For a couple of days I have been pulling my hair about this motorboating issue (I call it motorboating in the lose sense, see below)
It is for an ampifier with an ECC83 directly coupled to an ECC88 LTP, then RC-coupled to EL84 finals in triode mode. No CCS anywhere, just plain old resistors. Grid stoppers of appropriate values everywhere. The aim is to have about 20db of nefgative feedback around the whole circuit. A medium-quality transformer is used. Ir reverses phase at about 800Khz.
Step networks have been employed on the ECC83 anode resistor and leaad comp on the feedback resistor.
The circuit works as expecetd when open-loop. From the start I noticed some LG instability. It was slowly moving about at 2Hz but not a rail-to-rail (at we'd say in SS) oscillation The big issue started while I was optimising some circuit values. At some point I lowered the value of the decoupling resistor from the RC that feeds the 1st stage. There is a 220uF from the bridge, a 6K8 and a 22uf. The ECC88 feed from that. Then there is a 56K and another 22uf to decouple for 1 ECC83 anode.
So I replaced the 6K8 resistor wioth 4K9, and hell broke loose. It started stable, but when excited by the SG it violently motorboated at 5HZ. This continued even after the input signal was removed. Everything I did made no difference whatsoever.
-added more capacitance to the 2nd stage
-reduced the coupling caps from 68n to 22n (cutoff waas 7Hz, how it is more like 18Hz)
-increased EL84 grid stoppers to 8.6 Kohm
-re checked and changed the anode compensation of 1st stage
I now have reduced to feedback to something like 15db, and it is still marginal with a 10 KOhm resistor. I have noticed that when i change that resistor, The lower I go (4.7 Kohm the higher the motorboating frequency is (12 Hz wih 4K7)
The only solution I can think about this is to forgo the decoupling for the ECC83, but this will add possible hum.
Any good guesses greatly appreciated!
It is for an ampifier with an ECC83 directly coupled to an ECC88 LTP, then RC-coupled to EL84 finals in triode mode. No CCS anywhere, just plain old resistors. Grid stoppers of appropriate values everywhere. The aim is to have about 20db of nefgative feedback around the whole circuit. A medium-quality transformer is used. Ir reverses phase at about 800Khz.
Step networks have been employed on the ECC83 anode resistor and leaad comp on the feedback resistor.
The circuit works as expecetd when open-loop. From the start I noticed some LG instability. It was slowly moving about at 2Hz but not a rail-to-rail (at we'd say in SS) oscillation The big issue started while I was optimising some circuit values. At some point I lowered the value of the decoupling resistor from the RC that feeds the 1st stage. There is a 220uF from the bridge, a 6K8 and a 22uf. The ECC88 feed from that. Then there is a 56K and another 22uf to decouple for 1 ECC83 anode.
So I replaced the 6K8 resistor wioth 4K9, and hell broke loose. It started stable, but when excited by the SG it violently motorboated at 5HZ. This continued even after the input signal was removed. Everything I did made no difference whatsoever.
-added more capacitance to the 2nd stage
-reduced the coupling caps from 68n to 22n (cutoff waas 7Hz, how it is more like 18Hz)
-increased EL84 grid stoppers to 8.6 Kohm
-re checked and changed the anode compensation of 1st stage
I now have reduced to feedback to something like 15db, and it is still marginal with a 10 KOhm resistor. I have noticed that when i change that resistor, The lower I go (4.7 Kohm the higher the motorboating frequency is (12 Hz wih 4K7)
The only solution I can think about this is to forgo the decoupling for the ECC83, but this will add possible hum.
Any good guesses greatly appreciated!
Without a schemo it's harder to tell, but it looks like there are several problems here.
OPTs are the biggest culprits when it comes to stabilizing hollow state amps. 20dbv is a lot of NFB, and it puts a demand on the OPT at both ends of the audio spectrum. Excessive phase shift at either end might make it impossible to include 20dbv of gNFB and keep it stable. (Never designed in so much myself, as 20db sounded pretty bad.)
You need to fix this first. Your open loop design needs to be unconditionally stable before you connect the gNFB. If it isn't, you're asking for big problems once you close the NFB loop.
Without a schemo, it's not possible to really figure this out. If there's low frequency instability, then there's some source of phase shift. At the low frequency end, that includes any decoupling networks, RC coupling between stages, cathode bypass capacitors. All these should have their time constants staggered in order to prevent a phase shift pile-up that can turn negative feedback positive before the open loop gain drops below unity. Poorly designed OPTs can likewise have some squirrelly phase misbehaviors at both high and low frequencies. Sometimes, high frequency misbehavior can manifest itself as motorboating. This problem is often treated by connecting a Zobel between the ends of the primary and the center tap.
OPTs are the biggest culprits when it comes to stabilizing hollow state amps. 20dbv is a lot of NFB, and it puts a demand on the OPT at both ends of the audio spectrum. Excessive phase shift at either end might make it impossible to include 20dbv of gNFB and keep it stable. (Never designed in so much myself, as 20db sounded pretty bad.)
You need to fix this first. Your open loop design needs to be unconditionally stable before you connect the gNFB. If it isn't, you're asking for big problems once you close the NFB loop.
Without a schemo, it's not possible to really figure this out. If there's low frequency instability, then there's some source of phase shift. At the low frequency end, that includes any decoupling networks, RC coupling between stages, cathode bypass capacitors. All these should have their time constants staggered in order to prevent a phase shift pile-up that can turn negative feedback positive before the open loop gain drops below unity. Poorly designed OPTs can likewise have some squirrelly phase misbehaviors at both high and low frequencies. Sometimes, high frequency misbehavior can manifest itself as motorboating. This problem is often treated by connecting a Zobel between the ends of the primary and the center tap.
well I meant with feedback, my bad. Withouf freedback its just fine.
No cathode bypass caps except the output stage, come to think of it lowering the the value makes the motorboating go up in frequency. I have 220 uF at the moment.
Will try with the Zobel, didnt think of that , thanks. As for the schemo, it is still only in my head, I ll draw one up tomorrow, its really late here.
Thanks for your time!
Yes, this will usually happen in tube circuits unless there is a single dominant LF time constant, several times higher in frequency than any others.
It's best to set the first capacitor in the signal path as the dominant time constant, so that LF fluctuations are not amplified any more than necessary.
Ray
There are coupling caps from the LTP to the finals, 68n and 470KOhm, cutoff at 7Hz,( now I have decresased them to 22n) which are supposed to be the dominant LF pole. This is the part that baffles me the most.
I should also mention that the stage stage that is limiting is the 1st stage ECC83, but then again its the stage with the lowest margin, so it makes sense.
1) From your description sounds like its a variation of Mullard 520. Anyway, schematic is a mandatory for this type of questions.
2) What kind of load you use to diagnose motorboating? 8 Ohm wirewound resistor or open secondary connected to oscilloscope? Try to decrease NFB (use trim pot) until motorboating go away (with 8 Ohm load connected).
3) Try to shunt coupling caps leading to output tubes with 4.7 - 6.8 Mohm resistors, and rea-adjust bias. This hack will remove one LF pole.
4) What is value of grid leak resistor of the first input tube? Sometimes too high value leads to stability problem.
2) What kind of load you use to diagnose motorboating? 8 Ohm wirewound resistor or open secondary connected to oscilloscope? Try to decrease NFB (use trim pot) until motorboating go away (with 8 Ohm load connected).
3) Try to shunt coupling caps leading to output tubes with 4.7 - 6.8 Mohm resistors, and rea-adjust bias. This hack will remove one LF pole.
4) What is value of grid leak resistor of the first input tube? Sometimes too high value leads to stability problem.
Work out and write down your LF poles, including the OPT and any shelf caused by cathode decouplers. Make sure the highest one is well clear of the next lowest one or two. 'Well clear' depends on loop gain - the more feedback the clearer you need to be. If you have two poles near each other then add them. So if you have 10Hz, 2Hz and 1.5Hz then the dominant ratio is not 5 (=10/2) but 3 (=10/(2+1.5)). This ratio needs to approach the feedback ratio, although it need not necessarily exceed it.
Separately, work out how much LF coupling is going via the supply rail and at what frequency. Positive feedback there will muck up your loop stability calculations.
Separately, work out how much LF coupling is going via the supply rail and at what frequency. Positive feedback there will muck up your loop stability calculations.
Well, yes, it is Mullard 5-20 inspired.Load is 11 Ohm wirewound resistors (had to go to this value to adjust for the 9K anode to anode I desitre). Grid leak was a little high (and it is an air-wire situation) So I reduced it it 82K few days ago.
I like this hack idea, will try it later.
This is insightful, I have worked out all the LF poles, as mentioned on the first post, but did not include the OPT. Actually it is spot-on because When I removed the cathode bypasses, and it was totally stable, but at the time i thought it was just because of the reduced overall gain. Thanks!
Edit: How do I drive the transformer to full power from a low-impedance source to check for bandwdith? I am thinking treminating it with an appropriate resistor to make it look like an 8 Ohm load and using an SS amplifier. Come to think of it, the DCR of the secondary will be enough to make the primary it look like a speaker.
Last edited:
with the tubes removed except for the output tube. does it still motor boat? if the motor-boating went away, then your pulling too much current for your supply.
please post your schematics or ones your building off of. There is no such things as new tube circuits.
even those ccs circuits were done in the 60's.
please post your schematics or ones your building off of. There is no such things as new tube circuits.
even those ccs circuits were done in the 60's.
Tough to tell from the wording (I like schematics). If this is so, then it's a problem- the input tube and cathodyne should be fed from the same HV point.
Well, thank you all for your attention and brainpower!
As DF96 pointed out, I had forgotten to include the LF cutoff of the OPT in my calculations. Turns out that it is 8.4 Henry per side (small signal, 120 Hz), which paralleled wit the approximate 1k Ra of EL84 (triode, datasheet value) gives a LF pole of 25 Hz. So i changed the coupling cpacitors to 4.7 nF (closest value I had) of giving a cutoff of 72Hz. It's all good now.
Thanks to all
As DF96 pointed out, I had forgotten to include the LF cutoff of the OPT in my calculations. Turns out that it is 8.4 Henry per side (small signal, 120 Hz), which paralleled wit the approximate 1k Ra of EL84 (triode, datasheet value) gives a LF pole of 25 Hz. So i changed the coupling cpacitors to 4.7 nF (closest value I had) of giving a cutoff of 72Hz. It's all good now.
Thanks to all
You may need to be careful of LF distortion as the phase splitter has to drive the output stage via a high pass filter. I would be tempted to try putting the OPT as the dominant LF pole, but then limiting the LF range at the input too (i.e. outside the loop) so it is not asked to attempt things it cannot do.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.