The heavy read...For those who have access to the Radiotron Hd Bk 4th ed, go to page 356 (sect 3) The section regarding stability, phase shift etc. p. 371 corrective networks.
I have a different view/bench method in setting E&I cored push pull power amp stages using varying levels of global feedback, that I use a low freq sweep generator as a wobbulator to optimize the circuit Q factor. It requires a multi-function signal generator with also one shot functions to check for loop transient response. This a near duplicate from the method checking loop stability and response from an divide n counter/ phase locked loop. It requires experience but the solution is a more predictable end result than using PSpice tool, which doesn´t tell the whole truth and cannot predict the varying parameters of basically a "swinging output transformer" with non constant permeability, varying with excitation currents. The advantge of this is a live test rather than a simulated one which could be way out.....
Bench Baron
I have a different view/bench method in setting E&I cored push pull power amp stages using varying levels of global feedback, that I use a low freq sweep generator as a wobbulator to optimize the circuit Q factor. It requires a multi-function signal generator with also one shot functions to check for loop transient response. This a near duplicate from the method checking loop stability and response from an divide n counter/ phase locked loop. It requires experience but the solution is a more predictable end result than using PSpice tool, which doesn´t tell the whole truth and cannot predict the varying parameters of basically a "swinging output transformer" with non constant permeability, varying with excitation currents. The advantge of this is a live test rather than a simulated one which could be way out.....
Bench Baron
Thank You. Due to all You recommendations, I have now with 20db nfb LF steady stable.
HF is other story. This is what I get at 1 kz, no compensation applied.
As I saw there are two compensation points in this amp: one snubber around the first plate, the other one over the series nfb resistor. What is a good point to start please? Thank You!
HF is other story. This is what I get at 1 kz, no compensation applied.
As I saw there are two compensation points in this amp: one snubber around the first plate, the other one over the series nfb resistor. What is a good point to start please? Thank You!
The photo in #62 indicates 5 ringing cycles in 50us div, so circa 100kHz resonance that could cause instability if not managed. That resonant frequency could change for different loadings, such as no load or just capacitance loading.
Plot in #63 shows a much lower resonance frequency. Did you change the loading or anything else, other than adding a 1k5-250pF step network across a 470k plate resistor? Can you also link to the schematic you are now using, as I recall there were quite a few differences to a Williamson.
You have chosen to use quite a marked step from 470k to 1k5. Was there any reason for that, and did you calculate the likely start and stop frequencies of that step network?
You may want to add further squarewave testing (for a known applied compensation scheme) for no load connected, and for capacitor only connected loading - as an initial check as to how unstable the amp may get. But be mindful that such testing could cause immediate gross oscillation without any signal being applied, in which case turn-off the amp. The hoped for outcome from that type of testing could be a very ringing response for an applied squarewave, but if it doesn't cause gross oscillation then that shows a good outcome imho.
Plot in #63 shows a much lower resonance frequency. Did you change the loading or anything else, other than adding a 1k5-250pF step network across a 470k plate resistor? Can you also link to the schematic you are now using, as I recall there were quite a few differences to a Williamson.
You have chosen to use quite a marked step from 470k to 1k5. Was there any reason for that, and did you calculate the likely start and stop frequencies of that step network?
You may want to add further squarewave testing (for a known applied compensation scheme) for no load connected, and for capacitor only connected loading - as an initial check as to how unstable the amp may get. But be mindful that such testing could cause immediate gross oscillation without any signal being applied, in which case turn-off the amp. The hoped for outcome from that type of testing could be a very ringing response for an applied squarewave, but if it doesn't cause gross oscillation then that shows a good outcome imho.
Hey. No changes other than mentioned. #62= no compensation at all. #63= just 250pF+1.5K over first plate.I.ll draw the actual schematic and post it. Thanks.
It seems the plate compensation I cannot manage properly so I completely get rid of. I just keeped the 230 pF over 15k nfb series resistor with the following results:
1kHz:
10kHz:
It is not perfect but most usable than everything I tried. I.ll continue tomorrow I have not in hand all small capacitors values I need to handle in. Thank You.
1kHz:
10kHz:
It is not perfect but most usable than everything I tried. I.ll continue tomorrow I have not in hand all small capacitors values I need to handle in. Thank You.
Last edited:
Post #15 indicates triode mode operation of output stage - but schematic in post #72 shows pentode mode ? Also the schematic is ambiguous as to the B+ node, as it doesn't connect to the top rail feed. Pentode mode operation would have a significant effect on managing HF stability, and you likely have to dumb down the treble response a lot (compared to a Williamson) and even then may see marginal stability with no load or capacitor only loading depending on the output transformer.
In post #37 you identify the use of independent cathode bias of driver stage, and schematic shows a significantly low 15k anode loading. Did you measure significant anode voltage difference with the 6N6P valve and that led to changing to separate cathode biasing? Did you tube roll as a way of checking whether you could stay with original common cathode biasing? Why did you lower the anode loading to 15k - as a pro-rata to driving three parallel output stage valves?
Your input stage doesn't bypass the 1k5 bias - did you realise what that may do to triode operation?
In post #37 you identify the use of independent cathode bias of driver stage, and schematic shows a significantly low 15k anode loading. Did you measure significant anode voltage difference with the 6N6P valve and that led to changing to separate cathode biasing? Did you tube roll as a way of checking whether you could stay with original common cathode biasing? Why did you lower the anode loading to 15k - as a pro-rata to driving three parallel output stage valves?
Your input stage doesn't bypass the 1k5 bias - did you realise what that may do to triode operation?
Hello. Those project triode operated was gone. I had not such of stability issues on it as I used only 10db of feedback, and in the end I didn't used at all. This one is a pentode one, a new one as I mentioned in post #42. It need a lot of nfb to proper drive the speaker I have. Is a MI project for bass application. I used the experience I get from the last but used a 6sl7 in preamp to match with preamp section I have and the nfb amount I wish. Of course 6n6p is huge and not justified to drive in voltage the power tubes but works flawless and don't bother with. My only concern was to get it stable, and finally with You help it is. 130W clean output onset clipping steady stable with 20 db nfb applied. The bass is tight, precise and even for mild clipping it did not "fart". Otherwise I.ll be tempted to reduce the couplings in power stage for more fast time recovery, but was not the case. Thanks.
6n6p is a low mu tube with a internal resistance of 2k, using a 15 k plate looks a good choice as much browsed around the net I get reports it sound better with 12 - 15 mA.More think was convenient to get the full swing into 3x100k (33k) - 3x 220k (73k) load. (I did it able to drive anything from el34 to kt88). And yes, I much preferred to use local feedback around the stages from subjective sound consideration. I built this amp with the ears on the speaker to tailor the response I want by try and error. Maybe is not optimized from a design point but works good and sound good. Do you catch some horrible mistakes please ? Thank you.
6n6p is a low mu tube with a internal resistance of 2k, using a 15 k plate looks a good choice as much browsed around the net I get reports it sound better with 12 - 15 mA.More think was convenient to get the full swing into 3x100k (33k) - 3x 220k (73k) load. (I did it able to drive anything from el34 to kt88). And yes, I much preferred to use local feedback around the stages from subjective sound consideration. I built this amp with the ears on the speaker to tailor the response I want by try and error. Maybe is not optimized from a design point but works good and sound good. Do you catch some horrible mistakes please ? Thank you.
Last edited:
I assume you have a hefty dummy load, like an aluminum-cased 100 watt 8 ohm resistor. The "tough" tests Tim mentions would be:
--No load, or disconnect the dummy load and look for "freak out" of square wave;
--Resistive/capacitive load, or place a .1uF cap in parallel with the load and see if the square wave collapses completely;
--Capacitor only load. A very tough test. .01uF to start, then increase. Usually you won't get far, but even a small capacitor-only load, if stable, is a good sign.
--No load, or disconnect the dummy load and look for "freak out" of square wave;
--Resistive/capacitive load, or place a .1uF cap in parallel with the load and see if the square wave collapses completely;
--Capacitor only load. A very tough test. .01uF to start, then increase. Usually you won't get far, but even a small capacitor-only load, if stable, is a good sign.
