Tube Testing & Curve Tracing

[JMPGuitars]

There are obviously varying and polarized opinions on the subject of tube matching.

My opinion is that:

1. Most people that sell "matched" tubes are basically lying. They'll typically match a single variable like gm (or some B.S. magic number a tube tester arbitrarily decides, but is still typically a single variable). In all of the cases "matched" means within a given percentage of variance, but not everybody agrees on what that percentage is. As we know, tubes have more than 1 variable to consider.

2. If you want actually matched tubes, you would need to use a curve tracer and match the tubes' performance against the voltages that you will see in the circuit, and/or across the entire curve.

Now how many people do you know that actually does number 2?

I'm tempted to grab a set of tubes I purchased as matched and see how they actually perform on my curve tracer. Then compare them to other tubes not matched with the same magic numbers.

Thanks,
Josh

PS. For anybody that actually wants to accurately test / trace tubes, I'm working on a project PCB that people can build themselves. Eventually I'll probably also offer a prebuilt commercial version.

[TriodeLuvr]

My opinion is that:

1. Most people that sell "matched" tubes are basically lying.

LOL, I think you're right, Josh! But I also think McShane burns them in and takes the time to at least match them for Gm and idle current. He's a good source for people who can't do it themselves.

I match 6L6 types by first measuring Gm with a tube tester. Then I plug them into a modified 30W/ch stereo amplifier, heat them for at least 1/2 hour and record the bias current with all grids at the same voltage. Finally, if I'm feeling especially brave, I compare the distortion spectrum tube by tube at a medium power level. The latter is really time consuming and requires disabling NFB, so I don't do it often. Someday I'll build a small test jig that I can power from my bench supply and add switches for EL34, 6L6, EL84, etc.

You know, even with a curve tracer, the uncertainty of divergence at various screen voltages has to be considered. Two pentodes that test close with the screens at, say, +300V, might not be so close with the screens at +400V. For instrument amps though, the most important criteria is idle current. If that's balanced, hum will be minimal and the output transformer will be happy. I know, I know... preaching to the choir.

Jack

[JMPGuitars]

You know, even with a curve tracer, the uncertainty of divergence at various screen voltages has to be considered. Two pentodes that test close with the screens at, say, +300V, might not be so close with the screens at +400V. For instrument amps though, the most important criteria is idle current. If that's balanced, hum will be minimal and the output transformer will be happy. I know, I know... preaching to the choir.

Jack

That's the beauty of a good curve tracer. You can test at different screen voltages. My tester is similar to, but better than, a popular curve tracer (which was a huge PITA, and so was the creator if it). Here's an example of a recent test during beta testing the firmware for my device:

EL34.png

You can see specific measured data in the attached PDF file.

Thanks,
Josh

[JMPGuitars]

I think our firmware might actually be done. The results are pretty awesome. Here's some EL34 tests, I'll post the EL84 stuff soon too.

EL34-1_plot_1.png

EL34-2_plot_1.png

EL34-3_plot_1.png

EL34-4_plot_1.png

EL34-5_plot_1.png

EL34-6_plot_1.png

EL34-7_plot_1.png

EL34-8_plot_1.png

[JMPGuitars]

Here's the EL84 tests. These were done I think with 15 test points instead of 30, I might do higher sweep count tests eventually too if somebody has a specific request.

Tube 1:

EL84 v2-1_plot_1.png

EL84 v2-2_plot_1.png

EL84 v2-3_plot_1.png

EL84 v2-4_plot_1.png

EL84 v2-5_plot_1.png

Tube 2:

EL84 v2-6_plot_1.png

EL84 v2-7_plot_1.png

EL84 v2-8_plot_1.png

EL84 v2-9_plot_1.png

EL84 v2-10_plot_1.png

This is supposedly a matched pair. The only number written on the matching stickers is "33" - whatever that's supposed to mean. I'd post the data files too, but there's no point really. You can see that the results are not the same, so for them to have the same magical matching number doesn't make a lot of sense to me. Is it a huge variance? No, like I said, they generally match them within a margin of error. But the magic number makes it sound like they should be the same.

Don't worry ZP if you read this, I'm not advocating for (or against) matched tubes. Only stating observations.

Thanks,
Josh

[TriodeLuvr]

Josh, I'm designing several amplifiers around vintage output transformers that I removed from older commercial gear. The transformers are approximately 3.6K ohms or 3.8K ohms anode-to-anode. The original amplifiers used a tube type that is no longer available, and only the EL34 now matches this impedance well. It's problematic though, because whereas the original tubes were beam power, the EL34 is a true pentode. The screen current requirements of the EL34 are much more challenging, and because these are hi-fi amps, I can't just install a 1K or 1.5K screen resistor and call it a day.

One type of data that's not readily available in any of the EL34 sheets (including the extensive documentation provided by Mullard) is a chart showing anode current as a function of screen voltage. I'd like to see a graph of what happens to anode current with a fixed anode voltage of, say, +400V, a fixed grid voltage of perhaps -26V or -28V, and a screen that's swept from +250V to +400V.

This is something I can do point by point when my current project is off the bench, but that won't have the resolution of a curve tracer. Is this something you're able to do?

Jack

[JMPGuitars]

Jack, no problem at all.

EL34 v3-3_plot_1.png

I also ran the sweeps separately, but you've got all your data on here with a nice comparison in the PDF.

Let me know if you have any other tests you'd like me to run, or if you need more data points. I ran it with 20 points, but it can go higher.

Thanks,
Josh

[TriodeLuvr]

Well, that's a real eye opener! The effect of screen voltage on anode current is much more severe than what happens with 6L6 and 6V6 types.

Josh, what is the difference between the solid lines and dotted?

Many thanks for running this!

Jack

[JMPGuitars]

Well, that's a real eye opener! The effect of screen voltage on anode current is much more severe than what happens with 6L6 and 6V6 types.

Josh, what is the difference between the solid lines and dotted?

Many thanks for running this!

Jack

You're welcome!

The solid lines are Anode current, the dotted lines are Screen current.

But since you said that, now I want to run a similar test on a 6V6 and see what happens...but it will be a while before I can.

Because I love testing under stupid and/or extreme circumstances, I'm currently running a 300 point per sweep test, with 31 sweeps between 0 and -30 Vg, Anode voltage from 2 to 400, on an EL34. 39% complete after an hour. But it sure is pretty! lol

Thanks,
Josh

[JMPGuitars]

Here's my ridiculous / overkill test I did. 301 points from Va 2 to 400, with 31 sweeps of Vg from 0 to -30.

Ia:

EL34-1_plot_1.png

Is:

EL34-2_plot_1.png

I would attach the silly data file too, but it's 199 pages and over 6MB.

Thanks,
Josh

[TriodeLuvr]

What was the screen voltage during the plot of screen current? And what kind of hardware are you using for this work? This would be a terrific tool for designing new circuits, not just matching tubes.

Jack

[JMPGuitars]

What was the screen voltage during the plot of screen current? And what kind of hardware are you using for this work? This would be a terrific tool for designing new circuits, not just matching tubes.

Jack

Screen voltage was 400VDC. Both are the same test, but I plotted the results from A/S separately so they'd actually be visible.

The hardware is a tube tracer I designed in response to a string of bad experiences I had with a somewhat popular tracer...and it's creator. Ech.

The firmware for this appears to be good to go, so I will eventually be offering this as a PCB + programmed chip set, with a link to a BOM for advanced users to build it themselves. The only issue right now is that some of the parts are backordered until early next year. I probably have enough parts to sell a couple kits with the harder to find parts, but I'm not sure if I'm going to hoard those parts or not right now.

If this is something you (or anybody else) are interested in, you can email me at support at jmpguitars dot com to discuss it further. I don't have the website up for this, or any build docs done yet, so I dunno how fast anybody wants to rush on that. I miiiiiiight decide to auction the current unit I have after I build another one.

This is what my current build looks like:

_JMP2722.jpg

It's setup to test (from left to right):
12AX7 etc...
EF800 etc...
EF86
EL84
EL34/6L6/6V6 etc...

It can be setup for whatever tube types/sockets you want, but those are the ones I cared about, so that's what's on the tube pcb part.

Thanks,
Josh

[TriodeLuvr]

That's a terrific machine, and those are the same tubes I would be interested in tracing. I seem to remember seeing a tracer that used jumper wires to configure the various sockets. That would work for me too.

Let me know if you make any of this available, and what you think the cost might be. I have so many amplifier projects in the wings right now, not sure I can take on a DIY project. Still, it wouldn't hurt to know what it might cost, and I'm especially interested in this one if you decide to sell it. Very cool!

Jack

[JMPGuitars]

That's a terrific machine, and those are the same tubes I would be interested in tracing. I seem to remember seeing a tracer that used jumper wires to configure the various sockets. That would work for me too.

Let me know if you make any of this available, and what you think the cost might be. I have so many amplifier projects in the wings right now, not sure I can take on a DIY project. Still, it wouldn't hurt to know what it might cost, and I'm especially interested in this one if you decide to sell it. Very cool!

Jack

Thanks!

Well, for the PCBs and programmed chip I'm probably going to charge about $125 + shipping. Something like that. A bit more if people need me to do the SMT stuff for them. The BOM on Mouser for quality parts is between $176 to $200 depending on whether or not you need the SMT stuff. Then you need a laptop PSU (I paid $14 for the one I'm using, and it works great).

All that said, I don't know if I would sell the one I have for less than $1000.


The windows GUI is free/open source and awesome. I'm currently working on the theming for the GUI, and it's looking pretty noice:

futracer-gui.jpg

All the graphs above were exported from that.


In case you haven't guessed (or read my opinion somewhere else), the tracer that pissed me off was the utracer. So naturally mine is called the "FUtracer" in response to my experience with the utracer and its not-very-nice creator.

Thanks,
Josh

[TriodeLuvr]

Thanks Josh, I'll keep this in mind. Yes, the uTracer was what I was thinking about. Your comments are enough for me to steer clear.

Jack

[JMPGuitars]

Thanks Josh, I'll keep this in mind. Yes, the uTracer was what I was thinking about. Your comments are enough for me to steer clear.

Jack

Good call. It was an expensive nightmare. In the end, their design was poorly executed. The PCB quality was absolute trash, and their firmware was mediocre. On top of that, the supplied parts in their kit were all really low end. Their firmware, with the same connected hardware, couldn't give a proper trace of the same EL34 I posted tests of above. I'll post more of the story on the tracer website when I get around to building it, including comparisons of the same test sweeps.

Thanks,
Josh

[JMPGuitars]

We've been spending more time working on the GUI. I'm really happy how it's going. The plot tabs can now be resized or full screen. Resizing vertically is great to make the sweeps more visible:

EL34-1_plot_1.png

If anybody is wondering about why multiple tabs, because it's cool, that's why.

You can have separate tabs running during the test for different measurements. So that way if you have it auto-saving plots, it can save, for example, separate plots of Ia and Is, or other measurements at the same time. Yes, you can do 2 measurements on 1 plot, but that can be confusing. This way the information is nice and clear.

This screen current chart was run and auto-saved at the same time as the above:

EL34-1_plot_2.png

[lakelandmusic]

True, I run a small studio as well as restore guitars and amplifiers, and have a small online store in which vintage and NOS tubes are a part of my sales. 90% of online tube dealers sell them blindly, with mostly just Gm or pass\fail tests. In a never ending hunt for good tubes, I've bought thousands of them at auction over the last 15 years with varying test results from many different basic testers, and while most have been excellent, many of them tested so completely opposite of how they were presented, it just makes no sense to rely on those results.

I built a very nice curve tracer kit, and wouldn't think of selling or using a pair or quad not matched by relevant parameters, as well as a few other test circuits that track plate current variations while simulating voltage spikes, gas test, etc.

Great post, I just joined the forum, so hello as well!

[JMPGuitars]

Great post, I just joined the forum, so hello as well!

Thanks, and welcome to the site! I'm glad to hear there's people out there doing it the right way.

Thanks,
Josh

[southboundsuarez]

Brilliant work! Absolutely brilliant.
As an avid user of "NOS" valves, I would love having something like this to play with whatever latest discovery of treasures that I may bring back from vintage hunting safari.
My current method involves testing transconductance on Hickok (which leaves alot to be desired) I then take them and match for idle current on a single ended amplifier test jig. The final and most definitive test is actually installation in the intended amplifier, where basically all previous derived data is pretty much moot anyways, and I audition for my perception of sound, tone and essentially playing response.
The most critical parameter at this stage is the tubes ability to remain free of microphonics, fizz and noise.
Not too scientific and very reliant upon personal perception.
Curious if there is a parameter within your data that might predict the suceptibility to a tubes inherent character to be or become microphonic?
Thank you for sharing your posts and keep up the great work!