Matching tubes/ valves

[Chuck H]

So, one more way to get the Fender Fart out of my Bassman!!!???


Justin

We've all heard the tales of production BF/SF Fender amps that most are common but some were just magic Usually attributed to the +/- 20% component tolerance making it possible that any two amps could have one or more component values differing by as much as 40% value. Maybe a little. But in my experience, with the exception of gross drift, the components I've tested in BF/SF amps have always been within 10% and usually much closer. That the components were rated for +/- 20% doesn't mean they were. Sort of like the warning on the back of the Windex bottle that says "DO NOT SPRAY INTO EYES". The component manufacturers were just covering their back sides.

Just as probable is that the amps had differing mismatch of the power tubes. Perhaps it wasn't any given amp that was special, but the tubes. And not matched tubes either. But nowadays owners of those amps wouldn't consider putting unmatched tubes into them. They'll rebuild the entire board for no reason though. Always with +/- 20% carbon comp resistors

[bob p]

I'm fond of data. I've tried a few approaches to matching and the best I've come up with measures bias current around the desired operating point, gm1 (transconductance for grid 1) and gm2 (transconductance for grid 2). I sort the results giving by gm2, then gm1 and finally bias so that bias has the highest priority. I've found that gm1 is somewhat correlated to bias and has quite a narrow spread compared to the bias. Also that gm2 is really quite stable. Here is a chart of gm1 vs bias and you can easily see how stable gm1 is compared to bias for a batch of about 100 EL34s.

The bias was measured at Plate=425, Screen=425 Grid1= -38



[ATTACH=CONFIG]40252[/ATTACH]

Love that data. By any chance do you mind sharing whose tubes those were?

From a pragmatic standpoint, gm1 appears to be so well conserved that I probably would stop measuring it... unless you've weeded out some tubes with variant gm1 that you haven't listed in your chart. When it comes to matching tubes that behave like the ones shown in your graph, my inclination would be to disregard gm1 as it is essentially constant, and bin the tubes according to bias current under test conditions.

Enzo brought up a valid question -- what's the point? IMO there isn't all that much point when it comes to running a single pair of tubes in a p-p amp. I've used some obscenely mismatched tubes at times, just to get rid of tubes that I can't match for a parallel application, and to prove that almost anything will work in a single pair push-pull guitar amp circuit. You can make a guitar amp get by with tubes that differ significantly in terms of parameters and as a guitar amp they still sound pretty good.

When it comes to running multiple pairs in parallel outputs, that's where matching really matters, and I think it's more about operational longevity and robustness in the field than it is about tone. If you're slamming a sextet of 6L6 using a Super Twin as a bass head, or a sextet of 6550 in an SVT, then it pays to have matched tubes. I match them under load conditions. I don't want to have one outlier compared to the other two in the triplet on that side of the p-p circuit, as it won't last very long if it draws the majority of the current and ends up doing all of the work while the others take a coffee break. I'll take the extra effort to match tubes under load conditions, but that's as far as I go. Some designers go the extra mile, and have built comparator circuits in their cathode supplies to monitor these sorts of thing in real time and to shut the amp down if things go seriously bad. So yes, there are some cases where matching really does matter, and people do actually pay attention to it, matching the tubes under heavy load conditions. But I think those cases are in the paralleled output cases, where it matters most.

Regarding that question of balancing phase inverters -- yes Enzo, people do actually do it. Maybe not the average repair shop, where close enough is good enough, but there are builders out there who obsess over these sorts of things, match their tubes, and tune their pi before they ship the amp. Here's a photo of an amp (not mine) where the builder built a trimpot in his paraphase circuit so that he could balance the PI under real world load conditions.



And on the subject of blueprinting, yeah, people here do that too, but we don't seem to talk about it all that much any more. I remember when Plexibreath spent a lot of time fussing over optimal resistor and capacitor values when he was designing the TAB. If you look at those schematics, there are specific oddball parts that were chosen, specifically to match components for the application and to zone-in on the real-world values that sound best within the large tolerance bands. Yes, people really do fuss over that type of minutiae, but they're typically not the repair guys who do it, and they're certainly not the typical guys who build generic fender clones. It shouldn't come as a surprise that the majority of the guys who fuss over these sorts of things are the obsessiveness compusive kind of people that are/were found on this forum, moreso than anywhere else.

[bob p]

We've all heard the tales of production BF/SF Fender amps that most are common but some were just magic Usually attributed to the +/- 20% component tolerance making it possible that any two amps could have one or more component values differing by as much as 40% value. Maybe a little. But in my experience, with the exception of gross drift, the components I've tested in BF/SF amps have always been within 10% and usually much closer...

Chuck, on the subject of parts tolerance and matching, it's well known that you could sample 100 different Fender Bassman amps, and that you'd find that many would sound alike, a few would be stinkers, and a select few would be absolutely magical amps. All of these amps were built with components that were within the 20% tolerances when they started out, and under the heat of years of use, most of those components have drifted in value. It's sad to think about it, but the chances are that people like you and I are only going to get our hands on the vast majority of the amps that sound the same (or the stinkers), and that we're never likely to get to sample the select few "magical" amps that are out there, because the guys who have them don't let them go. I guess it's sort of like the difference between the KF Wrecks and the clones. Subtle differences add up to many reproductions that sound OK, but don't capture the magic of those special amps.

Considering that you and I might never lay our hands on one of the truly magical amps, we have to rely on other people here to do the blueprinting of the magical amps when they come across them. Then, instead of using design center values found on the schematics, maybe we can find the right combination of off-center values that add up to having great tonality in an amp. In response to Enzo's question, I'd say the answer is a definite YES when it comes to blueprinting of individual components. Sadly I don't see as much sharing of that information as we used to see in the old days. that might be because in the era of the internet, just about everyone and his brother wants to run a boutique amp building business, and they want to consider any knowledge that they might have gained to be a trade secret.

[nickb]

Love that data. By any chance do you mind sharing whose tubes those were?

They were Shuguang. I found them to be surprisingly rugged. The thing is the initial factory quality is rather poor so I prescreen them to remove the bad eggs from the basket. The problem ones are usually very microphonic.

When it comes to running multiple pairs in parallel outputs, that's where matching really matters, and I think it's more about operational longevity and robustness in the field than it is about tone.

Quite so. I think also matters where you have just a pair unless you have or add a bias balance. I really don't have the time to modify so just pop in a matched pair. I work on a few 'HiFi' tube amps where matching is essential to meet the THD specs. Also they often have toriodal OPTs.

My conclusion was the same as you, that gm1 isn't worth measuring for this brand and type. However I have the capability and for zero extra effort so I do it. It adds value to the end product too.

I have to caution about extrapolating the conclusions to other tubes, at least yet. I did test Shuguang 6V6's and saw the same pattern.

[bob p]

They were Shuguang. I found them to be surprisingly rugged. The thing is the initial factory quality is rather poor so I prescreen them to remove the bad eggs from the basket. The problem ones are usually very microphonic.

Shuguang. The quality of Chinese tubes certainly seems to have gotten better in recent years.

[J M Fahey]

Modern parts are GOOD .

I donīt care much about tubes but SS is my bread and butter.
I used to buy sticks or boxes of power transistors and now MosFets and measure and match them when using many paralleled in big amps.

Waste of time, find them so close that if all same production code, I measure a few just to get an average value (which has a surprisingly small spread) and write it on some paper tape on the box or stick.

I always use small 0.1 ohm ballast resistors and drop across them *swamps* the tiny differences in Vbe or Vgs so I can freely mix them without trouble.

They are *that* consistent.

I guess this does not apply to tubes which have far larger spread .

[doctor]

My two cents.
Inaccuracies are the random values. For this reason their overall inaccuracy has an order of magnitude equal to square root of (aA + bB +cC +…), where A, B, C, etc. are squared values of inaccuracies and a, b, c,… are the weighting factors.
For example, if there are two equal components with equal inaccuracies of 10% (0.10) and weighting factor of 1, the overall inaccuracy will be square root of (0.01+0.01) = 0.14,
i.e. 14%.
There are many components in the device, so mutual impact of their inaccuracies can be very significant. It follows from this fact that 10% or 20% dismatching of tubes is not significant at all. Matching of tubes is performed for the marketing.
There is the real way to compensate as far as possible inaccuracies of components by biasing of tubes. For this purpose each tube must be biased separately (tubes should not be prematched) and a criterion is a minimum output distortions.
The big question - whether it is good? A feature of guitar tube amps is their specific distortions: both linear and, particularly, non-linear. This is what is called “valve’s sound” and this is that is appreciated by musicians. It is possible to manufacture a tube amp having distortions less than 0.1%, but his sound will be far from “cool sound of tube guitar amp”. Take for example tube Hi-Fi amps.

[gbono]

Interesting discussion but let's face it tubes (or valves) are ancient technology that only stay in production because customers "want" them - IMHO usually for all the wrong reasons. Matching is useless if the design/quality control isn't in place to provide a reliable product and from what I can see you have to test all product you buy to guarantee reliability (forget controlling parameters like capacitance and frequency response) - I don't believe the average customer/musician understands how lousy tube quality really is considering the cost of a tube. What they "understand" is that tube amplifiers have tone, warmth, (insert more marketing hype here).

[J M Fahey]

Well, tone and warmth are real, at least in a classic old style full tube amplifier (no master volume).

As far as you stray away from that ideal, say thrash high gain metal amps, or a basically SS amp with just a tube inside as a gimmick, the less internal tube contributions matter .

My customers invariably ask for a tube amp first, until I talk them into SS, which is my bread and butter.
I tell them: You need sweet, warm and clean for Jazz/Tango? : SS
You need brain liquefying thrash distortion? : SS
You need not really clean not really dirty sound which is controlled JUST by pick attack and finger pressure, no pedals or fooswitches involved? (Blues/old RockīnRoll/Keith Richards/Neil Young): tubes

[diydidi]

Great discussion so far. Thanks for all the replies. I have come to the conclusion that 10% is good for me.
Another thing that I have noticed is that guitarists have freakin super hearing. So if I can use matching as a sales pitch, and convince them that I'm one up on other repair shops, its money in my pocket and more work coming my way.
I'm not a guitar player, so just gonna do what seems logical.

[diydidi]

Modern parts are GOOD .

I donīt care much about tubes but SS is my bread and butter.
I used to buy sticks or boxes of power transistors and now MosFets and measure and match them when using many paralleled in big amps.

Waste of time, find them so close that if all same production code, I measure a few just to get an average value (which has a surprisingly small spread) and write it on some paper tape on the box or stick.

I always use small 0.1 ohm ballast resistors and drop across them *swamps* the tiny differences in Vbe or Vgs so I can freely mix them without trouble.

They are *that* consistent.

I guess this does not apply to tubes which have far larger spread .

JM, a bit off topic, but don't you find that mosfets have greater spread even if using same batch numbers? I find that BJT are far more even between batch numbers, but my bin of fets don't. Some don't even switch with the same Vgs..

[Randall]

Great discussion so far. I'm not a guitar player, so just gonna do what seems logical.

Boom! The guitar players vs. the non guitar players in a tube amp discussion.

[J M Fahey]

JM, a bit off topic, but don't you find that mosfets have greater spread even if using same batch numbers? I find that BJT are far more even between batch numbers, but my bin of fets don't. Some don't even switch with the same Vgs..

I made this jig to measure them, and still use it on every new batch, but found vlues SO consistent , say: 3.78 - 3.74 - 3.82 - 3.67 - 3.72 - ... and so on, all within 20 or 30mV with an off flyer not much further away, that I stopped measuring ALL, one by one and sticking little labels to each.


I use ballast emitter resistors, even my preferred very low value is enough to equalize them: 0.1 ohms compared to modern standard 0.22 or older 0.33 , common for bipolars.

Consider that a 0.1 ohm resistor drops 500mV @ 5A or 1V @ 10A , compared to that a 40 or 50mV (and usually smaller) difference in Vgs means my "uncalibrated" MosFets are within 10%

Canīt insult them and call them "unmatched" because they ARE matched ... not by me but at the Factory.

FWIW I donīt buy them from "a shop" which even if in good faith might have them mixed in a bin, but by the sticks from the official representative/importer, now absorbed by ARROW, a World monster .
Also bought from its arch-rival: AVNET but they sell me as if I were in Miami and I have to pass them through Customs myself, a PITA.

In general I buy those made by Fairchild, IRFP250 or, if unavailable, IRFP240.

Donīt like nor buy PCh ones, prefer to go quasi complementary, designed a circuit which I will NOT patent, prefer to keep it for myself, where the high Vgs needed to turn them fully ON, some 8 or 9V , is NOT lost , mine saturate fully and I get 5 or 6V peak on the load more than others

EDIT:
Sorry, just checked, no ARROW in Africa, only branch offices in Egypt and Morocco.

[gbono]

Note Fairchild (FSC) may soon become On Semiconductor..............I only see Vishay (IRFP250PBF) and Infineon (IRFP250NPBF/MPBF) still producing this part with FSC having obsoleted IRFP250 in 2005.

[J M Fahey]

Out of curiosity I checked my last batch: they have a fat "S" engraved with what looks like an "I" embedded on it so I guess they are Siliconix-Vishay.
Didnīt buy many because they were out of IRFP250 so I bought IRFP240 instead.

Any of them is overkill anyway, which is good.