These look the same, but snivets in 807s appear at the plate current x-over, not near the peaks. It could be snivets, but in this case, it looks more like marginal stability here. You could include screen stopper resistors. It may not help, but it doesn't hurt either.
You may need to take more measures to improve the phase margin of this design and/or reduce the NFB.
Hi, Miles and any others watching this thread,
I tried a few things on the bench to see what might affect those bubbles (snivets?) on the peaks of the waveforms in my previous post.
1. Less feedback: Added some extra FB resistance. This made the bubbles grow worse, or at least they scaled with the increased sine signal amplitude.
Removed the added test resistance. Back to the little bubbles.
2. Parallel capacitance to alongside the FB resistor (there had been none in the original design): Added the following:
-- 100 pf, 560 pf, etc., up to 1000 pf: no change
-- up to 1500 pf, slightly worse: the bubbles grew
-- beyond 1500 pf: the waveform flipped completely upside down (?) and was a smeared mess.
Removed the added capacitance. Back to the little bubbles.
3. Switched to the 10X mode on the probe tip, and it cleaned up completely, no bubbles, nice and symmetrical sine wave shape, and I could increase input test sine voltage further and still retain a clean waveform.
The amp runs cool, sounds great and has a very clean output (including square wave) after the OPTs. Could this have been a probe anomaly?
I'd love to hear and learn more about what might have been happening in these tests.
Thanks for any thoughts.
--Jeff
I tried a few things on the bench to see what might affect those bubbles (snivets?) on the peaks of the waveforms in my previous post.
1. Less feedback: Added some extra FB resistance. This made the bubbles grow worse, or at least they scaled with the increased sine signal amplitude.
Removed the added test resistance. Back to the little bubbles.
2. Parallel capacitance to alongside the FB resistor (there had been none in the original design): Added the following:
-- 100 pf, 560 pf, etc., up to 1000 pf: no change
-- up to 1500 pf, slightly worse: the bubbles grew
-- beyond 1500 pf: the waveform flipped completely upside down (?) and was a smeared mess.
Removed the added capacitance. Back to the little bubbles.
3. Switched to the 10X mode on the probe tip, and it cleaned up completely, no bubbles, nice and symmetrical sine wave shape, and I could increase input test sine voltage further and still retain a clean waveform.
The amp runs cool, sounds great and has a very clean output (including square wave) after the OPTs. Could this have been a probe anomaly?
I'd love to hear and learn more about what might have been happening in these tests.
Thanks for any thoughts.
--Jeff
If reducing the NFB, error correction, makes the error worse then the NFB loop probably isn't causing it. You never provided a schemo or other details. If the finals are running Class A, and these snivets are appearing at the peaks, then that's what I'd expect here: the final goes into plate current cutoff at the outer limit of Class A, and these Barkhausen oscillations break out. 6L6-oids seem particularly liable to this. I had the same problem with a design that used PP 807s, but that was a Class AB, and the snivets appeared between the peaks and zero crossings.
Can you have a look at the snivet itself to determine its frequency? Does it match the natural resonant frequency of the OPT? If it does, it's not likely related to the feedback loop.
Did you install screen stoppers? These can be 470R -- 1K5 connected to the screen pin with the least connection length. Some types like the 807 need these screen stoppers to prevent these snivets at plate current cutoff. Other power pents don't seem to have this problem, but adding 470R -- 680R screen stops can't hurt.
Jeff - it seems you're probe was connected to plate node, and so would have been at high DC, and with a large AC voltage. Was the probe within spec for that reading, before you changed to 10x ?
The CRO probe on 1x would likely have had quite a noticeable impedance compared to other circuit values on that node, so the ground point may have been an influence, just as much as the position of the probe relative to nearby input grids.
Modern day ferrite beads over 6L6/807 anode leads at the anode terminal can also be useful (similar to the 10 ohm that was often located in the anode cap for 807's).
The CRO probe on 1x would likely have had quite a noticeable impedance compared to other circuit values on that node, so the ground point may have been an influence, just as much as the position of the probe relative to nearby input grids.
Modern day ferrite beads over 6L6/807 anode leads at the anode terminal can also be useful (similar to the 10 ohm that was often located in the anode cap for 807's).
Yup, those are exactly the parasitic oscillations I've seen with 807s. BTW, try HF Reject on the scope for a better synch.
You will need to put some grid stopper resistors (1.5K will do) in series with the grids of the outputs.
With high amounts of feedback, consider this:
The transformer bandwidth is limited. As you approach the limits in either extreme, the phase response starts to rotate. At the point where it reaches 90°, it begins to tip into positive feedback, thus above the amp's useable response, you see some positive feedback.
The solution to this is to band limit the feedback network. It's not necessary to have much feedback at 30KHz as at 2KHz. And you'll avoid the oscillation you get when the transformer phase starts to lag.
You will need to put some grid stopper resistors (1.5K will do) in series with the grids of the outputs.
With high amounts of feedback, consider this:
The transformer bandwidth is limited. As you approach the limits in either extreme, the phase response starts to rotate. At the point where it reaches 90°, it begins to tip into positive feedback, thus above the amp's useable response, you see some positive feedback.
The solution to this is to band limit the feedback network. It's not necessary to have much feedback at 30KHz as at 2KHz. And you'll avoid the oscillation you get when the transformer phase starts to lag.
Great advice. Thanks to all of you.
I will increase the screen and grid stoppers and try a different ground point for the probe. I like the ferrite idea and have some to try. I recall building an amp where the HF was rolled off in the FB loop via a small cap to ground. So I have several things to try. Still working on the schematic but hope to share it soon.
Current setup, walking backward through the amp: The OPTs are large Edcor CXPP30-MS-5K and the P-P amp is putting out about 7 watts per channel, with the 6Y6GTs set in "near class B" bias (fixed bias backed off to cutoff and then pulled up in current until crossover distortion is gone, about 6-7 mA plate current). Regulated screens. Stoppers on the 6Y6GT grids are 1K and on the screens are 330 Ohm. I am trying a concertina phase splitter using a MOSFET as an experiment -- an IRFBC20 with a 10k gate stopper. The input tube is a 6922 with 1.5K stoppers and a CCS anode load at 8 mA per side. The splitter is direct coupled to the 6922 plate -- the source to ground leg voltage matches the 6922 plate voltage well. There's a lot of fun stuff going on in this setup!
I will increase the screen and grid stoppers and try a different ground point for the probe. I like the ferrite idea and have some to try. I recall building an amp where the HF was rolled off in the FB loop via a small cap to ground. So I have several things to try. Still working on the schematic but hope to share it soon.
Current setup, walking backward through the amp: The OPTs are large Edcor CXPP30-MS-5K and the P-P amp is putting out about 7 watts per channel, with the 6Y6GTs set in "near class B" bias (fixed bias backed off to cutoff and then pulled up in current until crossover distortion is gone, about 6-7 mA plate current). Regulated screens. Stoppers on the 6Y6GT grids are 1K and on the screens are 330 Ohm. I am trying a concertina phase splitter using a MOSFET as an experiment -- an IRFBC20 with a 10k gate stopper. The input tube is a 6922 with 1.5K stoppers and a CCS anode load at 8 mA per side. The splitter is direct coupled to the 6922 plate -- the source to ground leg voltage matches the 6922 plate voltage well. There's a lot of fun stuff going on in this setup!
Last edited:
I built an 6v6 PP amp with simple SS bridge rectifier and CRCR (220uF ,50Ohm 220 uF) filter
it sounded quite good . I needed the power transformer for another project and when I got back to that 6V6 amp I used Hammond 600 Vct transformer as a substitute. Because I could not use tube rectifier due to space/heat constrains I did full way bridge with 2x1200V Cree diodes and was forced to trim voltage with pseudo choke IP arrangement (6H/200mA/150Ohm -marginal but seems to be working fine ) and 4uF vintage PIO cap. For output C I paralleled 220uF to give the amp little more umphhh.
Now , the amp sounds rather thin , sharp and sibilant. I hooked it up recently to my big 2x15" + horn project with "experimental" crossover and it refused to play at all becoming a generator instead . Now the amp is a well proven stock ,old design. The only thing changed is PSU arrangement .Could it be the source of oscillation ?
it sounded quite good . I needed the power transformer for another project and when I got back to that 6V6 amp I used Hammond 600 Vct transformer as a substitute. Because I could not use tube rectifier due to space/heat constrains I did full way bridge with 2x1200V Cree diodes and was forced to trim voltage with pseudo choke IP arrangement (6H/200mA/150Ohm -marginal but seems to be working fine ) and 4uF vintage PIO cap. For output C I paralleled 220uF to give the amp little more umphhh.
Now , the amp sounds rather thin , sharp and sibilant. I hooked it up recently to my big 2x15" + horn project with "experimental" crossover and it refused to play at all becoming a generator instead . Now the amp is a well proven stock ,old design. The only thing changed is PSU arrangement .Could it be the source of oscillation ?
I see this thread has been going almost 5 years now but the subject is as current as ever. Here are a couple of of hard-to-track-down oscillation stories:
One day my wife start saying that UHF television pictures were pixelating. Soon she noticed that the issue arose most often during louder passages of music from my Dynaco ST-70. I had just installed a brand new set of EL-34s in the Dynaco. Power pentodes have a natural tendency toward parasitic oscillation but a resistor, often 1K, in series with each control grid damps it right out. Unless of course you have just installed a set of Chinese EL-34s that are especially prone to parasitic oscillation. I went back to using Russian finals and the problem has never returned. Now for the one that was really well hidden.
There was a vague sense that all was not right with the Dynaco. It seems the bass was flabby as they say. I placed an AC voltmeter, with DC block, across the plate supply filter cap and watched the meter as music was played and I can tell you voltage swings were very high for an amp that was running Class A PP. At the time I simply could not account for this. So I checked each filter cap section but they all measured fine. So for a while I tried to satisfy myself that all was well if I just kept a 200uf cap paralleled to the existing plate supply cap. But I knew it was not alright. Not quite. Then I got a lucky break.
I happened to be checking for hum level at the speaker terminals one day (with input coaxes disconnected) and saw a sizeable signal. By sizable I mean 11.4 volts AC! But I could not hear any of it it. I switched the meter over to Frequency Counter and it read 38.4Kz. I also saw that this was only on the left channel. Just like the story above this problem had followed a change I had made. I had rerouted a wire to a voltage amplifier control grid to mod the amp for a different driver and in doing so I had routed this wire right over the top of a wire headed to the screen grid of a nearby left chanel final. This being an ultralinear amp, the screen goes to a 43% tap (in this case) on the OPT. The insulation of both wires was fine but there was enough capacitive coupling to get oscillation. The fix was merely separating these two as much as practical. The oscillation was gone has not come back. I removed the previously added supplementary cap and bass has fine ever since.
Hope this helps.
One day my wife start saying that UHF television pictures were pixelating. Soon she noticed that the issue arose most often during louder passages of music from my Dynaco ST-70. I had just installed a brand new set of EL-34s in the Dynaco. Power pentodes have a natural tendency toward parasitic oscillation but a resistor, often 1K, in series with each control grid damps it right out. Unless of course you have just installed a set of Chinese EL-34s that are especially prone to parasitic oscillation. I went back to using Russian finals and the problem has never returned. Now for the one that was really well hidden.
There was a vague sense that all was not right with the Dynaco. It seems the bass was flabby as they say. I placed an AC voltmeter, with DC block, across the plate supply filter cap and watched the meter as music was played and I can tell you voltage swings were very high for an amp that was running Class A PP. At the time I simply could not account for this. So I checked each filter cap section but they all measured fine. So for a while I tried to satisfy myself that all was well if I just kept a 200uf cap paralleled to the existing plate supply cap. But I knew it was not alright. Not quite. Then I got a lucky break.
I happened to be checking for hum level at the speaker terminals one day (with input coaxes disconnected) and saw a sizeable signal. By sizable I mean 11.4 volts AC! But I could not hear any of it it. I switched the meter over to Frequency Counter and it read 38.4Kz. I also saw that this was only on the left channel. Just like the story above this problem had followed a change I had made. I had rerouted a wire to a voltage amplifier control grid to mod the amp for a different driver and in doing so I had routed this wire right over the top of a wire headed to the screen grid of a nearby left chanel final. This being an ultralinear amp, the screen goes to a 43% tap (in this case) on the OPT. The insulation of both wires was fine but there was enough capacitive coupling to get oscillation. The fix was merely separating these two as much as practical. The oscillation was gone has not come back. I removed the previously added supplementary cap and bass has fine ever since.
Hope this helps.
Hello trobbins and thanks for the question. The tweeter had gotten hot to the touch and that is what finally put me on the trail. I'll bet it wasn't but hum but tweeter heating, as I think about it, that prompted me to put a voltmeter across the speaker terminals. It measured about 150 degrees F on it's frame. That happened a few years ago and no harm seems to have been done. I may just swap it out with a rebuilt tweeter on principle.
Would anybody be so kind as to list grid stoppers that seem OK. The sort of values that should get one started. How long is a piece of string comes to mind. My interest would be All ECC, All typical EL and beam power types , EF 86/184. I read somewhere this resistor could do all sorts of things. To be honest it seems to have to be very large to measure much. I usualy use 1 K and see where it goes. I never seem to have come unstuck when I did. Nice not to if a phono stage. When an input transformer to a power valve suddenly the world changes. I think in asking I know that answer. Good thread, thanks.
If it's not too much of a diversion . Could a NPN transistor be used to the UL tap of a SE EL 84 amp? The idea as a unity gain buffer. I suspect this would not work. Interesting if it did as it might get some more power . As UL can be trouble I suspect this is very much on topic. The ideal NPN might be a problem.
If it's not too much of a diversion . Could a NPN transistor be used to the UL tap of a SE EL 84 amp? The idea as a unity gain buffer. I suspect this would not work. Interesting if it did as it might get some more power . As UL can be trouble I suspect this is very much on topic. The ideal NPN might be a problem.
When my tube SET idles and I touch the output tube with a coax cable with 2 inches stripped, I get a dirty sinewave of 500kHz and 10-15mV amplitude. The dirtyness cannot be seen due to my low bandwidth, but its amplitude is approx. 2mV. The driver tube is EL802 with 1k CC grid stopper and the power tube is 6P45S with no grid stopper. The coupling is by IT.
This is interesting. Was she watching digital or analog TV ?
If it was digital TV I could imagine that a unmodulated UHF oscillation would cause the pixelation.
It it was analog TV, I can't see why it would pixelate. One would expect it to decrease the SNR and/or add "snow" to the image.
Slightly OT, but there is another reason for using high G1 stopper resistors, up to 15K if circuit (nfb) stability allows. That is to stop 'blocking' (charging coupling capacitors) during overload. The input (Miller) capacitance with pentodes is usually not high enough to cause trouble, and such values may just be sufficient to prevent blocking during brief overloads or decreasing it to unnoticable level.
Makes sense, because fb is such magic bullet to fix the distortion and freq response
Varying phase shift across frequency (opt + stages), can make nice oscilator.
Too much fb will make new, higher order distortion; sucking the details.
Do not use fb (too much), do not buy crap gear.
It will be no problem with high amount of feedback if the amp or the preamp can afford it , I mean that if the amp or preamp shows low phase shift ( in high frequency of course) , from it's input to it's output , then can afford " too much " feedback . But on the other hand too much feedback can "destroy " the output impedance of the stage that placed in .
