Discussion in 'Tube Audio' started by kward, Dec 22, 2016.
I am sure it takes some time to get the data for a single curve no?
It does take time. About 20 minutes per grid curve. I didn't show it all, just up to 15 mA, but my measurements took each one up to 15 watts plate dissipation and that took 20 minutes per curve. I also did 5V increments.
@kward, I think you mentioned that at short circuit (hopefully rare) that you run right at 42W plate dissipation? I guess that implies that one really should be using 6550As, yes? I note this because I was about to buy at quad of JJ 6550s, then realized they have the lower dissipation rating. (I ended up ordering some Valve Art 6550As ... price was right, anyway...)
Correct, with the screen voltage I specified of 116V relative to cathode, and the HV plate voltage used, a short on the outputs does run those plates at ~42 watts.
I'm not sure what the difference is electrically between the 6550 and 6550A. The data sheets don't seem to show any differences. Both 6550 (Tung Sol) and 6550A (GE) support the same plate and screen dissipation maximums of 42 watts and 6 watts, respectively. It might be the only difference is the glass shape....?
The original Tung Sol 6550 (the real USA one) was spec'd at 35 watts plate dissipation. The later ''A'' model,manufactured by GE,upped the plate dissipation to 42 watts.
Arguments concerning the use of a different rating system have been raised,but GE's use of a laminated (5-ply) plate material may be a better explanation.
In any event,either brand is damned near indestructible,even when used at near-meltdown conditions.Keep the G1 resistance at the proper value,and Bobs your uncle
Some of the current-production 6550s have the original 35W rating (JJ, for example), which is why I asked for the clarification. Looks like all of the New Sensor 6550s sport the 42W rating (well, I can't say for sure about the Sovtek-labeled variant, but the others do).
As far as the new variants go,most companies seem to have done a cut and paste of some original-manufacturers spec sheet or just copied them verbatim. The JJ spec sheet seems to be most of the original T-S specs,anything New Sensor related seems to be GE. As to exactly what the actual specs are,who knows.
Just as an example,there was once (1980's) a beautiful looking Chinese made 6550.You would swear it was an exact (visually,at least) copy of an original Tung Sol.
Right up until you tried to use it.If it didn't fail almost immediately,and usually in a very spectacular manner,it produced virtually no output in any 6550 application. Maybe 40% to 60% of an original American tube.I never determined why,perhaps limited cathode capability,or perhaps the grid-plate relations were just wrong.In fact,there was a local shyster who was re-branding and selling these as Tung Sols!
Mind you,the specs I had said it was identical.
As we've all seen,many types on offer today are not even,in fact,the actual type they are described/labeled as,but whatever the manufacturer felt they could pass off as such.
And running them using the manufacturer supplied specs as maximum operating guidlines can be less than optimum,in terms of both lifespan and reliability.
Not very helpful,I know. I do believe someone once did a torture test and comparison of modern vs original 6550's,but I can't recall who.
yup, noticed this myself. I'm curious what the true data is.
I feel comfortable recommending Tung Sol Reissue 6550's for this application because that's what I tested with and I tortured them pretty good. Any 6550 from New Sensor I think will work fine--Sovtek, EH, Gold Lion, Tung Sol.
I probably would not recommend the JJ 6550 for this application.
Regardless, here's how I would test the tubes used in this power supply:
1. Run the HV outputs short circuited for 30 seconds. (doesn't matter what voltage is dialed in).
2. At max dialed in voltage, while wearing gloves, hold the positive and negative leads in your finger tips and repeatedly touch them together for 1/2 second, then release for 1/2 second. Repeat for 30 to 60 seconds. (this will be no sweat for the power supply to handle, if those tubes are good)
3. I would not recommend this routinely, but a real torture test is the "rasp short circuit test." This abuses the 317 regulator circuit at least as much as the power tubes. Here's the test: Dial in max voltage on the bench supply (480V). Using a metal rasp with the course side up, hook the negative terminal of the HV supply to one end of the rasp. Wearing gloves, run the HV lead over the top of the teeth of the rasp, pressing down somewhat firmly as you move the lead across. A closed mouth alligator clip lead or banana lead works well as it gives some surface area to contact the teeth.
Take a soothing shower. That third test was about as much excitement as I felt I wanted for one day. Holding ~500V in your finger tips is no casual matter either.
If the tubes don't arc or otherwise fail after those tests, they should be good to go for many years of trouble free service in this application.
Ok, another dumb question... for the 6550 screen supply, what are the current requirements there? I plan to use an AnTek toroid for the HV power transformer, so no 80Vac @ 250mA winding. Instead, I plan to use a small Triad 120Vac transformer without the voltage doubler, and I suspect a 100mA transformer will be more than sufficient.
The conditions under which the max screen current will flow is when the tube is operating at minimum voltage drop (i.e, when the output voltage is set to deliver 480V), and when the load is set to the lowest possible resistance value to cause the supply to deliver the maximum amount of output current, but without the power supply entering current limit mode. Under those conditions, the 6550 is running with a plate to cathode voltage of (approx) 25V, and the screen current (per tube) is approx 17 mA.
A short circuit condition is not the worst case situation for screen current. Under short circuit, the tube is dropping about 515V. So the plate is 400V above the screens under that condition (screens are always at 116V).
Or in other words, when plate voltage is above screen voltage, the screens don't get the brunt of the electron bombardment. The screen gets beat on much more heavily when there is less voltage at the plate than there is at the screen, so that the electrons from the cathode are drawn more heavily to the screen than to the plate. When there's max voltage drop across the tube (plate at 515V), the screens sail along like it's sunny day and pull only about 6 mA current, per tube.
So the upshot of this is, I assumed about 25 mA max screen current per tube under worst case screen current conditions (for a little headroom). For 3x 6550 tubes (which is what I am using in my build), you'd need a screen supply that can source up to 25 mA x 3 or 75 mA DC current (for 3x tubes). Last thing you want (I think) is a screen supply that sags a bunch, because then current limiting will not be even across the voltage output range. That's not a requirement of course, but it just seems more consistent, and it's free if you make your screen supply stout enough.
Heh, fair enough. The F6-120 fits the bill at 250mA. (I plan to do a 4-tube / 400mA build.)
Hm... if that's the case, that 1.5K WW resistor seems a bit stout... under those worst-case assumptions, your screen supply would flop out of regulation because you'd drop 112V across that resistor. Definitely needs adjustment for 4x 6550s.
Now I'm trying to remember the detail here... (dang should have kept a full engineering log notebook on this build).
I believe that Edcor transformer put out about 85V on that winding, and I may have assumed a little less max current per screen also (like 20 mA insteadof 25 mA). But I am sure it is not flopping out of regulation under worst case conditiosn as I took some pretty good measurements at max screen current conditions, and I do remember I had plenty of headroom to regulate with still.
I am responding to the original first post. Why not buy a 0 to 400V power supply on ebay. They are about $200, most are 220V, some doesn't specify. Even if I have to buy a step up transformer, it's worth the while.
I use bench power supply to power up my SS amp during testing and building, it works. They are noisy, but who cares, it's just a temporary supply. You can't build one cheaper after you add up the parts and the chassis.
If I decide to design and build a tube amp, that would be my choice of power supply. It is 0 to 1A output current also.
Because building one is more satisfying?
For the fun of learning primarily.
Plus when done it provides an eminently useful tool.
So, with that doubler, I figure you've got about 220Vdc on that rail... and I guess the idea was that you could bleed off most of the "extra" voltage through that resistor before you got to the Zener? (You had a lot of extra...) If you did assume 20mA (rather than 25mA), that would still keep you above the Zener voltage even at full tilt.
In my case, I'll have 168Vdc on that rail (no doubler), so I won't have nearly so much voltage to throw away. A 330 ohm 10W resistor in that location will be about perfect in my case, and still have headroom at 100mA (if I assume I'll never exceed 80mA, a 470 ohm 10W resistor would be ideal).
That seems like the right thinking. One other thing to consider is how much current you dump down the Zener. My Zener runs a bit high. About 124V. So my screens run a little high too.
I could have "fixed" that by running the Zener at less current, I.e making the resistor in series with the Zener larger, but then I don't think it would regulate as well. So I picked a middle ground so to speak.
@kward, what was the thought process behind the choice of the 47K bleeder in the screen circuit?
Separate names with a comma.