This is a very simple circuit -- of course there are some compromises here and there.
One of them is the diff stage or input stage/long tail splitter, that it’s not very well balanced. Which will cause some distortion, especially 2nd harmonic (one phase is not equivalent in amplitude too the other).
The sad thing is that the global feedback will be feed through the other side of that stage and produce even higher order distortion. So, 2nd and 3rd order distortion will become 4th, 5th, and 6th order distortion at the output.
The main cure for that is to fix the problem where it's generated: In the diff stage.
Here are some spectrum analyzes of the output signal at 5V rms, before and after putting in a CCS to balance the driver stage:
Output with a 10k resistor as tail (input is 400Hz):
Output with a CCS as tail (input is 400Hz):
It is magic, isn't it?
Jan E Veiset
NO-6600
One of them is the diff stage or input stage/long tail splitter, that it’s not very well balanced. Which will cause some distortion, especially 2nd harmonic (one phase is not equivalent in amplitude too the other).
The sad thing is that the global feedback will be feed through the other side of that stage and produce even higher order distortion. So, 2nd and 3rd order distortion will become 4th, 5th, and 6th order distortion at the output.
An externally hosted image should be here but it was not working when we last tested it.
The main cure for that is to fix the problem where it's generated: In the diff stage.
Here are some spectrum analyzes of the output signal at 5V rms, before and after putting in a CCS to balance the driver stage:
Output with a 10k resistor as tail (input is 400Hz):
An externally hosted image should be here but it was not working when we last tested it.
Output with a CCS as tail (input is 400Hz):
An externally hosted image should be here but it was not working when we last tested it.
It is magic, isn't it?
Jan E Veiset
NO-6600
Part of my standard toolkit now are IXYS CCSs pre-mounted on heatsinks with 10-turn pot adjusts. Incidentally, you may already know but with an adjustable CCS I've found it's possible to null the 3rd and higher deeply if you ignore the 2nd. So far all my projects have been SE though it sounds very promising for PP.
Eliminating 2nd and oth. even harmonics does't lead to better sound by all means. The ratio 2nd/3rd/... is more important imho.
Agreed. The odd harmonics are not pleasant, even in small quantities, and they get worse the higher the order.
Tuning the sound.
Jan,
Thanks for posting the spectrum analyser outputs.
I have found that in using current sources you can in fact make the sound too "dry" - that is too much 3H or not enough 2H.
On solid state stuff I generally match the devices - then use a high value resistive tail OR a single transistor current source - occasionally deliberately compromised by a resistor added between collector and emitter. I use a maximum of 10% emitter degeneration on the diff pair.
On tube stuff I have generally found that I need a cascode transistor current source and rely on increasing 2H as I turn up the volume to balance the 2H and 3H. BUT its still possible to overdo the performance of the CCS. Once again its possible to compromise the CCS performance to "tune" the harmonic content by wiring a resistor from collector to emitter in the bottom transistor.
Cheers,
Ian
Jan,
Thanks for posting the spectrum analyser outputs.
I have found that in using current sources you can in fact make the sound too "dry" - that is too much 3H or not enough 2H.
On solid state stuff I generally match the devices - then use a high value resistive tail OR a single transistor current source - occasionally deliberately compromised by a resistor added between collector and emitter. I use a maximum of 10% emitter degeneration on the diff pair.
On tube stuff I have generally found that I need a cascode transistor current source and rely on increasing 2H as I turn up the volume to balance the 2H and 3H. BUT its still possible to overdo the performance of the CCS. Once again its possible to compromise the CCS performance to "tune" the harmonic content by wiring a resistor from collector to emitter in the bottom transistor.
Cheers,
Ian
Re: Tuning the sound.
The posting of Jan is wery useful anyway! The CCS makes the LTP more symmetric. Every triod of LPT becomes better voltage follower.
It's a good method to combine the advantiges - lower |-v| ("C-" on the circuit) and acceptable 2nd/3rd/... ratio. Nice idea, Ian.gingertube said:
The posting of Jan is wery useful anyway! The CCS makes the LTP more symmetric. Every triod of LPT becomes better voltage follower.
Wow that's an impressive result.
Does anyone have any links to schematics for CCS, or where they can be bought?
Thanks
Does anyone have any links to schematics for CCS, or where they can be bought?
Thanks
The basis of my first posting was pure technical. What sounds better, more sweet, what ever, is of course another discussion. 😉
The CCS that I use is a very simple one, just one single transistor (MJE340) and the reflected source impedance is about ~100k. I'll suppose the measurements would have been even better with a proper CCS...
Jan E Veiset
NO-6600
The CCS that I use is a very simple one, just one single transistor (MJE340) and the reflected source impedance is about ~100k. I'll suppose the measurements would have been even better with a proper CCS...
Jan E Veiset
NO-6600
In such a case - well done, Jan. I saw your beautiful design in the Photo gallery as well. I personally like PP sound. It's more precise compared to SE. Whereas SE is more human's ear friendly due to specific harmonic ratio IMHO.jane said:
Hi,
This result with low distortion is only valid if the 2 triodes are identical which they normally are in a spice simulator but not in real life, not even if they are in the same envelope......
It is possible and even more likely to get a similar low value of even order distortion without the CCS given that the differential pair has some means of balance control, for instance that part of the cathode resistor is a pot, the CCS is not necessary. The same goes for a push-pull amplifier, you can get extremely low values of even order distortion when using some sort of balance control.
Regards Hans
This result with low distortion is only valid if the 2 triodes are identical which they normally are in a spice simulator but not in real life, not even if they are in the same envelope......
It is possible and even more likely to get a similar low value of even order distortion without the CCS given that the differential pair has some means of balance control, for instance that part of the cathode resistor is a pot, the CCS is not necessary. The same goes for a push-pull amplifier, you can get extremely low values of even order distortion when using some sort of balance control.
Regards Hans
Would it be possible to combine a balance pot with a CCS? Something along the lines perhaps of a small value pot, wiper to the top of the CCS and cathodes on either end with safety resistors to cover failure? The potential is appealing because I've regularly seen 3rd and higher go 100 dB+ down (<0.001%) in SE with CCS plate loading and current trimming into reasonable loads, leaving only second. A means of nulling that in PP mode holds the promise of crazy performance.
It may be possible to use a pot with a CCS as you suggest, since the pot will alter the bias on each tube to make up for tube imbalance. Even if you get the input sections, and the phase splitter balanced to perfection, the output tubes and the OPT will never be perfectly balanced across the range of signal levels seen there.
I still think that puting a pot here is a worthwhile thing to do. The pot would also allow experimentation with the sound quality VS the balance.
I have found that an amp with more 3H than 2H doesn't sound that good, but that is my opinion. Almost everyone will agree that 5H and higher odds sound terrible.
I still think that puting a pot here is a worthwhile thing to do. The pot would also allow experimentation with the sound quality VS the balance.
I have found that an amp with more 3H than 2H doesn't sound that good, but that is my opinion. Almost everyone will agree that 5H and higher odds sound terrible.
tubelab.com said:
Hello Tubelab,
you're right, but you can give the amp a reference signal and watch the spectrum analyzer while turning the pot. So you can correct the "whole" amp balance 😉
If only I could make a software FFT work...
CCS to Pot wiper
Guys,
What you are doing with this arrangement is not so much varying the bias (although you will do that too) between one side and the other BUT you are degenerating the gm of the two sides (equally or unequally, depending on pot wiper position) to try to make them equal. That is: you are balancing both the operating point and the individual device characteristics.
For a take on what that does to a diff amp, search for the Doug Self article titled "Distortion in Power Amplifiers" - yep, its all about solid state BUT the descriptions of what happens in the diff amp front end and what contributes to even and odd harmonic distortions are worth a read and are applicable to tube diffamps. Like any paper on "blamless" amplifiers you need to read it keeping in mind that we are not trying to create a "blameless amplifier" BUT rather a "gorgeous amplifier". That applies whether we are designing a tube amp or a solid state amp - which is where the SS croud have largely lost the plot in the last 30 years in seeking ever more decimal points in harmonic distortion specifications.
The pot in the diff amp cathodes with wiper fed from the CCS is as good idea and one I have used a lot - ref "Baby Huey" etc.
Cheers,
Ian
Guys,
What you are doing with this arrangement is not so much varying the bias (although you will do that too) between one side and the other BUT you are degenerating the gm of the two sides (equally or unequally, depending on pot wiper position) to try to make them equal. That is: you are balancing both the operating point and the individual device characteristics.
For a take on what that does to a diff amp, search for the Doug Self article titled "Distortion in Power Amplifiers" - yep, its all about solid state BUT the descriptions of what happens in the diff amp front end and what contributes to even and odd harmonic distortions are worth a read and are applicable to tube diffamps. Like any paper on "blamless" amplifiers you need to read it keeping in mind that we are not trying to create a "blameless amplifier" BUT rather a "gorgeous amplifier". That applies whether we are designing a tube amp or a solid state amp - which is where the SS croud have largely lost the plot in the last 30 years in seeking ever more decimal points in harmonic distortion specifications.
The pot in the diff amp cathodes with wiper fed from the CCS is as good idea and one I have used a lot - ref "Baby Huey" etc.
Cheers,
Ian
Try this experiment- make a long tail pair with a good CCS (like a cascode bipolar), no balancing, and deliberately dissimilar tubes for each side- maybe a 12AX7 and 12AT7. When you're done looking at the balance over frequency and the distortion spectrum, you may reconsider the need for that pot.
SY,
I had a thunk about it and I'm convinced. With a good CCS then whatever one tube gives up in current the other must pick up (an vv). So as long as anode loads are equal then output signals at the anodes MUST be equal and opposite. So the pots can go.
Depending upon the amount of current swing it may still be worthwhile having EQUAL gm degeneration resistors in the cathodes to linearize differential gain vs input voltage swing. This is a symmetrical thing and so its odd order harmonics that it should reduce.
The ONLY experience I've had with this so far is with a solid state diff pair of matched 2N5401 with a relatively high value resistive tail. (the resistive tail imparted some "tubiness" which dissappeared when I changed to CCS tail so I took the CCS out again).
I found that about 10% degeneration of gm SOUNDED better but too much degeneration seemed to kill the presence. That would stack up with the view that degeneration affects odd harmonic distortion.
I'll need to repeat that test with tubes.
I have noted that on the "Baby Huey" and on an EL34 PP Amp in which I run diffamp front ends, the pot helps balance the DC anode voltages BUT swinging from one end to the other did'nt seem to affect the sound at all.
Cheers,
Ian
I had a thunk about it and I'm convinced. With a good CCS then whatever one tube gives up in current the other must pick up (an vv). So as long as anode loads are equal then output signals at the anodes MUST be equal and opposite. So the pots can go.
Depending upon the amount of current swing it may still be worthwhile having EQUAL gm degeneration resistors in the cathodes to linearize differential gain vs input voltage swing. This is a symmetrical thing and so its odd order harmonics that it should reduce.
The ONLY experience I've had with this so far is with a solid state diff pair of matched 2N5401 with a relatively high value resistive tail. (the resistive tail imparted some "tubiness" which dissappeared when I changed to CCS tail so I took the CCS out again).
I found that about 10% degeneration of gm SOUNDED better but too much degeneration seemed to kill the presence. That would stack up with the view that degeneration affects odd harmonic distortion.
I'll need to repeat that test with tubes.
I have noted that on the "Baby Huey" and on an EL34 PP Amp in which I run diffamp front ends, the pot helps balance the DC anode voltages BUT swinging from one end to the other did'nt seem to affect the sound at all.
Cheers,
Ian
As far as I can see there will always be some imbalance in a classic long tail pair. The ac current in one cathode is not equivalent to the ac current in the other cathode. The current in the second cathode Ik2 = Ik1 – It (where It is the ac current in the tail resistor)
The larger the tail is the better the balance is. A CCS will act as a very large tail resistor.
If I got it right, Ik1 and Ik2 will be:
Ik1[ac} = V/[(1/gm[1]) + ((1/gm[2])||Rt)]
Ik2[ac] = V/[(1/gm[1]) + ((1/gm[2])||Rt)] – V*[(1/gm[2]||Rt)]/[(1/gm[1])+((1/gm[2])||Rt)]/Rt
Rt is the tail resistor.
With a proper CCS this can be reduce to:
Ik1[ac] = V/[(1/gm[1])+(1/gm[2])]
Ik2[ac] = V/[(1/gm[1])+(1/gm[2])] = Ik1[ac]
The gain of a LTP will be ½ of a CC since Vgk is ½ Vin.
It also indicate that rp’ = rp + mu * 1/gm.
(I probably did some usual mistakes in my thinking or equations, so please correct me where I might be wrong 😉 )
Jan E Veiset
NO-6600
The larger the tail is the better the balance is. A CCS will act as a very large tail resistor.
If I got it right, Ik1 and Ik2 will be:
Ik1[ac} = V/[(1/gm[1]) + ((1/gm[2])||Rt)]
Ik2[ac] = V/[(1/gm[1]) + ((1/gm[2])||Rt)] – V*[(1/gm[2]||Rt)]/[(1/gm[1])+((1/gm[2])||Rt)]/Rt
Rt is the tail resistor.
With a proper CCS this can be reduce to:
Ik1[ac] = V/[(1/gm[1])+(1/gm[2])]
Ik2[ac] = V/[(1/gm[1])+(1/gm[2])] = Ik1[ac]
The gain of a LTP will be ½ of a CC since Vgk is ½ Vin.
It also indicate that rp’ = rp + mu * 1/gm.
(I probably did some usual mistakes in my thinking or equations, so please correct me where I might be wrong 😉 )
Jan E Veiset
NO-6600
- Status
- Not open for further replies.