Hi, I just stumbled this as I'm resurrecting a long-dormant Audio Innovations First. Can someone explain to me why the path to ground from filamentary cathode is so DIFFERENT for the two halves of each PP pair in this amp? Each pair of tubes has one with 39R resistors from each end of the filament joining then to cathode resistor, while the other half of each pair has a 50R pot with wiper to cathode resistor.
What is up with that???
What is up with that???
phase splitter
Okay, this thread is dead for some time.
Referring to post #17 with schematics. Do I see it correctly that the (paraphase) phase splitter uses 47k and 44k resistors (so: not exactly the same values but slightly different)? Or is it just a typo or bad printing of the circuit?
Why are these values different? Of course I assume to make the phases more symmetrical and to obtain less distortion (assuming matched output pairs etc). But for every other phasesplitter-tube this will work out slightly different, is it?
As background:
- AI did use this arrangement many times. They always seem to use 220k/220k as splitter values (AI 500 etc).
- The First has seen different versions. In the latest version MK3 they have added a trim-pot of value 47k in between (the same) 47k resistors. Of course then it is possible to cancel out distortion etc using an oscilloscope. Then again Peter van Willenwaard points out that cancelling out will not give the best subjective results.
As I have a MK3 version I want to leave-out the trimpot so I can use different tubes without the need to calibrate every time again (no equipment to do that anyway). And I hope to hear the better subjective results? I just tried this 'mod' but to be honest can't hear any differency. I keep using 47k/47k as they were already there. Should I change to 44k for one of them?
Okay, this thread is dead for some time.
Referring to post #17 with schematics. Do I see it correctly that the (paraphase) phase splitter uses 47k and 44k resistors (so: not exactly the same values but slightly different)? Or is it just a typo or bad printing of the circuit?
Why are these values different? Of course I assume to make the phases more symmetrical and to obtain less distortion (assuming matched output pairs etc). But for every other phasesplitter-tube this will work out slightly different, is it?
As background:
- AI did use this arrangement many times. They always seem to use 220k/220k as splitter values (AI 500 etc).
- The First has seen different versions. In the latest version MK3 they have added a trim-pot of value 47k in between (the same) 47k resistors. Of course then it is possible to cancel out distortion etc using an oscilloscope. Then again Peter van Willenwaard points out that cancelling out will not give the best subjective results.
As I have a MK3 version I want to leave-out the trimpot so I can use different tubes without the need to calibrate every time again (no equipment to do that anyway). And I hope to hear the better subjective results? I just tried this 'mod' but to be honest can't hear any differency. I keep using 47k/47k as they were already there. Should I change to 44k for one of them?
Hi rmgvs
Phase splitting does not need to be perfect to work. I would keep it as is. There are other ways to improve this circuit, such as by using constant current sources and some mods to the PSU if you feel inclined.
Unfortunately, the chassis precludes many potential mods. Its a tight fit already.
Phase splitting does not need to be perfect to work. I would keep it as is. There are other ways to improve this circuit, such as by using constant current sources and some mods to the PSU if you feel inclined.
Unfortunately, the chassis precludes many potential mods. Its a tight fit already.
in the middle?
Thanks for your feedback Soulmerchant.
Would you suggest to put the pot in the middle-position? Just to go better with different tubes without having to fiddle every time?
BTW: now the pot (from factory settings) is in 33k / 14k position (totalling 47k), both channels.
And if true that it doesn't matter that much: why not taking out the pot altogether? It's a cheap small pot that can't do much good to the sound, is it? So: in that case we go back to the former circuit of post #16?
Thanks for your feedback Soulmerchant.
Would you suggest to put the pot in the middle-position? Just to go better with different tubes without having to fiddle every time?
BTW: now the pot (from factory settings) is in 33k / 14k position (totalling 47k), both channels.
And if true that it doesn't matter that much: why not taking out the pot altogether? It's a cheap small pot that can't do much good to the sound, is it? So: in that case we go back to the former circuit of post #16?
In that schematic the 44k and 47k are inverted.
If you put the 44k where the 47k drawn then the other resistance (Rfb) becomes
44k x (30+1)/(30-1) = ~ 47k where 30 is the (approx) gain of the ECC88.
That is without taking the gridleak (220k) into account, big enough to forget.You can even leave it out.
Mona
If you put the 44k where the 47k drawn then the other resistance (Rfb) becomes
44k x (30+1)/(30-1) = ~ 47k where 30 is the (approx) gain of the ECC88.
That is without taking the gridleak (220k) into account, big enough to forget.You can even leave it out.
Mona
ecc82
Thanks Ketje, that makes sense.
Actually, in this MK3 version ECC82's are used. So the factor would become (20+1)/(20-1) without taking into account the grid resistor. And that is assuming exactly same gains from the two halves of the triode.
So 47k becomes 48k5 (against 44k).
Given all kinds of tolerances (resistors with tolerance of 5% could just as well be 45k or 49k assuming no matching has been done), no problem at all to use 47k/47k or 100k/100k or whatever.
And not taking into account that all things perfectly matched and distortion cancelled out will sound kind of clinical. Why? This is asserted here: http://www.audiomagic.nl/archief/hf309_5.jpg (in dutch). See attachment taken from that link.
Thanks Ketje, that makes sense.
Actually, in this MK3 version ECC82's are used. So the factor would become (20+1)/(20-1) without taking into account the grid resistor. And that is assuming exactly same gains from the two halves of the triode.
So 47k becomes 48k5 (against 44k).
Given all kinds of tolerances (resistors with tolerance of 5% could just as well be 45k or 49k assuming no matching has been done), no problem at all to use 47k/47k or 100k/100k or whatever.
And not taking into account that all things perfectly matched and distortion cancelled out will sound kind of clinical. Why? This is asserted here: http://www.audiomagic.nl/archief/hf309_5.jpg (in dutch). See attachment taken from that link.
Attachments
There is nothing assymetrical about it.The lower triode must just perform a minus one gain.If it had a very big gain itself the two resistor can be equal.To get the voltage needed at the grid there is a difference.Unfortionally the gain of a tube varies with signal and so the resistors are only right on one point of operation.The greater the gain of the tube the better result, so an ECC82 isn't the best choice the ECC83 does much better job here but doesn't have much drive capablilities.
Mona
Mona
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.