Well, if there was one oin there.....
My fault, actually, for not showing a schematic for ppl to look at.
The OPT can be the dominant pole in the feedback loop, but not necessarily the dominant pole in the whole system.
If the OPT rolls off at 25Hz then any attempt to make it go much lower than that is likely to lead to distortion. With reasonably heavy feedback the loop will try to force it lower, even with a 70Hz roll off from a coupling cap in the loop. I would try a bigger coupling cap within the loop (say 2-5Hz rolloff) but then a 20Hz rolloff before the loop. 20dB feedback means the loop will try to amplify down to 2.5Hz so you just have to stop subsonic stuff from ever reaching the loop.
If the OPT rolls off at 25Hz then any attempt to make it go much lower than that is likely to lead to distortion. With reasonably heavy feedback the loop will try to force it lower, even with a 70Hz roll off from a coupling cap in the loop. I would try a bigger coupling cap within the loop (say 2-5Hz rolloff) but then a 20Hz rolloff before the loop. 20dB feedback means the loop will try to amplify down to 2.5Hz so you just have to stop subsonic stuff from ever reaching the loop.
OK, so I backed off the feedback (from 27 db to 16 db) and I decreased the 470uf cathode bypasses of EL84 to 47uF (the 470uF was casing LF instability too, when turning on/off the signal gen). Now it is very steady.
Next step is to decrease the cathode resistors of the finals and to go pentode mode. This will be interesting .
Next step is to decrease the cathode resistors of the finals and to go pentode mode. This will be interesting .
I changed the gain structure to be more fitting for the ECC83 and the ECC88, and after fine tuning the compensation networks in pentode i can have up to 30 db global feedback without instability.
I set it for 25 db feedback and I get the following results
input sensitivity .24 V rms
LF goes down to 30Hz @ full power -3db
15 Hz @ 1W -3db
-3db @110Khz full power
-3db @ 95 Khz 1 Watt
and all that with a cheap guitar output transformer
I am starting to like it ...
I set it for 25 db feedback and I get the following results
input sensitivity .24 V rms
LF goes down to 30Hz @ full power -3db
15 Hz @ 1W -3db
-3db @110Khz full power
-3db @ 95 Khz 1 Watt
and all that with a cheap guitar output transformer
I am starting to like it ...
Last edited:
Here is the schematic.The gain without feedback is around 560. This should be very good for EL34's + feedback, but too much for the EL84.
As I am re-making an EL84 amplifier I bought some time ago, there are a few differences, the outputs are ultralinear here. The PS decoupling is shared between channels.
As it is it has way too much gain so feedback to give input sensitivity of about 0,3 V rms has to be about 30 db. This is way more then the output transformers will take. On the load resistor , it is stable, but it breaks into motorboating on the loudspeakers. Seems like the "guitar" transformer I used when testing was better quality one.
Next step it to replace the 1st stage with a ECC82 and adjust the circuit for that,
There isn't a ground on the output transformer secondary. Also, are those 1uF coupling caps?
Ok, I made the change to a ECC81 with 47Kohm load and the feedback now is about 16db, which is much more manageable. Coupling caps are now 180nf, As Predicted, there is distortion st very low frequencies.
Questions :
- There is a 100K pot at the input, and after grid leak of 470K for the ECC81. I am planning to change the grid leak to 220KOhm and solder a 38nF between pot wiper and grid, For 220KOhm and 39nF I get a corner freq of 18,55Hz as DF86 recommended in #19. The question is, do I have to include the 100K of the pot in the calculation? I think not, but I d like to verify
- Because of the lower gain, feedback resistor is now at 180K. Do I leave it or should i better make a voltage divider with the input stage cathode resistor? I am concerned about noise pickup, that case is tiny...
Thank you
Questions :
- There is a 100K pot at the input, and after grid leak of 470K for the ECC81. I am planning to change the grid leak to 220KOhm and solder a 38nF between pot wiper and grid, For 220KOhm and 39nF I get a corner freq of 18,55Hz as DF86 recommended in #19. The question is, do I have to include the 100K of the pot in the calculation? I think not, but I d like to verify
- Because of the lower gain, feedback resistor is now at 180K. Do I leave it or should i better make a voltage divider with the input stage cathode resistor? I am concerned about noise pickup, that case is tiny...
Thank you
A 100k pot feeding 39nF coupling cap to a 220k grid leak will give corner freq from 18.55Hz to 16.66Hz - the source impedance will vary from 0 to 25k (pot at -6dB) so the effective resistance will be 220k-245k. Whether this change is audible may depend on your speakers and room.costis_n said:
Then it will be doing something quite different - and much correction will be required when the input finds out
If you want an amplifier output to be a good approximation to a voltage source then the feedback has to come from the output point. If the OPT is outside the loop then maybe the OPT should be moved from the amp chassis to the speaker box? Treat the OPT as part of the speaker instead of part of the amp.Globulator said:
Like having an electrostatic speaker with its transformer. Whatever. I m in the final stage of tweaking, will post the finished schematic to see if you can spot anything terrible.
Nonetheless, I am actually interested in applying that. Based on the schematic I posted, how would I go about applying that (feedback from both anodes possibly) ?
I once saw a British 50's vintage amplifier, not from the famous ones, that had 5 or 6 feedback loops. It was with KT66's ? I think? I can't remember, does it ring a bell to anybody?
I once saw a British 50's vintage amplifier, not from the famous ones, that had 5 or 6 feedback loops. It was with KT66's ? I think? I can't remember, does it ring a bell to anybody?
Yes - very good question! At this point most people go back to the easy single-rail secondary
Fact#1: It's MUCH easier to leave the OPT out of the loop in an SE (Single Ended) design.
Push-pull has issues because you have a (sort of!) balanced drive so you actually need to treat both the 'push' and the 'pull' bit as separate amplifiers and run the feedback from the output tube anodes to their drivers.
This requires decent components that can cope with the odd 800V difference across them.
On my SE I had the driver tube with a 1k cathode resistor. From the output tube anode via a 10k resistor, via a 22uF Poly I fed back to the driver cathode. This did mean I could not bypass the cathode with a capacitor. So it's not as easy but I made the driver+output pentode form a tightly coupled lowish gain low-impedance driver which really made the OPT sound far better than it ever had before.
Transformers seem to get along just fine if they are driven properly - without feedback of some form (lowZ triodes or followers) they are going to distort terribly and need lots of 'almost-on-time' feedback to make it better again
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.