Matching transitors, HfE and test sets......?

Ok sorry for all the stupid questions. I read over a bunch of posts and links to matching transistors. It seems you have to build a test jig to do that. I noticed that on my Triplett meter there is a place to test hFE for transistors. Is this the same thing? Can I use this to match transistors?

Populating a PCB is one thing, but understanding the design and all the little nuances is totaly a different story! UgH!!!!!!


Thanks,

Joe

just for the heck of it, I tested some transistor using the Tripplet function. These transistors are from a video amp. I was getting numbers like 155, 198 and 268. If this is the same thing as what other were refering to, then when I get my BC550/560 I should be looking for numbers well over 400. Right?

So, I guess the question I have is, is this the same thing as "matching" transistor? Do I just look for pairs that have close to the same numbers?

Thanks,
Joe

IMO to match transistors you need to build a circuit that measures the transistor at the range it will be used at.

Say you want to match some that idle at 1ma Ic and swing .5ma each way. I would build something to test Hfe at .5ma and 1.5ma. Now if the transistors are from the same batch a one point match would be OK.

Problem with a DMM transistor tester is it uses one Ic maybe not the place the transistor will be used in circuit IIRC.


Look at transistor graphs things like Hfe vs Ic sometime the sound is in the curve or operating point of a Si IMO. Some transistors have flatter Hfe vs Ic some max the Hfe at different Ic etc.....................................................

How you match transistors depends on what's important in the given circuit.

Hfe is just one parameter. Another one that may be important is base-emitter voltage. It all depends on the circuit. Gus's point, to do the match under the conditions of use, is a good one. That includes the operating temperature as well---both Hfe and Vbe have significant temperature coefficients. If I'm trying to match some devices where Vbe is critical, for example use in a VCA, I will tape them down to a surface and let them stabilize (just handling them will induce significant temp changes) and then go in and measure with some probes. The diode forward voltage function on my old fluke DMM works well for this.

Beware as well of mixing manufacturers. Even though the datasheets may read identically, the constructions are almost always different and with them the typical performance.

A "well-designed" circuit should be relatively insensitive to Hfe. But sometimes we sacrifice uniformity for other potential performance benefits---vide the popularity of low-global-feedback designs.

Look as well, as Gus suggests, at the extremes of signal swing in the circuit considered. You may even want to check some of the matching parameters at the extremes of voltage and current, not just at the nominal operating point. And be wary of the self-heating of the device while you are performing the measurement---a consistent pattern over time will help.

Thanks for your reply. I read several of your other posts on this subject as well. Some or even most of this is over my head. I am building some 2520 opamps first, then some 1731 next. And finally a FET BOY preamp. So I was kind of generalizing the question to get some advice.

I think I may match the first ones as best as I can for now, and learn as I go. I can't imagine when Melcor or API were building opamps back in the 70's that they sat and match each pair of transistors. That may actually be some of the magic about older gear. Things that were not supposed to sound good, and have terrible specs, sometimes turn out to really rock. The LA3 compressors come to mind>

Joe

Let me know what is unclear. I'll never be as good at explanations as our master but I can try...

One note: when used in op amps the extent of transistor matching may depend as well on how the op amp itself is going to be used. If it's in an active filter with a d.c. gain of 2.5 (say), and is far down the signal chain, and a.c.-coupled to the next stage, d.c. offset and temp drift of this offset will be pretty unimportant. If you are using the amp as a high gain stage and are trying to eliminate as many caps as possible out of concerns about sonics, you may want something with low offset and drift. This may not only require matching of certain parts but may rule out some topologies to begin with.


EDIT: interesting that when written this was post 2520 (cue twilight zone incidental music bits...)

> I can't imagine when Melcor or API were building opamps back in the 70's that they sat and match each pair of transistors.

It's possible. I do think people today obsess over "matching" much more than is warranted.

In circuits like you describe, Hfe match only makes sense if the external DC resistance is matched. In many audio systems, it isn't. In that case, selecting for high Hfe does reduce errors, but buying from the high-Hfe bin of a high-Hfe type usually takes you as far as the basic design will go.

Vbe mis-match is multiplied by DC gain and appears at your output, ultimately rejected by your loudspeakers if not before. A 10mV Vbe mismatch and a DC gain of 1,000 will set your stage output at 10V, which with typical 11V peak outputs will kill headroom. However Vbe for two transistors from the same reel is likely to be better than 10mV, and audio gain of 1,000 can usually be designed to an AC gain of unity. A 5mV output offset is nothing compared to 11V peaks.

laugh :grin:

[quote author="PRR"]In circuits like you describe, Hfe match only makes sense if the external DC resistance is matched. In many audio systems, it isn't. In that case, selecting for high Hfe does reduce errors, but buying from the high-Hfe bin of a high-Hfe type usually takes you as far as the basic design will go.

Vbe mis-match is multiplied by DC gain and appears at your output, ultimately rejected by your loudspeakers if not before. A 10mV Vbe mismatch and a DC gain of 1,000 will set your stage output at 10V, which with typical 11V peak outputs will kill headroom. However Vbe for two transistors from the same reel is likely to be better than 10mV, and audio gain of 1,000 can usually be designed to an AC gain of unity. A 5mV output offset is nothing compared to 11V peaks.[/quote]

OK, I did not under stand any of this. I going to go back to square one and re-read about transistors again. Then I will come back to this and see if it makes anymore sense!

The one thing I did get is that there apears to be many parameters that can be match on transistors. In the case of Opamps for preamps, the Vbe seems to be more critical. Now I just have to look up what that means!

Yours truely, the electronic idiot!

Joe

confused now. :shock: :? :?

So is transister matching in 1731s and 2520s in the same category as matching output tubes in amps (which I never bother with)? I put off stuffing my doa boards because I don't get the matching thing either.

Kiira

Now to confuse you even more

I tend to think of Si BJTs as small med and large PNP NPN and often use them that way. I try to make any Si in that size work by the circuit around the transistor. For more fun look up SMD parts.

Yes there are different die sizes and masks for making transistors BJT and Fet. Some of the older data books would have info on the process used.

Lets take fets, when you see different numbers sometimes all it is is a different case, pinout or IDSS sort same process step.

I believe I read years ago that the 2n3055 was a power BJT that could be ANY DIE that met the min specs.
That said when I repaired laser power supplies the pass bank was a large darlington 10 2n3055 in parallel with emitter bal resistors and a 2n3055 driver, used as a constant current pass device into a argon tube as a constant voltage device for constant power(a gas laser can act like a voltage reg tube)
We found only Mot and RCA 2n3055s would last I would match all ten from batchs of a few hundred when they came from the same batch they often matched well enought to just install them, never did that because it was a pain to change them so every thing got tested before being installed.

When I see some IC opamp in a circuit design sometimes I note the input balance can't be good because the the input pair has different base or gate currents to ground or with a DIY module say the emitter resistors if used are not matched. so any transistor matching is almost a waste of time. Now sometimes you might see this but often it will be cap coupled and the "designer" is aware there is an offset but it is small compared to the signal swing


You can "force" an input pair to be better matched with emitter resistors but the gain goes down(this can be good) but the noise goes up look at a lm318.

buy a bunch of the same batch/reel test them on a DMM and use them if the offset is to much try another transistor

I alway try to match any power devices that are used in Parallel for better current sharing even with emitter resistors

If you read at Aron's stompbox page you might find some of my effect schematics in the schematics section most have the bias dependance on Hfe "designed" out of them.

Global feedback and absurdly large open loop feedback were invented and refined to eliminate/reduce the need for measuring and matching of components. (but doesn't help with DC offset due to Vbe mismatch on the input pair)

If you are building a Melcor or a 2520 look alike and the circuit you use it in has a capacitor on the output of the opamp you might not want to worry about matching. Get the parts from the same batch though. With the price of 100 transistors hitting $6 it is worth to get a bunch. The Fairchild "CTA" parts on tape are always close in parameters.

Above all, don't put off building one. Build one with what you have now and with what you can easily source. You can come back later and build more when you finally figured out what paramateres you want to match why and how.

If you want to try discrete opamps that need no matching you can build the FetBloak or the BigBloak. They were designed to require NO matching of components.

Cheers,
Tamas

Picking two similar output tubes makes some sense, because they work so hard that it affects distortion, and because they live so close to melt-down that a big mismatch in bias will make one tube carry most of the load while the other loafs.

Still, matching output tubes is done more often than needed. We sure never did it back in the old days. There may be more justification with the loose tolerances on recently-made tubes.

Matching input transistors is critical for DC accuracy. We don't listen to DC.

Matching input transistors can reduce subtle problems with DC in the signal path, but only if the op-amp is properly balanced. The 990 is nearly perfect (and the good vendors hand-trim them to practical perfection). The oldest audio opamps will be badly "unbalanced" no matter how perfect the input pair is.

The 2520 has a trim: the "TRIM" pin between the 68K and the 22K resistors near the input pair is for a 25K-0 ohm variable resistor connected to V+. Put the amp in-circuit, the output DC should be mighty close to zero Volts un-trimmed, the TRIM resistor lets you get the last milliVolt out. However, it will also work fine with 22K at the TRIM pin, or even with TRIM left unconnected: the input pair will run about 10% off-balance, a few mV of DC error. This also increased measured THD, but in music there is reason to maybe prefer the off-balance condition.

But all audio opamps should work without input matching, especially if you pull two transistors from the same tape. You have to get into the outer limits of audio refinement before input pair matching matters; it may be one of the last things to trim.

Really: just stuff it and make music.

Really: just stuff it and make music

Click to expand...

LOL...Yep thats what I am going to do! I will learn as I go! Not stuffing the boards doesn't seem to be getting me anywhere!

Thanks for the help. I ordered 50 of each transistor 550/560 so I will have a small batch to work with.

Joe