Quad set of matched valves vs 2 pairs of matched valves.

[diydidi]

When replacing valves in smps like Classic 30's or 50's, and some 100W Marshalls that use EL34's, I usually try and buy matched Quads.
Is it a bad idea to use two different matched pairs in the place of a quad set?
What are the pros and cons in doing this? Obviously if two valves are running in parallel, and one has a much higher gain than the other, it will do more work than the weaker one.
Any thoughts?

[eschertron]

I read a discussion about matching pairs and sets here just a few days ago, and my take away from that was that it was much more important to have matched valves on the same side of the output, less so on opposite sides. Sorry I don't have a link to that thread.

[Richard]

If your matched pairs are close enough you will have a matched quad. If escherton read the thread I'm thinking of, I think another takeaway was that tubes are matched at one data point and there is no guarantee that they will be matched throughout the entire operating range.

[The Dude]

Agree to the above. + It would also matter if the amp had tubes matching or separate bias controls for each side. If there are controls to match the two halves then two separate matched sets wouldn't be a problem at all. Just put a set in each side and use the tubes matching or separate bias controls to match them. Saying that, even if not so, unless the tubes were grossly mismatched, it's not something I would worry about in the least.

[R.G.]

Let's talk about what "matching" means.

Tubes have a lot of things that get measured about them - mu, transconductance, plate resistance, electrode capacitances, lots. What matters a lot to audio tubes are some of the prosaic ones, how the stuff above relates to what I call:
- DC gain; how much current the tube lets through with zero bias on the gate, and how much DC it takes to turn it off. This isn't a straight line relationship, but for simplification, imagine it is. Many "matched" sets are matched for how much current flows through them with a given plate voltage, screen voltage, and grid reverse bias. So two DC-matched tubes will bias to close to the same DC current with the same grid off-bias
- AC gain; how much the tube current changes for small changes in grid voltage, given that it's biased somewhere into the middle of its operating range. AC gain is NOT constant for all DC bias conditions. This is obvious, as it's how bias-based tremolos work. So two identical tubes would have the same AC gain at all conditions. Two "AC gain matched" tubes would have the same AC gain at similar operating points. How many points are tested to gain match? Something tells me that no tube matcher provides you with that info, as it's a long tedious test, even with a computer curve matcher
- Then there are the variations of the two over varying conditions. Two identical tubes would have exactly the same DC current for all DC grid voltages. Two identical tubes would have exactly the same AC gain at all points as their AC gain varied over all operating conditions.

Tube matchers want to sell you the promise that the sound is magically better if you buy their matched tubes. The problem is, they haven't told you what they match. Even if they do, they haven't swept the conditions enough to tell you that the tubes are matched within X % over what range. They won't do this, as finding tubes matched for more than one condition makes the yield of "matched" tube nosedive into the ground.

On top of that, power amps with any feedback tend to cover up any AC gain offsets with the feedback, and some guitarists by actual test LIKE mismatched tubes for the very soft asymmetrical distortion.

My opinion is that it's better to buy unmatched tubes and institute separate DC bias controls to get the DC idle current matched in your OT so the OT isn't "walked" into the edge of satuation on one side or the other, and then to swap tubes till you like the sound, letting the AC gain variations fall where they may. It is entirely possible you'll like the sound better, and you will save a ton of money over buying dreams, er, promises of sonic purity with some unspecified matcher.

[Dave H]

I read a discussion about matching pairs and sets here just a few days ago, and my take away from that was that it was much more important to have matched valves on the same side of the output, less so on opposite sides.

Have I been doing it wrong all these years then? If I have two matched pairs I've always put one of each pair on opposite sides of the output to balance the DC in the OT. Say you have a matched pair which bias at 30mA and a second pair which bias at 40mA at the same bias voltage. If you put the 30mA pair on one side of the OT and the 40mA pair on the other side there will be 20mA DC imbalance across the OT.

[diydidi]

Now i'm really confused. Surely if all the valves share the same Cathode resistor, they need to be matched. Look at the schematic for a Vox AC30 (1964 Jennings era)

[eschertron]

Have I been doing it wrong all these years then? If I have two matched pairs I've always put one of each pair on opposite sides of the output to balance the DC in the OT. Say you have a matched pair which bias at 30mA and a second pair which bias at 40mA at the same bias voltage. If you put the 30mA pair on one side of the OT and the 40mA pair on the other side there will be 20mA DC imbalance across the OT.

My understanding is that the tubes on each side need to be 'equally yoked' to keep one from doing all the work and overheating while the other just idles. A completely different issue than DC bias in the OT. But you are right to bring that up, it does affect operation.

[Dave H]

Surely if all the valves share the same Cathode resistor, they need to be matched. Look at the schematic for a Vox AC30 (1964 Jennings era)

I agree. I was thinking of fixed biased amps (as in your OP). A common cathode resistor will make any DC bias differences worse than with the same tubes in fixed bias.

[R.G.]

It's back at that DC versus AC gain stuff. You want the DC balanced so the OT is happy, including the idle/crossover having equal and opposite amounts of B+ so the ripple on the B+ cancels. Ripple comes up if the idle DC is not matched and cancelling in the OT.

You want AC gain matched from side to side for least distortion, which may or may not be better than asymmetrical gains from side to side, and you want AC gain matched between the two tubes on one side (if the amp has four output tubes) to keep the tubes working in tandem, equally sharing the load as you mention. This will result in asymmetrical distortion as the gain between a matched pair on one side and a differently matched pair on the other side amplify each half of the AC wave a different amount.

Here's another quirk: phase inverters are good, but not perfect, and in general the AC signal to one side of the output tubes is not as big as the signal to the other side, causing asymmetrical drive on top of the different AC gains. Most differential PIs have a different gain on the driven versus passive side. You can tinker this by changing the resistors and that's why the plate resistors are not matched in many amps with a differential PI. It's also a way to even out the AC gains on each side for non-"matched" tubes if that's even something you want to do.

Bottom line: there are many things that could be matched, but it's unlikely that even "matched" tubes are matched for all of them.

I suspect that matching first came up when the big US and Euro tube makers started producing a smaller and smaller volume of tubes, making uniformity go down. It's hard to get more matched than sequential tubes from the old GE tube line. They worked HARD at making them all come out the same. When uniformity went down, tubes started having different DC biases, so bias balance came up in amps and then "matched pairs", then independent biases when they really got less uniform. Then we get to today when a tube may be almost anything within (we hope) the specs for that tube.

"Matching" offers the same appeal as quoting a single gain number for a transistor - it's easy to understand for laymen and beginners, and is a single handle that covers up a huge amount of variation that's really going on. But it's a truly vast oversimplification of the reality under it.

[eschertron]

"Matching" offers the same appeal as quoting a single gain number for a transistor - it's easy to understand for laymen and beginners, and is a single handle that covers up a huge amount of variation that's really going on. But it's a truly vast oversimplification of the reality under it.

So it's my hope, as a layman and a noob, that when I get a 'matched pair' (or set) then I'm getting tubes that are close together on the broad spectrum of uniformity.

As opposed to a random selection of tubes with wildly different characteristics, which is what you're suggesting today's production is. I don't care - and I probably don't know - if I get "3's" or "7's" or "reds" as long as the tubes in that set are pretty close in operation.

[R.G.]

That's a reasonable situation. It's better if, since you need to rebias anyway, that you put on an individual bias.

And dangitall I need to finish up that "stoplight" biaser setup for one to six tubes that shows blue(cold) green(good) red(hot) per tube.

That would let you do my preferred thing, which is to toss in unmatched tubes, bias them the same in about 10 seconds after warmup, then listen and see if the non-uniformity of AC characteristics is a problem or not.

[Dave H]

You want the DC balanced so the OT is happy

I'd like to turn that around and ask how much DC imbalance would it take to make the OT unhappy? Of course you are going to tell me it depends on the transformer so there's no easy answer.

[Enzo]

Consider this when worrying about matched tubes: Nobody in the amp making world installs matched sets of tubes in their amps. I mean Marshall, and Fender, and Peavey, and so on. I am sure some small boutique guys do. But point being all those thousands and thousands of amps with unbalanced tubes manage to thrive regardless.

[Dave H]

But point being all those thousands and thousands of amps with unbalanced tubes manage to thrive regardless.

Back when I started out if they both said 'Mullard EL34' on the box they were matched. For fun I once tried running my 4 tube amp with two 6V6 in one side and an EL34 in the other side. It sounded fine.