now that kit did contain projects for transformer coupled (driver and output transformers) push-pull amplifers. I do wonder how critical the matching was in a kit like that. Could you have 'got away' with just soldering any old AC132 onto the PCB?

kalee20 7th Dec 2017 7:54 pm

Re: Matched Transistors
 
Ah, AD161/162, that's NPN/PNP.

Matching these is about as similar as getting a man and a woman of the same height and weight.

Argus25 7th Dec 2017 10:17 pm

Re: Matched Transistors
 
Quote:

Originally Posted by kalee20 (Post 997811)
Ah, AD161/162, that's NPN/PNP.

Matching these is about as similar as getting a man and a woman of the same height and weight.

That is a very good way to put it :-D

I've never had much joy from these transistors, they are quite easy to damage in experimental amp setups. But the NPN/PNP combo did help eliminate output transformers.

In the end one of my favorite audio circuit topologies uses a split driver transformer, which is often not to bulky at all, a pair of PNP's like AD149's. That way you have the best of both worlds and eliminate the output transformer mass and can even match the transistors, if you believe it makes a difference.


All times are GMT +1. The time now is 10:55 pm.

Powered by vBulletin®
Copyright ©2000 - 2024, vBulletin Solutions, Inc.
Copyright ©2002 - 2023, Paul Stenning.