Just completed my latest build, kind of a mutt design from the Mullard 5-20 with variations taken from other builders on DIYAudio. So far so good, no really big problems to contend with. AC heated, absolutely immeasureable hum at the output. Contrary to most of my other builds, this one has no RF issues (yea !!)
I have a few observations, and a few questions was hoping you could help out with.
Feedback is about 20dB (correct me if my math is wrong, open loop gain measured at about 105). This is a reasonable amount of feedback when using A470 iron, not too much, but enough for a decent damping factor ???
Compensation network (R11-C9) seems a little heavy, but that's what works for clean 1kHz squares and slightly rounded 10kHz squares. These values are also a compromise between triode and UL operation, with triode looking the best (preferred position anyway). Yes, or is something else wrong?
Finding a little bit of motorboating at the output. The power supply has a tendency to 'burp' or 'bounce' every now and then, which is pretty much unavoidable without resorting to really large storage capacitance, which I prefer to avoid. No more room for motor runs. Each time the PS bounces, I see the output react as well, with approximate 1 Hz decaying exponential. The magnitude is fairly low, with normal bouncing at the 4 ohm output around 1 mV, maximum excursions maybe up to 5 mV. I haven't really been bothered by this, but wondering if I should be. Is the preferred solution to decrease the coupling capacitors C5-C8, or is the problem strictly PS related? Grounding is immaculate, feedback reference taken directly to the first stage common, star grounding throughout, etc.
Suggestions welcome, just no major modifications, please. Amp is built and functional. Thanks !!
I have a few observations, and a few questions was hoping you could help out with.
Feedback is about 20dB (correct me if my math is wrong, open loop gain measured at about 105). This is a reasonable amount of feedback when using A470 iron, not too much, but enough for a decent damping factor ???
Compensation network (R11-C9) seems a little heavy, but that's what works for clean 1kHz squares and slightly rounded 10kHz squares. These values are also a compromise between triode and UL operation, with triode looking the best (preferred position anyway). Yes, or is something else wrong?
Finding a little bit of motorboating at the output. The power supply has a tendency to 'burp' or 'bounce' every now and then, which is pretty much unavoidable without resorting to really large storage capacitance, which I prefer to avoid. No more room for motor runs. Each time the PS bounces, I see the output react as well, with approximate 1 Hz decaying exponential. The magnitude is fairly low, with normal bouncing at the 4 ohm output around 1 mV, maximum excursions maybe up to 5 mV. I haven't really been bothered by this, but wondering if I should be. Is the preferred solution to decrease the coupling capacitors C5-C8, or is the problem strictly PS related? Grounding is immaculate, feedback reference taken directly to the first stage common, star grounding throughout, etc.
Suggestions welcome, just no major modifications, please. Amp is built and functional. Thanks !!
A quick check of the various time constants does'nt show any real problems.
The Time constant of the coupling caps into the output tube grids (0.22uF & 121K + bias network resistance) is perhaps a little close to the time constant of the power supply feed to the diff amp BUT the current source tail should prevent any major problem here (C4 and C6 should perhaps be X2 or X3 of current value, to increase that time constant, or leave them and add caps from the bottom of each 121K bias resistor to 0V to reduce that time constant).
Did you see the problem in UL mode or triode mode or both?
The screen resistors (100R) are too low for UL operation. If they are going to cause a problem it normally shows up as "squegging". If you look at say a 1kHz sine wave into a dummy load at reasonably high output level squegging can be observed as a burst of HF oscillation just after the peak on one side or other (or both). If that occurs you can get a serious current demand surge from the Power Supply. If there is no evidence of that then I can think of anything else which may cause a problem. If there is evidence of that, increase the screen resistors until it goes away.Your diagrams don't seem to show how you connect your output tube bias.
Cheers,
Ian
The Time constant of the coupling caps into the output tube grids (0.22uF & 121K + bias network resistance) is perhaps a little close to the time constant of the power supply feed to the diff amp BUT the current source tail should prevent any major problem here (C4 and C6 should perhaps be X2 or X3 of current value, to increase that time constant, or leave them and add caps from the bottom of each 121K bias resistor to 0V to reduce that time constant).
Did you see the problem in UL mode or triode mode or both?
The screen resistors (100R) are too low for UL operation. If they are going to cause a problem it normally shows up as "squegging". If you look at say a 1kHz sine wave into a dummy load at reasonably high output level squegging can be observed as a burst of HF oscillation just after the peak on one side or other (or both). If that occurs you can get a serious current demand surge from the Power Supply. If there is no evidence of that then I can think of anything else which may cause a problem. If there is evidence of that, increase the screen resistors until it goes away.Your diagrams don't seem to show how you connect your output tube bias.
Cheers,
Ian
zigzagflux,
Even though you don't want larger caps in the power supply, I think that supplementing the four 15uF caps with larger caps will stop the motorboating. You may try looking for some photoflash caps. I've had very good luck with them over the years, and they sound pretty good to me. They are very space efficient. I have them in 210uF 330V, 350uF 330V, 600uF 330V and 800uF 330V. Obviously, you'd have to series connect a pair for your rail voltages. The 210uF 330V are about 1" in diameter and 1" long.
Ditto on gingertube's notation about 100R being too small for screen grid connection in UL. Made the same mistake with KT88s. worked fine in triode, but standing current went through the roof in UL in addition to hearing wild squealing through the speakers when turning the power off.
820R 3W CF resistors fixed it for me.
Stuart
Even though you don't want larger caps in the power supply, I think that supplementing the four 15uF caps with larger caps will stop the motorboating. You may try looking for some photoflash caps. I've had very good luck with them over the years, and they sound pretty good to me. They are very space efficient. I have them in 210uF 330V, 350uF 330V, 600uF 330V and 800uF 330V. Obviously, you'd have to series connect a pair for your rail voltages. The 210uF 330V are about 1" in diameter and 1" long.
Ditto on gingertube's notation about 100R being too small for screen grid connection in UL. Made the same mistake with KT88s. worked fine in triode, but standing current went through the roof in UL in addition to hearing wild squealing through the speakers when turning the power off.
820R 3W CF resistors fixed it for me.
Stuart
Curious operation. Added double the capacitance to the driver stage, input stage, and both stages. No real change at the output, maybe a little, but not significant enough that I would pursue 100uF at these stages. Seems to be primarily generated at the main 420V B+. Added up to 70uF of additional capacitance to the output stage; again, maybe a little bit of improvement, but the bounce is still there. Being that I am running at about 200mV of ripple at the B+, I can't see more and more capacitance as helping the problem, IDK.
For kicks, I used a 120VAC ferroresonant regulator to feed the amp. Problem was essentially gone, though there was just a slight bit of wiggle when the house well pump would start; so there is a little getting through the regulator, which indicates to me this is a PS issue.
Bias supply is rock solid, no bounce there. But I do see the bounce at the EL34 grid. No bounce at the EL34 cathodes. No bounce at the LTP grids, just a slow time constant rise and fall with no ripple. LTP cathode is fairly quiet. Hope I have all of this correct; I'm running on memory here.
Well, adding additional 15uF to each stage did not seem to change things, and adding 30uF to just the driver stage also did not improve things. The input stage is quite quiet regardless of capacitance; most of the bounce occurs at the output and driver stages. I'll try to muster up enough capacitance at the driver stage to completely eliminate PS bounce there; my thought is if the speaker output bounce still exists, this is definitely due to output stage issues.
Any other thoughts?
For kicks, I used a 120VAC ferroresonant regulator to feed the amp. Problem was essentially gone, though there was just a slight bit of wiggle when the house well pump would start; so there is a little getting through the regulator, which indicates to me this is a PS issue.
Bias supply is rock solid, no bounce there. But I do see the bounce at the EL34 grid. No bounce at the EL34 cathodes. No bounce at the LTP grids, just a slow time constant rise and fall with no ripple. LTP cathode is fairly quiet. Hope I have all of this correct; I'm running on memory here.
The bias supply is very stiff relative to the bounce. Is that good enough?
Both, although all my testing is occurring in triode mode. At the end of the day, I suspect the amp will never leave triode mode; the switch is there for little reason.
Page 5 shows the bias supply, with the bias adjustments.
Well, adding additional 15uF to each stage did not seem to change things, and adding 30uF to just the driver stage also did not improve things. The input stage is quite quiet regardless of capacitance; most of the bounce occurs at the output and driver stages. I'll try to muster up enough capacitance at the driver stage to completely eliminate PS bounce there; my thought is if the speaker output bounce still exists, this is definitely due to output stage issues.
Any other thoughts?
Good call sepolansky:
added 100uF to both driver and input stages, bounce is gone. Basically I tried to push the time constant of the driver PS to .5 Hz, and the input stage to 0.1 Hz. Since I had good star wiring with the existing polypropylene caps, I tossed in 1 ohm carbon resistors in series with big cheap 100uF electrolytics to avoid any strange resonances. Does just fine in killing the bounce.
Very happy with the result; I'll keep this design for possible future builds. I have a few other amps to compare this one to (including a Karna), and I might just prefer this one to all the others. Probably due in part to the lower damping from the GNFB, bass is a little punchier. Highs are also significantly more detailed than the Karna, which I attribute to poor output transformers (some day I'll sell some stock and get amorphous Lundahl).
However, the sound is a little more compressed than the non-FB amp, so there is a tradeoff of punch and detailed highs versus superior midrange. Tradeoffs.
The fun continues...
added 100uF to both driver and input stages, bounce is gone. Basically I tried to push the time constant of the driver PS to .5 Hz, and the input stage to 0.1 Hz. Since I had good star wiring with the existing polypropylene caps, I tossed in 1 ohm carbon resistors in series with big cheap 100uF electrolytics to avoid any strange resonances. Does just fine in killing the bounce.
Very happy with the result; I'll keep this design for possible future builds. I have a few other amps to compare this one to (including a Karna), and I might just prefer this one to all the others. Probably due in part to the lower damping from the GNFB, bass is a little punchier. Highs are also significantly more detailed than the Karna, which I attribute to poor output transformers (some day I'll sell some stock and get amorphous Lundahl).
However, the sound is a little more compressed than the non-FB amp, so there is a tradeoff of punch and detailed highs versus superior midrange. Tradeoffs.
The fun continues...
Excellent zigzagflux!
It's always such a relief when a persistent problem is resolved. Then you can finally sit back and enjoy the fruits of your labor.
My mentor pointed out that in feedback amps, each output tube and circuit type seems to have its own sweet spot with respect to feedback. He suggested installing a pot in the feedback loop and setting the pot to where the circuit sounds right to you. Incredibly small changes can be noticeable.
For the KT88 (ala Citation II) I used, about 20dB was the starting point. The actual value we used was pretty close, I forget it offhand. But once we found the setting, replaced the pot with a Mills resistor, then final trimmed the feedback cap for a slight peak on the 10kHz square wave leading edge. Adding capacitance to remove the overshoot really takes away the "air".
If you haven't messed with those feedback settings yet, you may want to diddle around a bit there. It could really wake up a feedback amp.
Good luck,
Stuart
It's always such a relief when a persistent problem is resolved. Then you can finally sit back and enjoy the fruits of your labor.
My mentor pointed out that in feedback amps, each output tube and circuit type seems to have its own sweet spot with respect to feedback. He suggested installing a pot in the feedback loop and setting the pot to where the circuit sounds right to you. Incredibly small changes can be noticeable.
For the KT88 (ala Citation II) I used, about 20dB was the starting point. The actual value we used was pretty close, I forget it offhand. But once we found the setting, replaced the pot with a Mills resistor, then final trimmed the feedback cap for a slight peak on the 10kHz square wave leading edge. Adding capacitance to remove the overshoot really takes away the "air".
If you haven't messed with those feedback settings yet, you may want to diddle around a bit there. It could really wake up a feedback amp.
Good luck,
Stuart
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.