Thanks all - yup waiting for some bits and a faster scope.
Fair assumption that the amp by design is stable and works. That is to say there's a fault here.
Bit reluctant to reduce R23/R31 value IF V2/V3 cathode follower stage can clip.........but cant at this time see how common anode V2/V3 grid could go positive wrt cathode (blocking) - need to understand that one ?
As to output stage, reading of 6550 curves with Vg2 ~ 400V and Va 600+ suggests Vg =70V reasonable for Ik -20mA or so? Besides, op stage is well balanced when it works at full power and the fault (when its there) seems to be in the drive. Already systematically gone through the 6, swapping out with known good pair, no effect on fault. Naturally don't have 6 spare 6550s!
Thanks again for all help and ideas.
Fair assumption that the amp by design is stable and works. That is to say there's a fault here.
Bit reluctant to reduce R23/R31 value IF V2/V3 cathode follower stage can clip.........but cant at this time see how common anode V2/V3 grid could go positive wrt cathode (blocking) - need to understand that one ?
As to output stage, reading of 6550 curves with Vg2 ~ 400V and Va 600+ suggests Vg =70V reasonable for Ik -20mA or so? Besides, op stage is well balanced when it works at full power and the fault (when its there) seems to be in the drive. Already systematically gone through the 6, swapping out with known good pair, no effect on fault. Naturally don't have 6 spare 6550s!
Thanks again for all help and ideas.
I don’t think they actually run AB2. It’s just to provide low impedance for the grid bias. Three paralleled 6550’s would need ~15k or so grid leak. With a follower driving it you get 1/gm, in parallel with the tail resistor.
Just reading about grid blocking here's extract from Radiotron 4th Edition 1953 Ch 2.2
I dont profess to fully understand, but the symptom circled red perhaps seems a good fit here?
I dont profess to fully understand, but the symptom circled red perhaps seems a good fit here?
This book is known to be a very reliable source of info. I have the printed version edited in 1999. I after searched and dowloaded several of the reference he gave in each chapter. Most from magazines.
But in your paticular issue, I continue with a suspicion of oscillations. An easy way to corroborate it is playing some audio at low level as to prevent the output stage to block. If still persists, I am almost sure the amp oscillates.
But in your paticular issue, I continue with a suspicion of oscillations. An easy way to corroborate it is playing some audio at low level as to prevent the output stage to block. If still persists, I am almost sure the amp oscillates.
Looking at the curves for the 6550 I arrive somewhere between -45 to -50 V for 20 mA, the bias voltage indicated on the schematic looks OK to me. If you only get the required bias current at -70V something is definitely wrong.
But it could still be the driver, most probably it is indeed oscillating like others also mentioned. But the cause of the oscillation might very well be one or more output tubes sailing west.
The part of the Radiotron manual that you showed is about tubes that are supposed to be used at zero bias, and have their grids driven positively (like the 811A). A 6550 will die quickly when used in this way. In practice, the driver circuit is not able to make the grid enter the "grid blocking" region. The blocking distortion that we're talking here is caused by a different phenomenon, and I guess we'd do better by calling it farting distortion to keep the phenomena apart. You can read more about it here: https://www.aikenamps.com/index.php/what-is-blocking-distortion
You'll find that most of the measures mentioned in the article to prevent the amp from farting are actually taken in the SVT. This is not strange, because musical instrument amps are often severely overdriven, and it is the overdrive behavior that determines most of the tone of the amp. As nobody wants a bass guitar to sound like someone's trouser trumpet, it's best to leave those measures in place.
Thanks I suppose the Radiotron chapter might describe any valve which shows inflection in its grid curves .......no idea if that includes ECC82/12AU7. But 'latch up' as described would seem a good fit so something to keep in mind. At issue here is V2/V3 behaviour which are ECC82/12AU7s - not the 6550s - of course.
I'll check again bias potential. But the power stage appears healthy and happy at Ik 20mA or so per valve. both sides and proper B+ and screen grid potentials.
I'll check again bias potential. But the power stage appears healthy and happy at Ik 20mA or so per valve. both sides and proper B+ and screen grid potentials.
The fault happens even at low levels - I've been using 500Hz sine test tone at just a few watts (that's loud and annoying enough !)
They drive grids of 12AU7 which has very low leakage and is rated 1Meg grid path resistance. Also this is actually one of the most reliable high-power amps ever made. The series resistors ARE high because otherwise when the power stage touches AB2 (rare but it does happen in this amp) the poor 12AU7 will bust a gut pulling-up three 6550 grids positive. With 150k in the 12AU7 grid path it is unlikely to burn anything.
That said: I too am mystified how it can be one or the other randomly. I know in the heat of battle I sometimes get mixed up but Dog seems to have it on the bench for calm observation. There are many ways an SVT can become ill but good driver DC with no audio makes no sense.
I don't have all the experience of the contributors here, but I once had an infuriating fault which I eventually traced to a poor connection in one of the octal pins of the output tube (in a loctal-octal adaptor I had made, so fulfilling the usual support question 'what did you do last?'). It tested fine in my tube tester (using loctal or octal), but not in the amp.
Perhaps there is something that is not electrical that is a facrtor here? Is it always the same part of the circuit that has the issue, even if all tubes are swapped around?
Perhaps there is something that is not electrical that is a facrtor here? Is it always the same part of the circuit that has the issue, even if all tubes are swapped around?
Yes indeed. The amp is calmly on the bench and been a puzlzle for some weeks now, on and off. The fault is repeatable after a few minutes, and certainly affects at random one or the other half of drive circuit. And sometimes both. Im reluctant to change parts at random. Besides its interesting and Id really like to understand. That said, I do believe Ive been through the 'usual' suspects.
As I wait for parts and retrieve my faster scope, Im also away from the bench for a few days. So that gives good chance to think and make a cunning plan.
Thanks to all ideas, presently Im led to favour rf oscillation as most likely. Perhaps implicated is grid blocking (Radiotron meaning).
I still cant see how V2/V3 ECC82/12AU7 cathode followers could ever have a positive grid versus cathode in this circuit? Unless perhaps V2/V3 circuits can oscillate. Cathodes each drive 6550 grids via three seperate 47k stopper resistors. Is 6550 grid current always into its grid ie a current sink? Can 6550 grid source current?
So maybe R23/R31 looks unnecessarily high by design, but that's quite probably my lack of understanding. Yes V2/V3 cathodes each drive 3x6550 grids perhaps sometimes into conduction on crests. However, ECC82/12AU7 is not lightweight and curves go up to Ik 35mA........
I forgot the 47k parts. Yes, the grid current (if any) is normally into the grid (electrons come out). A straight cathode-follower driving the grid can see large cathode current, OTOO a 1k resistor (+10V makes 10mA). But the 47ks swamp that. 48k/3 = 16k is not a dangerous load for 12AU7 (showing Ampeg were smart (or got educated on the 6146 version?)).
The other function of the 150ks is to reduce grid blocking by equalizing the RC constants positive and negative. (I pity the player who had to test this.) Also grid blocking would be clear in DC readings. Except of course the gol-darn always stops oscillating when you lay a meter-probe on it.
Isn't there an Ampeg-oriented forum? Or did they fade away?
Well Wiki (of all places) might offer the explanation for how V2/V3 150k grid resistors came to be. R23/R31. Apparently the very first Ampeg SVTs from the late 60's had the anode of the cathode followers fed from 220V whereas the preceding stage from 430V. And so the cathode followers could be overdriven as described here:
"The issues with the earliest version of the SVT were in the early design of the driver circuit, not the 6146B tube which is a stable and reliable tube. The earliest SVT driver circuit's design would on occasion result in blown 6146B tubes. Because the front end 12BH7 voltage amplifier is fed from the 430 V node, during loud transients and overloads this will produce an AC signal that far exceeds the 12BH7 follower that has 220 V on the plates. So when the follower grid is driven well over its own plate voltage, it saturates on the positive half of the signal and thus takes over the BIAS voltage, forcing it very positive. The time for this voltage to come back to normal is based on the time constant of the 150 kΩ mixer resistors and the coupling cap - but by this time it is too late, the bias is pushed too far into the positive and the current gets pushed through the 6146, resulting in blown tubes."
This was addressed by supplying the cathode followers from the screen grid supply in later issues c 385V. But the 150k grid stopper resistor has remained.
PS: The SVT here is a late issue based on 6550s and has the later circuit, of course.
"The issues with the earliest version of the SVT were in the early design of the driver circuit, not the 6146B tube which is a stable and reliable tube. The earliest SVT driver circuit's design would on occasion result in blown 6146B tubes. Because the front end 12BH7 voltage amplifier is fed from the 430 V node, during loud transients and overloads this will produce an AC signal that far exceeds the 12BH7 follower that has 220 V on the plates. So when the follower grid is driven well over its own plate voltage, it saturates on the positive half of the signal and thus takes over the BIAS voltage, forcing it very positive. The time for this voltage to come back to normal is based on the time constant of the 150 kΩ mixer resistors and the coupling cap - but by this time it is too late, the bias is pushed too far into the positive and the current gets pushed through the 6146, resulting in blown tubes."
This was addressed by supplying the cathode followers from the screen grid supply in later issues c 385V. But the 150k grid stopper resistor has remained.
PS: The SVT here is a late issue based on 6550s and has the later circuit, of course.
Last edited:
Thinking about this, apparently grids can source current, the negative gradient part of grid current versus grid potential curve from Radiotron posted here
https://www.diyaudio.com/community/...blem-ampeg-svt-300.395328/page-2#post-7257577
Into a resistive grid load, this would be positive feedback and unstable - and hence the Radiotron text? And latch up.
Worked on a couple of these so far, Intermittent problems for both. You're going to have to bite the bullet and reflow almost all the solder joints on the main board. They are notorious for microcracking solder joints due to the lead free solder used on modern environmentally friendly manufacturing processes
As power tubes age, some of the cathode material may evaporate onto the grid over time. When that happens, the grids start to source current (and pull themself towards more positive bias) even at negative grid voltage, because they start to work like a weak cathode. If the grid starts to draw significant current, the bias voltage on the cathode of the CF will be lower than the voltage on the power tubes' grids. That's why I asked you to check the bias voltage directly at pin 5 of each individual power tube and see if you spot any noticeable differences. If you spot nothing out of the extraordinary, at least you have ruled out the possibility of the output stage upsetting things.
Still chewing the cud for a few days til the parts, scope me and the amp can be together again.
timpert wrote:
As power tubes age, some of the cathode material may evaporate onto the grid over time. When that happens, the grids start to source current (and pull themself towards more positive bias) even at negative grid voltage, because they start to work like a weak cathode.
That would be positive feedback. And runaway - depending on grid circuit resistance (load line) according to Radiotron. Toward a latch up at point B on Radiotron's example curve. I suppose that might happen for any valve with a negative slope to part of its grid i-v characteristic, in the wrong grid circuit. Not just ageing power valves. No idea if ECC82/12AT7 or 6550 show that characteristic.
This SVT is a later model, so includes seperate 10R cathode resistors for each 6550. A circuit senses bias current through each resistor, and lights green or red leds bias in/out of range for all 6 6550s in 2 halfs of 3. This is one reason to believe all 6550s are biased properly, even when fault occurs. Both green lights stay on. Seperate cathode Rs means its easy to verify and measure bias of each valve exactly. The cathode resistors are also sensed individually by a fault detection circuit for over-current, which shuts off and latches off the whole B+ transformer via a relay. This never trips.
Already systematically been through all 6 6550s replacing in pairs with known good pair. There would have to be an odd fault on 2 separate 6550s to defeat this check. I believe the 6550s and ancillary circuits are working fine. Will confirm exact bias when Im next in a few days. My notes say -70V, though its possible I measured V2/V3 grids rather than cathodes I suppose. Bias self-monitoring leds shows all in range.
timpert wrote:
As power tubes age, some of the cathode material may evaporate onto the grid over time. When that happens, the grids start to source current (and pull themself towards more positive bias) even at negative grid voltage, because they start to work like a weak cathode.
That would be positive feedback. And runaway - depending on grid circuit resistance (load line) according to Radiotron. Toward a latch up at point B on Radiotron's example curve. I suppose that might happen for any valve with a negative slope to part of its grid i-v characteristic, in the wrong grid circuit. Not just ageing power valves. No idea if ECC82/12AT7 or 6550 show that characteristic.
This SVT is a later model, so includes seperate 10R cathode resistors for each 6550. A circuit senses bias current through each resistor, and lights green or red leds bias in/out of range for all 6 6550s in 2 halfs of 3. This is one reason to believe all 6550s are biased properly, even when fault occurs. Both green lights stay on. Seperate cathode Rs means its easy to verify and measure bias of each valve exactly. The cathode resistors are also sensed individually by a fault detection circuit for over-current, which shuts off and latches off the whole B+ transformer via a relay. This never trips.
Already systematically been through all 6 6550s replacing in pairs with known good pair. There would have to be an odd fault on 2 separate 6550s to defeat this check. I believe the 6550s and ancillary circuits are working fine. Will confirm exact bias when Im next in a few days. My notes say -70V, though its possible I measured V2/V3 grids rather than cathodes I suppose. Bias self-monitoring leds shows all in range.
The process I described is indeed a possible cause of runaway. But if you measured bias on the CF grids, then your -70V figure becomes more credible. The 12AU7 is used at a very high anode voltage in this position (actually way outside its spec), so it needs quite a lot of negative voltage to bias properly, and this comes on top of grid bias of the output valves. I didn't know that you had deluxe protection circuitry in your amp, if that says all is well then it probably is.
I still wouldn't rule out a bad component or joint though. Especially because it typically starts after a few minutes of operation, that points to something thermal. Then I am thinking along this line: part heats up -> fault occurs -> oscillation starts -> oscillation stops when probed.
One more question: dues the fault occur silently (the amp just quits working properly without making a sound) or do you hear a loud pop or bang when the fault occurs?
I still wouldn't rule out a bad component or joint though. Especially because it typically starts after a few minutes of operation, that points to something thermal. Then I am thinking along this line: part heats up -> fault occurs -> oscillation starts -> oscillation stops when probed.
One more question: dues the fault occur silently (the amp just quits working properly without making a sound) or do you hear a loud pop or bang when the fault occurs?
Well you still might be right. I either measured -70V on pin 5 of 2 empty 6550 sockets (amp run up with 4 6550s) or with the amp chassis upside down on the cathodes of V2/V3. I dont remember, but at least thats what I intended. But its just possible we are both right and here's how.
IF those 6550s were sourcing about 600uA grid current each, there's a 47k grid stopper resistor on that would drop about 24V. Then -70V on V2/V3 cathode and -46V on 6550 grids could both be right. That seems plausible in context of total current in R24/R26 47k cathode resistors for V2/V3 returned to -180V ? The amp would bias properly too. V2/V3 would be a bit starved. If so, I doubt its design intention.
As you say, measurement will confirm - but that'll have to wait a few days unfortunately - I'm interested.
The fault comes on suddenly and over a second or so with increasing rustling crackle sound and distortion.
IF those 6550s were sourcing about 600uA grid current each, there's a 47k grid stopper resistor on that would drop about 24V. Then -70V on V2/V3 cathode and -46V on 6550 grids could both be right. That seems plausible in context of total current in R24/R26 47k cathode resistors for V2/V3 returned to -180V ? The amp would bias properly too. V2/V3 would be a bit starved. If so, I doubt its design intention.
As you say, measurement will confirm - but that'll have to wait a few days unfortunately - I'm interested.
The fault comes on suddenly and over a second or so with increasing rustling crackle sound and distortion.
There is no way that a 6550 G1 is going to supply 600 uA when properly biased, if the grid current reaches that value then the tube is shot. The 47k grid resistors of the power tubes should drop no more than a fraction of a volt if all is well.
Just a stupid question: how did you count tube pins? The right way is clockwise when seen from the bottom. If you expect to find a cathode at pin 2 of the 12AU7, but accidentally count counterclockwise, you'll actually probe a grid. I've made that mistake myself on more than one occasion, it is especially confusing when you work on a thing that contains both tubes and ICs and you work on both sides of a PCB.
Just a stupid question: how did you count tube pins? The right way is clockwise when seen from the bottom. If you expect to find a cathode at pin 2 of the 12AU7, but accidentally count counterclockwise, you'll actually probe a grid. I've made that mistake myself on more than one occasion, it is especially confusing when you work on a thing that contains both tubes and ICs and you work on both sides of a PCB.
I agree timpert, 600uA g1 current source for 6550s seems pretty unlikely. Just sayin 'IF' .....that might reconcile most things here. I doubt that though. nevertheless once you eliminate the impossible, whatever remains, no matter how improbable, must be the truth and all that.
Where sensible before taking things apart I remove power valve(s) and measure dc conditions from the socket top - so pins count counterclockwise from pin 1. Screen grid is on pin 4 before g1 on pin 5. In this case, the 6550 g1 (valve being absent) would have zero grid current so no drop across 47k stopper resistor.
Just saying, but if R24/26 did have 3x 600uA from g1s, V2/V3 would be close to cutoff..........
On the other hand I might have just mismeasured - we'll see.
Where sensible before taking things apart I remove power valve(s) and measure dc conditions from the socket top - so pins count counterclockwise from pin 1. Screen grid is on pin 4 before g1 on pin 5. In this case, the 6550 g1 (valve being absent) would have zero grid current so no drop across 47k stopper resistor.
Just saying, but if R24/26 did have 3x 600uA from g1s, V2/V3 would be close to cutoff..........
On the other hand I might have just mismeasured - we'll see.