Same idea, different way of saying it.DualTriode said:
I suspect that in a genuine balanced SRPP (i.e. same top and bottom, no cathode bypass) the 12AU7/ECC82 would work quite well as there is some cancellation. For an unbalanced one, including mu follower, there is much less cancellation so you have to rely on the valve's intrinsic linearity - other valves can beat the ECC82.
Both of the things you mention are not always true. For even harmonics things may not be equal and opposite. Cancelation may not be perfect, perhaps on a good day. For odd harmonics it is not so much canceling but transfer characteristic curves sloping in opposite directions.
Many tube data sheets include “Average Transfer Characteristic” curves. The slope of the curve is gm. From low grid voltage on the left to less negative grid voltage on the right the tube data sheets show increasing slope (gm). The difference between straight line constant slope and the tube data sheet is the source of the 3rd harmonic. Now add a cathode resistor to the average tube in the data sheet, with an increasing cathode resistor value the Transfer Characteristic” curve begins to slope in the opposite direction minimizing the 3rd harmonic as the composite curve approaches a straight line.
So yes, on a good day you can tune a circuit’s composite “Transfer Characteristic” curve to minimize both even and odd harmonics.
Sometimes they do not tell you everything in class.
DT
True, but if signal levels are sufficiently low then higher orders will be low too.Wavebourn said:
This only works if the sign of the intrinsic 3rd is opposite to the added distortion. Sometimes it works, sometimes it doesn't. For an ideal 3/2 law triode it doesn't work. For an ideal remote-cutoff triode it does work. Real valves are somewhere in between.DualTriode said:
Works in SPICE using a 12AU7 Marshall Leach 3/2’s Child’s Law model, worked on the bench with a (made in Canada Amperex 9AU7’s) test Circuit and Audio Tester FFT. Could be because the tested output voltage was at 2 Volts RMS where the 3rd and higher harmonics are tiny to start with.
Page 5 of the linked 12AU7 data sheet (real tube data) shows a positive rate of change of plate current as grid voltage becomes less negative (this curve is without a cathode resistor). Add increasing values of cathode resistor and the rate of change goes from positive to zero and then negative. The added 3rd due to the cathode resistor harmonic is opposite in sign to the original (I believe this is how you put it).
http://www.mif.pg.gda.pl/homepages/frank/sheets/093/1/12AU7A.pdf
DT
Because of the sign of the second-order distortion, cathode degeneration will always introduce compressive 3rd-order. Thus for 3rd-order cancellation you need intrinsic expansive 3rd-order. The 12AU7A data sheet (on page 5) appears to show compressive 3rd-order for high currents, as the curves approach a straight line, but I accept that they may be expansive at low currents. The risk is that in order to cancel some 3rd-order you end up accepting higher amounts of other orders because you are operating in a non-linear region of the valve characteristic. However, the 12AU7 is unusual in that it is almost square-law in places (this is why it makes a good receiver mixer). Of course in a real circuit there is also the effect of varying mu to complicate things (see page 3 of the linked data sheet).
I worked out some of the maths here, in the context of receiver performance. Note the difference in sign of the third-order term when comparing 3/2 with exponential responses.
I think we can agree that a little cathode degeneration may sometimes reduce 3rd-order. The wrong amount, or at the wrong bias, or with the wrong valve, could make things worse.
I worked out some of the maths here, in the context of receiver performance. Note the difference in sign of the third-order term when comparing 3/2 with exponential responses.
I think we can agree that a little cathode degeneration may sometimes reduce 3rd-order. The wrong amount, or at the wrong bias, or with the wrong valve, could make things worse.
I have just thought of an explanation. At low currents, where island effect is present, many valves follow a 5/2 law. This would give expansive 3rd, so a little cathode degeneration will cancel it.
It is possible that a "3/2" model may include some 5/2 at low currents, or achieve the same effect in a different way such as reducing mu.
It is possible that a "3/2" model may include some 5/2 at low currents, or achieve the same effect in a different way such as reducing mu.
DF96,
Thanks for sharing the IP3 theory and Maths. Also thanks for sharing your thoughts plus explanation. I like your solution oriented approach to see how the data fits the models.
A couple of years ago I looked for and did not find SPICE models for 5670 and 12B4. I adjusted similar models to match the performance tested on the bench. This worked pretty well above the knee of the curves.
I found a set of plate curves on line for a triode connected 6BQ6, not your ideal 3/2’s law triode. The plan is to tweak the value of the cathode resistor plus varying degrees of cathode resistor bypass by adjusting the value of a resistor placed in series with the bypass capacitor. It will be interesting to see what happens to with the FFT and how clean I can get a few watts of Push-Pull output to be.
Just for fun!
DT
Thanks for sharing the IP3 theory and Maths. Also thanks for sharing your thoughts plus explanation. I like your solution oriented approach to see how the data fits the models.
A couple of years ago I looked for and did not find SPICE models for 5670 and 12B4. I adjusted similar models to match the performance tested on the bench. This worked pretty well above the knee of the curves.
I found a set of plate curves on line for a triode connected 6BQ6, not your ideal 3/2’s law triode. The plan is to tweak the value of the cathode resistor plus varying degrees of cathode resistor bypass by adjusting the value of a resistor placed in series with the bypass capacitor. It will be interesting to see what happens to with the FFT and how clean I can get a few watts of Push-Pull output to be.
Just for fun!
DT
I mentioned "cancellation" by another device with similar curve, like in SRPP example, so it is always true in this case. The main advantage of SRPP is not cancelation. It is dynamic load of bottom tube and using the top one as cathode follower as well, i.e. as the result the bottom tube sees higher load resistance, that is the source of lower distortions.
Last edited:
I will save the never and always argument for when it is important, like with my wife. If there is an exception to a True False question the answer is always False.
DIYAUDIO is a good place. there is a world of knowledge and experience, much to be shared and learned. I worked with a Mechanical Engineer from Transilvania Romania. Paul always had a fresh view on how to solve a problem. We argued a lot. I think that those discussions produced better solutions.
In terms of SRPP I agree that there is likely a lot of distortion cancelation. I also agree that the performance is highly dependent on load. If the connected load is not considered in the design the balance between push and pull can be upset and what is thought to be equal and opposite cancelation is not true. This is the exception that makes the answer False.
People love to argue about SRPP.
DT
DIYAUDIO is a good place. there is a world of knowledge and experience, much to be shared and learned. I worked with a Mechanical Engineer from Transilvania Romania. Paul always had a fresh view on how to solve a problem. We argued a lot. I think that those discussions produced better solutions.
In terms of SRPP I agree that there is likely a lot of distortion cancelation. I also agree that the performance is highly dependent on load. If the connected load is not considered in the design the balance between push and pull can be upset and what is thought to be equal and opposite cancelation is not true. This is the exception that makes the answer False.
People love to argue about SRPP.
DT
In a balanced SRPP (i.e. top and bottom identical, no cathode bypass, only lightly loaded) cancellation is a major effect. In an unbalanced SRPP (e.g. bottom cathode bypassed, or mu-follower) there is little cancellation, so the higher load resistance is the main source of lower distortion. In both cases there is further advantage from the reduced output impedance due to the modified CF behaviour of the top valve. Most people use an unbalanced SRPP, so your statement is true for the most common case.Wavebourn said:
Sure. When linearizing by feedback all possible means of linearization of open loop amp are valid, in terms of linearity. But in no feedback amp the only way to go is to put the tube into most privileged conditions.
I see both cancelation (feedback included) and parameter restriction as design tools.
Speaking of parameter restriction we can increase the load impedance to the extreme of the active load of a mu-follower. When we do this the best we can do is approach the inherent distortion limit imposed by the 3/2’s law of the Ideal triode. There is a limited selection of ideal triodes!
When the SS JFET guys get to this point they add the cancelation of a differential pair.
I scratched the above to post half a day ago. Wavebourn you posted just as I was to pull trigger. I thought the topic was beaten to near death, I guess not.
When we reduce swing to limit distortion one trade off is to reduce the Signal to Noise Raito. Or put another way, say we reduce the output from 20 volts where it popular to test and report distortion data. At 20 volts the higher harmonics are measurable above the grass of the noise floor and within the ability of the test gear to detect. At 2 volts output which is greater than a line amplifier will be used or 10 db down the noise floor is still where it was, the fundamental and harmonics are 10 db less. The higher harmonics are likely lost down in the grass where no one sees or cares.
DT
Speaking of parameter restriction we can increase the load impedance to the extreme of the active load of a mu-follower. When we do this the best we can do is approach the inherent distortion limit imposed by the 3/2’s law of the Ideal triode. There is a limited selection of ideal triodes!
When the SS JFET guys get to this point they add the cancelation of a differential pair.
I scratched the above to post half a day ago. Wavebourn you posted just as I was to pull trigger. I thought the topic was beaten to near death, I guess not.
When we reduce swing to limit distortion one trade off is to reduce the Signal to Noise Raito. Or put another way, say we reduce the output from 20 volts where it popular to test and report distortion data. At 20 volts the higher harmonics are measurable above the grass of the noise floor and within the ability of the test gear to detect. At 2 volts output which is greater than a line amplifier will be used or 10 db down the noise floor is still where it was, the fundamental and harmonics are 10 db less. The higher harmonics are likely lost down in the grass where no one sees or cares.
DT
About the question on the existence or non-existence of ECC80, some could be confused about a single triode EC80, which existed for sure, I have a pair of NOS Valvo, it's in B9A format, with multiple pin connection, I wonder why they did this way, maybe for redundancy? They are 3 pins connected together for the grid...
The data on EC80 therefore excludes the replacement even with socket/PCB modification, it's much higher mu and S than half of an E80CC or the 6SN7/6N8S in discussion. I think it could be an interesting triode, for example for a phone amplifier, also for a small speaker amplifier, what can you say about the linearity of these high mu and high S triodes? The anode power is very tempting - 4W (maybe because of the space being single in the tube).
The data on EC80 therefore excludes the replacement even with socket/PCB modification, it's much higher mu and S than half of an E80CC or the 6SN7/6N8S in discussion. I think it could be an interesting triode, for example for a phone amplifier, also for a small speaker amplifier, what can you say about the linearity of these high mu and high S triodes? The anode power is very tempting - 4W (maybe because of the space being single in the tube).
Last edited:
Hi,
That valve has nothing to do whatsoever with a 6SN7 but this document should answer most of your questions I believe:
http://frank.pocnet.net/sheets/046/e/EC80.pdf
See also :
http://www.diyaudio.com/forums/tubes-valves/230357-ec80-preamp-tube.html
Cheers,
That valve has nothing to do whatsoever with a 6SN7 but this document should answer most of your questions I believe:
http://frank.pocnet.net/sheets/046/e/EC80.pdf
See also :
http://www.diyaudio.com/forums/tubes-valves/230357-ec80-preamp-tube.html
Cheers,
Last edited:
EC80 is probably quite unconnected with ECC80 or E80CC. Multiple grid pins are because it is intended for grounded grid RF service, so a low inductance connection to the grid is needed. Not relevant for audio. This sort of triode is intended for either small signal RF gain or for RF oscillation. In either case the sort of linearity we want for audio are not a design aim.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.