I had an incident with my KT88/KT90 PPP amps last night. I had them on the bench to check the bias and decided to hook up the function generator to perform another THD test with another 12AX7 driver. Well, here is what happened:
- I set the function generator to 20KHz full power and took the reading.
- When I set the generator to 1kHz, I heard a snap and smelled smoke. I looked at one of the tubes and I saw red plating.
- I powered down the amp and saw one of the 100 ohm screen stoppers was toasted.
I put in a new resistor and returned the amp to service. It is so far working fine now. What likely caused this anomalie? Was it the sudden jerk at full power from 20K to 1K?
- I set the function generator to 20KHz full power and took the reading.
- When I set the generator to 1kHz, I heard a snap and smelled smoke. I looked at one of the tubes and I saw red plating.
- I powered down the amp and saw one of the 100 ohm screen stoppers was toasted.
I put in a new resistor and returned the amp to service. It is so far working fine now. What likely caused this anomalie? Was it the sudden jerk at full power from 20K to 1K?
Making an audio amp pump out full RMS power with a uniform wave form is the most stressful thing you can do. That's why the ratings for TV HD finals, and xmtr finals are so down rated for "brick on the key" operation such as FM and packet. ICAS v. CCS ratings for xmtr finals, or 12W Pd limits for a tube that doesn't red plate while dissipating 20. In either case, it was meant to pump out full power with a uniform wave form for hours a day, for several years. Audio finals aren't so conservatively rated since any audio program (or AM or SSB) will spend most of the time bouncing around a low average power with the occasional excursion into max power or even clipping.
George has described how that happens: overload the screens, and they'll get red hot. If that happens, then the screen becomes an electron emitter. If the plate voltage soars, as it will when the signal suddenly disappears as you're switching the decade scale of the signal generator, and POOF! Goodbye screen stopper.
Just be thankful that the screen stopper was small enough to act as a fuse that popped before something else got borked.
I take it I just unknowingly abused my circuit.
Would a better way to handle testing be to lower the amplitude, shut off the generator, switch the frequency, then turn it on slowly again?
Would a better way to handle testing be to lower the amplitude, shut off the generator, switch the frequency, then turn it on slowly again?
Yep.
If you are running a test at max power, keep the test short, and watch for glowing screens, especially if you're running audio finals, or xmtr finals at ICAS specs. In either case, you could be running on the edge of the tube's SOA. Most tests don't require running max power, as the amp won't be running max power when playing an audio program unless it's been severely compressed (becoming distressingly common these days).
I've done two projects where I could run a max power sine wave for extended periods of time, but that was running 6BQ6GTAs at a max power of 30W, and that's pretty conservative for this type, as it could do 70+W if pushed a good deal harder, and with higher plate voltages.
The other project was with 807s operated at below ICAS specs. Audio finals might not run so conservatively, even when in spec, for the reasons listed above.
Turning down the signal generator (or the volume control) is a good idea. If your screens go red, best to not hit the finals with any discontinuities that could provoke sharply rising plate voltages.
Actually the explanation that my statement came from referred to a screen driven sweep tube where the screen voltage is driven by the audio signal. In sweep tube applications the screen voltage is always far less than the plate voltage so a glowing screen WILL lead to this kind of runaway since a glowing screen can emit electrons that will travel to the plate.
In a typical audio tube amp, like the KT88 the screen grid is usually operated at a voltage near, or equal to the plate voltage so prolonged screen emission should not occur since there is no higher positive voltage for the emitted electrons to travel to.
A glowing screen can emit electrons during the instant that the plate voltage swings above the screen voltage by 100 volts or more. Any element other than the cathode that glows red inside a tube will release metal ions that polute the vacuum inside the tube. This is often referred to as gas.
Probably. I don't know what your generator does when you change the frequency abruptly over such a wide range, but some, like an HP204C will emit all sorts of transients when the range switch is turned. Some of these are very low frequencies that can saturate the OPT leading to a temporary but extreme surge of current through the output tubes.
When playing with an amp that runs near the max spec for the tubes, it would be wise to turn the generators amplitude down, set the new freq, wait a few seconds for the generator to stabilize, then turn the amplitude back up. Actually switching the generator off and back on, may create an even bigger transient.
Have I blown screen stoppers in my "typical audio" amps under "normal" test conditions? Yes....
Every amp I build gets run through a series of torture tests to see if I can blow it up. I figure that if I can kill it with a few minutes of testing, someone else will find a way to kill it, so I try my best to "test" anything that I will publish. Everything will get tested at full power with continuous tones for several hours. This finds any parts that are not capable of a long life under more normal conditions.
Note that full power at 10 or 20 KHz is NOT a normal condition. Typical music does not have much energy in this region and a tube amp will draw more current (including screen grid current) at a high frequency due to the capacitance in the OPT. The excess current at high frequency will vary with each amp due to the capacitance effects.
A tube amp will also draw excessive current at low frequencies if the OPT saturates. When testing at or above 10KHz or below 100Hz watch the output tubes for signs of glowing grids or red plates (or monitor the plate current)and turn down the generator if either occur.
Every amp gets to eat my guitar playing at levels from mild to full metal racket (20+ db beyond clipping) for at least an hour if it is NOT a guitar amp, and far more if this is the intended application. This is where I blew screen stoppers in the SSE board.
When designing the SSE board about 8 years ago, the board passed all the normal and not so normal tests with no incidents. I plugged in my guitar preamp, dialed up my best Jimi Hendrix preset and cut loose. Within a minute one channel was dead. The screen stopper was open. I replaced it, and promptly blew another. New resistor, swap the output tubes with each other, repeat.....poof. This time the blown resistor was in the other channel. The blown resistor followed the output tube!
The resistors were 100 ohm 1 watt carbon film. They fried with minimum smoke and smell, some were slightly brown, some were almost normal. All had gone open where the metal ends were crimped over the coating on the ceramic body. I swapped the 1 watt resistors to 2 watt 100 ohm carbon film and never blew another one.
I put a scope probe on each end of the resistor and used the scopes differential mode to measure the voltage across the resistor. Typical testing revealed no unexpected results. However cranked guitar testing did!
The SSE can use 6L6GC's, EL34's and KT88 tubes. I tried several of each type. The screen current varies quite a bit from tube to tube. Old stock tubes were quite consistent from tube to tube of the same type. I tried 6 NOS GE 6550's, 5 used Mullard EL34's and several NOS 6L6GC's from different US brands. The new production tubes, especially the EL34 were all over the place. These tubes should have the grids aligned such that the screen grid is in the shadow of the control grid resulting in low screen current, but this is not always the case with new production tubes.
Some tubes, but not all, exhibited a spike in peak screen current as the tube goes into clipping. The peaks are short but were in excess of 100 mA. No screen grid glow was obsevred in a dark room. The tube that happened to blow the resistors was an EL34 and was one of the worst. This should not blow a 100 ohm 1 watt resistor, but it did blow several. The SSE has been a very reliable design with 2 watt resistors and the screen current hungry tubes seem to work fine for normal listening even when bass heavy music is played at the onset of clipping.
As suggested, maybe a larger wattage resistor is needed, or you can consider the ones you have as safety fuses in the event of an excessive current situation that should not happen in normal operation.
Reduce the generator level when changing freqs, and keep full power testing at high (above 10KHz) and low (below 40Hz depending on OPT size) frequencies to a short duration, unless you are monitoring tube current.
It's way more complicated than that. I have a big bunch of OPT's that were designed for guitar amps. There is zero interleaving no end bells, maximum cheapness. I got 200 of them for $16 each. The spec sheet says 6600 ohms, 80 VA from 80 Hz to 5 KHz. Real world testing puts the upper limit from 18KHz to 30 KHz depending on what's driving them. In a smallish HiFi amp using EL84's the open loop frequency response is 30 Hz to 20 KHz at 20 watts. When driven by 300B's in push pull zero feedback the same transformers go 25 Hz to 30 KHz. In both of these applications the distortion is rather high 3 to 5% at full power in the low bass region, 25 to 50 Hz due to transformer saturation. It drops reaching a low point under 1% in the 1 to 3 KHz range, then starts to increase as the frequency is increased up to the 5 to 10 KHz region then starts dropping again. At 20 KHz the distortion is quite low, under 1% on the 300B amp. Neither amp will make more than 10 watts (I don't remember the actual numbers) at 20 KHz. WHY??????
Well the transformer sucks, so it will not pass the frequencies above 20 KHz. When you run a single tone through an amp, most of the distortion is THD. If you run a 15KHz tone through an amp even if it distorts badly, the distortion can't pass through the OPT, so you don't see it or hear it! A 15 KHz tone in a P-P amp has most of the distortion at 45 KHz. Even if the amp is making 50% distortion it all gets lost in the OPT and only 10 watts of fundamental tone comes out of the OPT!
Granted this is worse case, but the amps, espercially the 300Beast as I call it sounds quite good, and every time I tried better OPT's, I wound up putting the junk ones back! Yes, this flies directly in the face of convention, and not all amps I built with these OPT's sound this nice, but some do.
I have found that crummy OPT's like to be fed by a low impedance source. The 300B is a low impedance triode. These OPT's also work quite well with sweep tubes with local (Schade type) feedback wrapped around them. I have an amp that will make 125 WPC with these transformers, but they tend to saturate on low frequencies at high power. I still get 50 watts at 30 Hz and 5% distortion. I recently got a pair of 100 watt Edcors so a comparison at high power levels will happen soon.
With a better OPT capable of passing the harmonic distortion the usual reason for higher distortion at higher frequencies is imbalance. Even the best OPT's will have different distributed capacitances between the two halves. Output tubes can be perfectly balanced on a tester at one particular set of parameters, but likely aren't well balanced across the entire range of conditions seen in the amp. The tube balance will also change with tube wear. Less than perfect OPT's usually will show a lower distortion at higher frequencies with a slight DC imbalance to help offset the built in imbalance in the OPT. I will usually set the DC (and AC balance if there is one) for lowest distortion at about 4 KHz and the average power level the amp sees in normal use. This is usually a low number, like 1 or 2 watts with my 87 db speakers, even less with high efficiency speakers, and a bit higher if you like to listen LOUD!
- Status
- Not open for further replies.