A perfect CCS guarantees perfect current balance. Perfectly matched anode resistors then guarantee voltage balance.
There are no perfect CCS, but any reasonable CCS will be near enough perfect when compared with typical triode 1/gm. Possible exception is HF, when stray capacitance may degrade the CCS.
Ballpark figures: common-mode gain = Rload/(2 x Rtail), differential gain = mu.
We want the latter to be at least 50 times the former, preferably 100 times.
For a resistor tail, values are such that C-M gain ~ 1, so we want mu > 50.
For CCS tail, we want 2 x mu x Rccs / Rload > 50. This is not difficult to achieve, unless stray capacitance causes problems at HF. This could be worse with a high impedance stage (e.g. ECC83), where a resistor tail might be better as it could have less stray capacitance.
So use an ECC83 with a resistor (if driving pentodes), or your favourite 'linear' valve with a CCS (if driving beam tetrodes).
There are no perfect CCS, but any reasonable CCS will be near enough perfect when compared with typical triode 1/gm. Possible exception is HF, when stray capacitance may degrade the CCS.
Ballpark figures: common-mode gain = Rload/(2 x Rtail), differential gain = mu.
We want the latter to be at least 50 times the former, preferably 100 times.
For a resistor tail, values are such that C-M gain ~ 1, so we want mu > 50.
For CCS tail, we want 2 x mu x Rccs / Rload > 50. This is not difficult to achieve, unless stray capacitance causes problems at HF. This could be worse with a high impedance stage (e.g. ECC83), where a resistor tail might be better as it could have less stray capacitance.
So use an ECC83 with a resistor (if driving pentodes), or your favourite 'linear' valve with a CCS (if driving beam tetrodes).
The stray capacitance of a competent CCS (say, cascoded DN2540) is minuscule when considering scale. Take a look at the measurements from Walt Jung that appeared in AX a couple years ago- the impedance held up well to the limits of his measurements (1MHz, if memory serves) and was in the multi-gigohm range.
Multi-gigohm should be good enough! However, there will be strays from the cathodes (e.g. to the heaters) in parallel with the CCS. OK, I think I may be arguing against myself here as this will be a bigger issue in a high impedance circuit. That is the trouble with typing as I think! So a CCS with ECC83 will not do any harm, but I guess the type of person who likes to use CCS is probably the type of person who would never use an '83.
Hmmm. You may be right for the moment, but... since the ECC83 with a high plate load is exceptionally linear, there's always some idiot (often with initials SY) who might try doing just that. The LTP's outputs would have to be buffered to take advantage of the ECC83's high intrinsic linearity, which would discourage most duffers.
It's easy to be brave when you have data. 
http://www.diyaudio.com/forums/tubes-valves/164920-guess-tube.html#post2150434
Interestingly, datasheets generally do not give mu versus plate voltage but rather mu versus plate current. One can sometimes get two or three points if they show multiple curves, but the omission makes it harder to judge a tube's linearity under a CCS load.
http://www.diyaudio.com/forums/tubes-valves/164920-guess-tube.html#post2150434
Interestingly, datasheets generally do not give mu versus plate voltage but rather mu versus plate current. One can sometimes get two or three points if they show multiple curves, but the omission makes it harder to judge a tube's linearity under a CCS load.
I had not seen that particular Guess that Tube thread. I suppose the problem is people try to use an '83 in an '81-like circuit and are disappointed. Interestingly (as someone once pointed out to me), one of the 6SN7-family valves (I forget which one) gives a low current '83-like circuit in its datasheet (RCA?). Data doesn't cut much ice with some people - they prefer pronouncements from gurus!
Yes, the curves can be annoying for what they omit. I suppose it might be possible to fit some parameterised curves to the published data, then plot different views. A bit like preparing Spice models, but with a different aim.
Yes, the curves can be annoying for what they omit. I suppose it might be possible to fit some parameterised curves to the published data, then plot different views. A bit like preparing Spice models, but with a different aim.
In another nearby thread I made the recent discovery that a PNP concertina splitter
following an amplifying triode can be made to have equal impedances at both collector
and emitter. Because PNP phase returned to ground is flipped with respect to "normal"
can be tapped at 1/Mu to return Schaded negative feedback to the cathode of equal
influence to that of Mu from the plate. Equal feedbacks make for equal impedances.
This is actually a rediculously simple circuit requiring no constant current source. And
indeed provides bonus option free (but not terribly stiff) CCS for the amplifying triode.
Just by strapping a small bleed resistor across the near constant voltage of VBE.
I calculated a plate load slope of 190K with my very first try.
Full simulation here:
http://www.diyaudio.com/forums/tube...hase-splitter-yaqin-mc-84l-5.html#post2443445
Post#46
----------
I remember OldEurope's search for some very special triode with Mu of exactly 1 for
much the same purpose. With Mu of 1, Plate and Cathode both fight equal feedback.
There is perfect symmetry of misbehavior when unequal loads are switched position
from anode to cathode.
I found the secret lies not in Mu = 1, but Voltage feedback to Cathode = 1/Mu.
Meaning that high Mu voltage gain causes no problem. 12AX7 + MJE350 perfect!
Similar end result to Long Tail Pair with sandy current source, yet fewer parts.
Plenty of concertina current to drive the big bottles slightly into A2.
following an amplifying triode can be made to have equal impedances at both collector
and emitter. Because PNP phase returned to ground is flipped with respect to "normal"
can be tapped at 1/Mu to return Schaded negative feedback to the cathode of equal
influence to that of Mu from the plate. Equal feedbacks make for equal impedances.
This is actually a rediculously simple circuit requiring no constant current source. And
indeed provides bonus option free (but not terribly stiff) CCS for the amplifying triode.
Just by strapping a small bleed resistor across the near constant voltage of VBE.
I calculated a plate load slope of 190K with my very first try.
Full simulation here:
http://www.diyaudio.com/forums/tube...hase-splitter-yaqin-mc-84l-5.html#post2443445
Post#46
----------
I remember OldEurope's search for some very special triode with Mu of exactly 1 for
much the same purpose. With Mu of 1, Plate and Cathode both fight equal feedback.
There is perfect symmetry of misbehavior when unequal loads are switched position
from anode to cathode.
I found the secret lies not in Mu = 1, but Voltage feedback to Cathode = 1/Mu.
Meaning that high Mu voltage gain causes no problem. 12AX7 + MJE350 perfect!
Similar end result to Long Tail Pair with sandy current source, yet fewer parts.
Plenty of concertina current to drive the big bottles slightly into A2.
Last edited:
I know it's an old thread, but I would like to ask a question related to this point without opening a new thread.
Talking about a standard cascoded CCS below a PI like this one:
Tail CCS PCB test – Bartola(R) Valves
Can the current through the leds (so the value of R1) change the performance of the PI?
From simulations I'm doing, I see a reduction of the 3rd harmonic as the current increase. Is it a software bug, or there are some technical reasons behind and audible effects in reality?
Thanks
Talking about a standard cascoded CCS below a PI like this one:
Tail CCS PCB test – Bartola(R) Valves
Can the current through the leds (so the value of R1) change the performance of the PI?
From simulations I'm doing, I see a reduction of the 3rd harmonic as the current increase. Is it a software bug, or there are some technical reasons behind and audible effects in reality?
Thanks
Check your sim and see if the PI tail current changes when you change the led current. I suspect it does. Depending on the leds used and the current, you may find they are not perfect voltage sources at that operating point.
Anyway, it's irrelevant. Just change the led current until you get the desired tail current and do the same with the real hardware. Make sure you're at the appropriate point on the led curve.
It's unlikely the sim software has a bug, but the device models are rarely perfect. So small discrepancies are acceptable.
Anyway, it's irrelevant. Just change the led current until you get the desired tail current and do the same with the real hardware. Make sure you're at the appropriate point on the led curve.
It's unlikely the sim software has a bug, but the device models are rarely perfect. So small discrepancies are acceptable.
I found an article online thats been pretty helpful when selecting pre amp tubes, 12AX7 in this case. Ive not worried too much about the type for the phase inverter and have gotten pretty lucky so far.
But man good quality tubes are becoming harder to find. maybe this guide can help someone else too? Worth a look at any rate.
12AX7 Tube Guide with Reviews | Best Options 2021 | Killer Rig
But man good quality tubes are becoming harder to find. maybe this guide can help someone else too? Worth a look at any rate.
12AX7 Tube Guide with Reviews | Best Options 2021 | Killer Rig
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.