Last year I was in Florida for the Thanksgiving week. While the rest of my friends were slugging it out at the mall, I drove from Orlando to Tampa to eat turkey with my brothers......Somehow my car took a detour on the way. I met Stan at ESRC for some Thanksgiving day shopping. My "deal" included several BIG sweep tubes, and 100 6DG6's (another 6W6 in disguise). Some will be "stress tested" using my new concept.
I have not (yet) tested enough 6W6's to see the large gain and distortion variation that Lynn saw. That should change. I have seen several failures in triode strapped pentodes over the years, even when the tubes were operated at the upper edges of (but within) the published specs.
I have managed to blow two 6W6's by operating them in triode mode at the upper edge of the specs shown for TV vertical sweep. They work fine for a while, often a long while, but go into a slow cook leading to an eventual runaway death. I have seen this in several triode wired pentodes, especially TV sweep tubes. These were running in triode mode in an SSE amp right at 300 volts. I switched to 6EZ5's which have a 300 volt screen grid, and there have been no failures.
At first I thought it was bad tubes, but it happened too often, and the runaway death part occurred when the amp was left idling. It finally dawned on my that there is no "idle" or "turn the volume down" situation on TV vertical sweep. It runs full on, "maximum volume" with a continuous 50 or 60 Hz sawtooth wave at its input every minute that the TV is on.
Idle is worse case dissipation on a class A amp, and at idle, the tube doesn't know whether it's triode, UL, or pentode wired. It's just cooking away with 300 volts applied to a 150 volt screen grid in the case of the 6W6.
That's it, and there is some info about torturing (oops, I mean testing) those little tubes scattered throughout that thread. You will even see the 150 watt KT88 schematic that started this thread in post #73.
The thinking that was provoked in that thread convinced me to restart some experiments that had failed miserably at least twice before. I was trying to squeeze maximum power from TV tubes without all the complex power supplies. This time I focused on running some of these tubes in triode mode where they would blow up before. Soon there will be some power output comparisons between the different circuit designs.
Thanks!!! for the spreadsheet. It gives me a starting point. I dont have much tube experience, just recapping a few organ and console amps that I plan on using for guitar. I have lots of high voltage experience though, I work as a lineman, so i fully understand my responsibilities for safety.
I've got two 700v ct xfmrs, variac, triad x68 isolation xfmr, dc 2-37v negative supply, and a bunch of sockets.
Hopefully I can get a test rig put together soon and get going.
I've got two 700v ct xfmrs, variac, triad x68 isolation xfmr, dc 2-37v negative supply, and a bunch of sockets.
Hopefully I can get a test rig put together soon and get going.
I have been playing with tube stuff for nearly 60 years. In the early years I blew up well over 90% of everything I touched, but it was all free stuff that I got from other peoples trash, so nobody cared.
Now these things cost money, so set your sights, and power supply voltages well below the numbers in that spreadsheet. Those were maximum levels coaxed carefully by someone who knew what they were doing yet still used expendable parts. I would recommend starting from the "typical operation" data for a given tube in the tube manuals. Get that working first, THEN turn it up.
Thank you for the spreadsheet! This is pure gold - I can't even begin to guess how many engineers, techs, DIY builders, inventors, and crackpots have tinkered with these tubes for the last four or five decades without ever even coming close to the kind of power outputs you coaxed out of them. Awesome work!
-Gnobuddy
In a strange coincidence, my wife came home from work at a fabric shop that is at least 70 years old, and being renovated. She reaches into her pocket and says I found this under some junk in the back room when we were cleaning up. She hands me a 50C5.
Another thing I forgot to mention that the 50B5 is the same tube internally as the 50C5, but often far easier to find and cheaper.
The 50B5 was the standard 7 pin miniature output tube until around 1950 when it was decided to rearrange its pinout to suit the new UL rules for product safety. These rules are now called "creepage rules".
Much of my testing is being done with 50B5's because I got 10 NOS tubes and all of my 50C5's are used.
Yeah, but they didn't have mosfets until recently, and by then nobody cared except us few crazy DIYers. Expect some more experiments in the near future.
Another thing I forgot to mention that the 50B5 is the same tube internally as the 50C5, but often far easier to find and cheaper.
The 50B5 was the standard 7 pin miniature output tube until around 1950 when it was decided to rearrange its pinout to suit the new UL rules for product safety. These rules are now called "creepage rules".
Much of my testing is being done with 50B5's because I got 10 NOS tubes and all of my 50C5's are used.
Yeah, but they didn't have mosfets until recently, and by then nobody cared except us few crazy DIYers. Expect some more experiments in the near future.
In case anyone else cares, the 25C5, 17C5, and 12C5 are all still on the dollar list at ESRC. (The 50B5 and 50C5 are not.)
If one abandons the "series heater string" idea, a 24V DC "brick" will conveniently power 25C5s with heaters in parallel, and groups of two 12AX7s with heaters in series.
A 12V power brick will conveniently power 12C5s and 12AX7s heaters in parallel.
An old 18.5 or 19V laptop brick will power 17C5 heaters (with a small series resistor added to drop the extra volt or two), and, as George pointed out in a previous thread, the one-buck 18FY6 is half-a-12AX7, and will also run happily on a laptop brick.
We are far from the OP's desire to blast eardrums to instant destruction with a pair of 6L6s set to stun, but what tube-head can resist the idea of extracting 30 watts RMS from a pair of one-buck radio tubes that have been despised and scorned for half a century? I have been slowly accumulating bits and pieces for such a build for some time now.
-Gnobuddy
Interesting...my rather inexact estimation gave a mutiplyer of 10.3, which isn't terribly far off.
Alas, the KT-88 Mu is only 8. I may need to go to 200 V on the screens at peak, but that is still quite a savings over 400 V.
This further convinces me that I'm on the right track. At peak, I should be at no less than 60 V on the plate and no more than 200 V on the screen. That's hardly ideal, but still much better than the 400 V G2 I would need with Class AB1 operation.
My working "bingo numbers" are 16W instantaneous on G2 and 2W instantaneous on G1. If you have any data on what dissipation constitutes a similar level of strees on G1 and G2 (for example, tube X's grid starts glowing at X mA and Y V, and its G2 starts glowing at A mA and B V), that would help. Ultimately, I want to "one hoss shay" this thing so that G1 and G2 are both operating at the same % of maximum average dissipation (hopefully far less than 100% for both!)
That's both impressive and encouraging. If I can get to where I need to go with 200 V G2 and +20 V G1, I may be able to get to the same place with only 100 V G2 and +32 V G1. So long as I am limited by screen dissipation, that's a good trade.
I think I see that for a sine wave or a more or less typical music signal, but I tend to think in terms of torture by a signal so clipped it might as well be a square wave. So I look for what happens at B+/2, since that's where the loadline crosses deepest into the plate dissipation curves.
And that's my other limiting case, even if I use G1 to lower stress on G2. The only saving grace is that at high plate current, the diode line limits just how low the plate can go.
Well, it was KT-88s, but I'm not the only participant here, and so long as someone is getting something out of the discussion, it's all good.
And to summarize, this is all helpful, and I think I'm on the right track with regard to dealing with the screen dissipation problem.
An idea I've had for estimating the maximum G1 dissipation relative to the published G2 dissipation: Can we estimate the surface area of each? If they both use the same size filament wire and lay it out in the same pattern, it might be a simple as measuring the size of the oval that forms each grid as it wraps around the posts. For example, if the oval for G2 has twice the perimeter length of G1, you probably can't dissipate more than twice as many watts through G2 as through G1. Admittedly, just thinking out loud here, but that might explain why we blow up screens but not grids. 1 W might be very conservative.
The electrons leave the cathode with very little velocity, but as they move through the electric field from cathode to anode, they accelerate dramatically.
This means electrons smashing into G1 (which is very, very close to the cathode) will carry much less energy per electron, and so cause much less heating. In fact, if G1 is negative compared to the cathode, the electrons are even slower when they hit G1 than they were when the just worked themselves free from the cathode (thermionic emission.)
By the time electrons make it to G2, which is hundreds of volts more positive than the cathode, they have accelerated a great deal, and will dump a lot more energy into the grid wires when they smash into it.
I did a simple Newtonian (not relativistic) calculation, and came up with an electron velocity of over 10,000 kilometers per second by the time they smash into an anode or screen at +400 volts from the cathode. Still only 4% of the speed of light, but geez, those electrons are moving when they smash into the screen grid or anode!
Put another way, ignoring electron mechanics and only looking at the big picture of volts and amps, if G1 is at 10 volts and we're pumping 10 mA into it, that is only 100 mW of dissipation. But if G2 is at 300 volts and there are 10 mA of screen grid current, that is 3000 mW of heat dissipation. In other words, it's harder to blow up G1 because the voltages there are so much lower, even in AB2...
-Gnobuddy
It would be nice if we could just scale up the power level by using a bigger tube, but there are other factors at work here. TV sweep tubes are often called "high perveyance" tubes. The simple explanation for this term is that the cathode is far bigger than it needs to be for a given job, and the "electron cloud" surrounding it will always have more than enough to support all possible uses short of a nuclear meltdown. The typical guitar store tubes simply run out of cathode emission at high current demands. The cheap radio tubes fall somewhere in between these extremes.
In theory, yes. In practice I have seen the plate yanked all the way to zero with the amp loaded into a resistor, and below zero into a speaker.
Refer back to this thread:
Tube sale at AES In post #98 I have a crude screen drive setup connected to some tubes that I got for $2 each.
In post #104 I lean on it hard enough to make 130 watts from a pair of tubes. I have one scope probe on the Screen grid, and the other on the plate. The screen drive voltage goes from about -15 volts to +250 volts. The plate voltage touches zero and settles to about +35 before heading to some extreme positive voltage above 1KV.
Post #105 shows the same amp turned down a bit. In post number 106 the load impedance was changed from 6600 ohms to 3300 ohms. The power went up to 154 watts.
Post #108 shows a crazy drive experiment.....it blew up.
Actually: they "can't see" the plate. Only G2. They move toward G2. G2 is what "sucks" electrons away from the cathode.
With good geometry, 90% of those electrons "miss" the thin G2 wire. But as you say: the ones which hit, hit hard.
They miss because at a distance they think G2 is a "sheet" of uniform potential. Not until they get close do they discern separate wires with gaps between. By then it is too late to swerve.
Ouch.
It has dawned on me that there is a fairly simple way to figure out just how much impedance I need for any given B+.
Just build my "ideal" output transformer, but start with "too much" impedance, and give it a kajillion taps on the secondary side. I can just put a big load resistor across the secondary side and...
First tap...no problem.
Second tap...no problem.
Twelfth tap...starting to redplate. Ok, back off from that...now we know.
Someone must have thought of this. Are there any obvious flaws in this plan?
I wanted to come back to this, now that I'm starting to understand dual/twin/crazy drive and how it might apply to this project. Time to go back to the plates and the bottom right corner of the load line. I wanted to put this in a separate post so that it didn't get buried in all the screen and grid talk.
So far I've dug up:
1. One or more serial nonlinear devices from plate to ground that clamp very high voltage to ground.
2. One or more serial head-to-tail pairs of nonlinear devices from output tranformer primary to the other primary that prevent very high voltages from building up across the primary terminals.
I suppose you could describe both of these approaches as using "flyback diodes", except that I've also seen MOVs suggested for this.
I have read a claim from at least one amp designer that a string of three 1N4007s from plate to ground will be transparent for all situations save when arcing would have occurred. I have seen traces of this circuit coping with squarewaves, and to my eyes, the only effect was to clip some ringing that followed each leading edge, ringing at an inaudibly high frequency.
I take all such claims with a dose of salt. Any observations on this?
You could do that, but there is a much easier way, that I habe been using for years. Start with a good OPT that has "too much" impedance. For my highish power experiments I use 6600 ohms because I have several. The transformer should be able to handle as much power as you are going to throw at it at some mid range frequency. If it has multiple secondary taps like the usual 0-4-8-16 ohm winding, that's even better.
Note that the power rating for most OPT's is pure BS. The best measure of an OPT's power handling is weight! Say you have an OPT like mine which is specified for 6600 ohms to 0-4-8-16, 80VA at 80 Hz. This is a true "80 watt" spec from engineering which quotes a power level at a specific frequency. The usual "rule of thumb" is that it takes 4 X the weight (iron and copper) to extend that frequency range downward one octave (to 40 Hz). So this OPT is good for maybe 20 watts at 40 Hz, and testing seems to agree with this. On the flip side, I should be able to feed it a LOT more power at a higher frequency. There is of course a voltage limit where the insulation could fail, and a current limit where the wire could melt, but I have stuffed over 180 watts through these guys at 400 Hz without issue.
Most HiFi testing is done at 1KHz. I tend to use 400 Hz for guitar amps since it's more in the middle of a guitar's frequency range. Do not test big power levels at low frequency with a wimpy OPT. Saturation will blow tubes!
The OPT will have the same impedance ratio for a reasonable load over a fairly wide range, so instead of building an OPT with a bunch of different ratios, we build a load with a bunch of different impedances. You can collect a bunch of big resistors in the low ohm range and use them in series parallel combinations to make several different loads in the 2 to 10 ohm range as I have done, or you can buy a .5 ohm, 1 ohm, 2 ohm and two 4 ohm resistors, which in combination can make anything from .5 ohms to 11.5 ohms.
Use resistors capable of handling a lot of power. If a load resistor blows, bad things will happen. I set one of my 6600 ohm OPT's on FIRE when the Radio Shack load resistor blew at somewhere north of 170 watts through my 80 watt OPT. It's the only one that I have blown, but an arc from plate to ground at 150+ watts severely magnetized another one....I successfully degaussed it, and it lives on.
Then apply the ratio from the transformer to the test load to get the load on the tube. My 6600 to 8 OPT is 5775 ohms with a 7 ohm load, 4950 ohms with a 6 ohm load, 4125 ohms with a 5 ohm load, and 3300 ohms with a 4 ohm load. This works pretty well over a 2:1 range, that's why I use a multi tapped OPT and switch taps when going beyond 2:1, so my OPT is als 3300 ohms with an 8 ohm resistor on the 16 ohm tap, and 2887.5 ohms with the 7 ohm load on the 16 ohm tap..... This way you don't need that many load resistors.
Not so easy. With big power the screen will glow red first as power is increased under light load, and the plate will glow first under heavy load (lower impedance). This range will move with different plate and screen voltages and drive schemes.
If you really want to squeeze maximum power out of a given tube you need to put the screen grid, and plate B+ on separate adjustable power supplies and make several spreadsheets like those I posted where plate and screen dissipation levels ate measured VS power output. Then you can really know what thew tube likes. You are looking for maximum plate efficiency with screen dissipation held to acceptable levels. It takes quite a bit of knob twisting for a given tube.
I tend to look at whatever OPT's I have available, then test those impedances only, even though there may be a better option.
Wouldn't it be simpler to have a fixed-ratio off the shelf output transformer that's roughly in the ballpark, and just vary the load resistance on the secondary side to find your optimum (reflected) primary impedance?
As an example, let's say your Tx has a 1000:1 impedance ratio, and you find you get optimum results with a 3.1 ohm resistive dummy load; so now you know the optimum primary Zaa is 3100 ohms.
Once you've found your optimum primary Zaa, you can then wind your own transformer with just the right ratio to create that primary impedance with the appropriate loudspeaker impedance connected to the secondary. In this case, to create 3100 ohms at the primary when your 8 ohm (or 4 ohm, or whatever) speaker is connected to the secondary.
FWIW, I've read more than a few on-line reports by owners and repair techs about amps which fried despite the protective diodes. Damage reports variously included all the usual victims: the OT itself, output valve sockets, output valves, carbon-tracked PCBs, et cetera.
I've never experienced this myself, so it's only hearsay. But it gives one pause for thought; unless every one of those online reports is false, it suggests the clamping diode approach is, at the least, not 100% effective in preventing collateral damage.
Personally, I question the effectiveness of using multiple reverse-biased diodes in series. Reverse leakage current varies widely from one diode to another, which means their effective resistance in reverse bias varies widely. Which means the three diodes do not share reverse voltage equally. Which in turn means a high probability of the diode with the lowest reverse leakage current failing due to excessive reverse voltage across it.
It might work better if equal-value resistors are wired across each of the three diodes. Maybe 1 mega ohm or something like that. (There will still be uneven voltage sharing when there is an AC signal voltage present, because of the varying junction capacitances, but if the parallel resistors are low enough, that should swamp out the effect of the small diode capacitances.)
My other observation is that the diode string approach is that it is based on the belief that if one end of the secondary is clamped to prevent it dropping below 0V, then the other end can't rise higher than 2 B+. But this relies on the OT having perfect transformer action; we know transformers have all sorts of imperfections, including leakage inductance and stray capacitance. Can we be sure both halves of the OT winding truly remain in perfectly balanced anti-phase under extreme conditions, of the sort that actually require protection circuitry to kick in?
-Gnobuddy
I see George got there long before I did, as usual. My apologies for essentially saying the same thing three hours later - I must not have refreshed the browser, and so never saw George's post on the (new) next page of the thread.
-Gnobuddy
There is a discussion thread in the tubes forum about this and it seems that there are many opinions, a few possible solutions, none of which are foolproof, or inaudible. My post #26 in that thread explains why the OPT bursts into flames when the load is removed at full power.
Tube output protection
It appears that preventing the voltage on the plates of the tubes from going below zero may prevent or reduce the risk of OPT damage.....experiments have shown that the usual diodes from plate to ground will affect the sound under extreme operating conditions.
Minor excursions below zero (OK, several hundred volts) are not uncommon on a guitar amp cranked into clipping feeding into a real guitar speaker. Clamping them to ground will eliminate some of the "fatness" in a distorted signal and replace it with a high frequency component. Note that this is memory from some experiments done years ago. My current guitar amp only makes 4 watts and will run just fine with no load despite the $5 OPT. It doesn't have enough power to blow itself up, kinda like an old slant 6 Dodge Dart.
Experiments with devices connected from plate to ground, especially gas tubes or some slowish MOV's reveal another issue. Once the protection device fires, it puts all of the energy in the power supply across half of your OPT. When this happens something will blow. These type of devices rely on a AC supply that crosses through zero every 10 to 20 mS, so they get a chance to reset. On DC they never get the chance until something blows.
As I will be building a big power (500 WPC) tube amp in the future, and my expensive Plitron OPT's are no longer available, I need to figure this out on a smaller scale with some expendable OPT's before blowing up something irreplaceable.
I am currently of the belief that multiple methods are needed on a big amp, with the majority of the protection on the secondary side of the OPT.
On a reasonably low powered amp it is often the case that a suitable power resistor wired across the speaker terminals and sized to waste a small fraction of the output power (2 to 5%) is sufficient to prevent a flaming disaster if the speaker is removed (or blows) at full crank since it provides a path for current flow. Perhaps a second resistor that eats more power could be activated at a level beyond max output voltage with a set of zener diodes in series.
For a bigger amp sensing the voltage across and the current through this resistor could be used to activate gain reduction in the amp. Yes, this is the definition of a compressor, and would be audible, but maybe acceptable if it doesn't activate until way beyond the amps expected maximum power.
As a last ditch fail safe gas discharge tubes, TVS diodes, or both, made to protect phone lines from lightning can be wired across the speaker lines to clamp the voltage if it rises to say twice the voltage seen at full crank.
Tube output protection
It appears that preventing the voltage on the plates of the tubes from going below zero may prevent or reduce the risk of OPT damage.....experiments have shown that the usual diodes from plate to ground will affect the sound under extreme operating conditions.
Minor excursions below zero (OK, several hundred volts) are not uncommon on a guitar amp cranked into clipping feeding into a real guitar speaker. Clamping them to ground will eliminate some of the "fatness" in a distorted signal and replace it with a high frequency component. Note that this is memory from some experiments done years ago. My current guitar amp only makes 4 watts and will run just fine with no load despite the $5 OPT. It doesn't have enough power to blow itself up, kinda like an old slant 6 Dodge Dart.
Experiments with devices connected from plate to ground, especially gas tubes or some slowish MOV's reveal another issue. Once the protection device fires, it puts all of the energy in the power supply across half of your OPT. When this happens something will blow. These type of devices rely on a AC supply that crosses through zero every 10 to 20 mS, so they get a chance to reset. On DC they never get the chance until something blows.
As I will be building a big power (500 WPC) tube amp in the future, and my expensive Plitron OPT's are no longer available, I need to figure this out on a smaller scale with some expendable OPT's before blowing up something irreplaceable.
I am currently of the belief that multiple methods are needed on a big amp, with the majority of the protection on the secondary side of the OPT.
On a reasonably low powered amp it is often the case that a suitable power resistor wired across the speaker terminals and sized to waste a small fraction of the output power (2 to 5%) is sufficient to prevent a flaming disaster if the speaker is removed (or blows) at full crank since it provides a path for current flow. Perhaps a second resistor that eats more power could be activated at a level beyond max output voltage with a set of zener diodes in series.
For a bigger amp sensing the voltage across and the current through this resistor could be used to activate gain reduction in the amp. Yes, this is the definition of a compressor, and would be audible, but maybe acceptable if it doesn't activate until way beyond the amps expected maximum power.
As a last ditch fail safe gas discharge tubes, TVS diodes, or both, made to protect phone lines from lightning can be wired across the speaker lines to clamp the voltage if it rises to say twice the voltage seen at full crank.
That would be a neat trick.
I helped a friend make his dead 6 cylinder Mercury Comet (Ford Falcon) into a running 5 cylinder by removing the broken piston and rod then sealing the oil hole in the crank with a hose clamp.
I had probably a dozen cars with the slant 6 of 1965 to 1970 vintage in all three displacements (170 CID, 198 CID, and 225 CID). The slant 6 got a new "emission friendly" cylinder head around 1975, and nobody could make those things run right in hot weather. All the cars were purchased in the zero to $250 range and junked or given away. I think I sold the Duster for around $50. Despite treating them like I do tubes, I never had a power train failure. The front suspension would fall apart, or the body would rust off of the car.
I have determined that a 1 ohm, 1.5 ohm, 2.7 ohm, and a 4.7 ohm in various series-parallel combinations can get me from 0.44 ohms to 9.9 ohms with no two steps being more than 11.5% apart (which ought to be plenty close enough with 10% resistors).
I have a sorted table of all the resulting resistances, with the circuit configuration for each resistance ("R3 in parallel with (other leg) R1 in series with (together) R2 and R4 in parallel"). I can post that somewhere on this site if there is any interest.
But I think I can do this separately, plate on one hand, and screen/grid on the other, right? My thoughts are to give it just enough screen and a squarewave signal on the grid to get it comfortably in the range of B+/2 without saturating. Then play with B+ and output impedance until I have some potentially useful combinations, and when that's done come back and check the better candidates at max power with a more sophisticated screen and grid driving system like the one you mentioned.
I think that if I try to map every combination of grid, screen, output impedance, and B+, I'll be handing off the project to my grandchildren at some point. But if I can "get in the ballpark" with regard to impedance and B+, then come back and map everything over a narrower range of possibilities, it becomes manageable.
Regardless, I'm running out of theory and paper research, and moving to designing some kind of test rig.
Currently leaning toward MOVs from B+ to each side of the primary. Mr. Robbins is very persuasive. (Read through that entire protection thread and also his updated (Dec 26 2018) paper on OPT protection.)
Also looking hard at an MOV or resistor there, too.
That makes good logical sense because the MOV is right across the two points where the troublesome energy is stored, in the OPT primary windings.
But I'm wondering what happens if the MOV starts conducting at the same time the corresponding output valve is also conducting hard. B+ is now clamped to ground directly via the MOV and valve in series. Will a fuse in the B+ line be enough to protect the output valve under these conditions?
-Gnobuddy
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.