Yes, that seems odd. Perhaps then there is at least some charge carrier leakage between the triode sections, unless the screen is grounded? Essentially a noise floor which obscures the dip?
Possible. When I tried redoing the test today (shield ungrounded) I got a strange flat trace at about -30dB. I ran it a few more times with the same result and was totally confused, until I tried running the test one more time, and touched the shield pin with my finger and let go, half way through the test. The trace suddenly jumped to the normal path! I couldn't get it to repeat the phenomenon after that, although I think it might have something to do with applying the HT before the heater... I have heard of strange things happening with a floating shield in another thread years ago... hmm
Indeed, but I find it difficult to see how even a charged shield, sitting a few millimetres away from and between two high-voltage anodes, could have such a serious (and I mean serious) effect on crosstalk between those anodes?
EDIT: Unless perhaps it causes the glass to become charged too, via the valve pins, allowing direct conductivity between valve pins? DF96, I need a physicist...
EDIT: Unless perhaps it causes the glass to become charged too, via the valve pins, allowing direct conductivity between valve pins? DF96, I need a physicist...
Last edited:
If the shield/screen becomes significantly charged then this suggests that it can 'see' or 'be seen' by the electron stream in at least one of the triodes - probably through an anode hole which was there for assembly alignment purposes? An isolated piece of metal being hit by electrons can gain a negative charge, or if secondary emission dominates than it can gain a positive charge. In either case it can then affect, to a limited extent, the electron streams via the same holes.
My guess is that provided the shield/screen has a standing potential which is not too far different from the other electrodes it will have little effect. If it becomes very positive then it will attract electrons and due to its low capacitance could experience quite a large AC swing. Unfortunately this will be almost impossible to measure as attaching test equipment will immediately remove the standing charge.
My guess is that provided the shield/screen has a standing potential which is not too far different from the other electrodes it will have little effect. If it becomes very positive then it will attract electrons and due to its low capacitance could experience quite a large AC swing. Unfortunately this will be almost impossible to measure as attaching test equipment will immediately remove the standing charge.
Curiouser and curiouser. I'd expect the ungrounded shield to reach an equilibrium potential in the presence of charge carriers equal to the field potential at that location. If charge carrier density/field potential varied (eg signal in one triode) then this potential would vary. However, I'd expect the equilibrium to embrace any intial charge present on the shield, ie to re-establish a set point no matter what initial conditions - one exception is if the intial charge is so high as to prevent charge carriers from reaching the shield at all. In which case, I think a 'virtual capacitor' would form between charge carrier zone and the screen, and the electrostatic potential of the screen would remain as per initial charge, modified by signal/ac as though it were an unconnected plate of a capacitor. But to fit the observation, a large electrostatic potential of the screen would have to modify triode behaviour - not sure this stands scrutiny though.......although the potential could be very large in principle I think?
You don't just have fields, but also moving electrons.
If the screen starts sufficiently negative to repel electrons then it is likely to keep its potential. If not, electrons will hit it. If they hit at low velocity then they are likely to make it go more negative, so eventually it will repel them and stabilise. If they hit at high velocity then some secondary emission will occur. If this causes more than one electron to be emitted for each incoming one then the screen goes positive. This increases the collision velocity so encourages more secondary emission. We know that the metals used in valves can suffer from secondary emission, as this is why pentodes were invented to solve the problem for tetrodes. So thw screen can go into 'electric runaway'.
So now you have a very positive electrode, which can be 'seen' by the electrons in each triode. It will attract some proportion of the cathode current, thus both affecting and being affected by the current.
If the screen starts sufficiently negative to repel electrons then it is likely to keep its potential. If not, electrons will hit it. If they hit at low velocity then they are likely to make it go more negative, so eventually it will repel them and stabilise. If they hit at high velocity then some secondary emission will occur. If this causes more than one electron to be emitted for each incoming one then the screen goes positive. This increases the collision velocity so encourages more secondary emission. We know that the metals used in valves can suffer from secondary emission, as this is why pentodes were invented to solve the problem for tetrodes. So thw screen can go into 'electric runaway'.
So now you have a very positive electrode, which can be 'seen' by the electrons in each triode. It will attract some proportion of the cathode current, thus both affecting and being affected by the current.
Great thread!
I find this topic very interesting because I do a lot of balanced circuits with double triodes. There of course if you have crosstalk at high frequencies, the signals cancel each other out partially, and the frequency response should slope down a bit.
However I've measured 6N16B (which is super tiny and would probably have a lot of crosstalk) and 6N3P and also 6SN7 balanced circuits to have flat response up to at least 30 to 50 kHz, sometimes a lot more.
Looking at measurements in this thread, I wonder shouldn't crosstalk effect it more? What is going on?
I find this topic very interesting because I do a lot of balanced circuits with double triodes. There of course if you have crosstalk at high frequencies, the signals cancel each other out partially, and the frequency response should slope down a bit.
However I've measured 6N16B (which is super tiny and would probably have a lot of crosstalk) and 6N3P and also 6SN7 balanced circuits to have flat response up to at least 30 to 50 kHz, sometimes a lot more.
Looking at measurements in this thread, I wonder shouldn't crosstalk effect it more? What is going on?
Also regarding the "acceptable level of crosstalk":
If you have a system with let's say 5 stages between cartridge and speaker, and each stage has 30dB of channel separation, the total effect is that the channels mix much more along the way. Each stage mixes the signals a bit more. If you want that 30dB at the output of the system, you should have as much channel separation as possible in the first stages. Probably have separate envelope tubes in the RIAA stages.
If you have a system with let's say 5 stages between cartridge and speaker, and each stage has 30dB of channel separation, the total effect is that the channels mix much more along the way. Each stage mixes the signals a bit more. If you want that 30dB at the output of the system, you should have as much channel separation as possible in the first stages. Probably have separate envelope tubes in the RIAA stages.
You need an awful lot of cross-talk before it affects the frequency response. 30dB crosstalk from antiphase could drop the signal by 3% or 0.26dB.
To get high HF crosstalk by this method you need high impedance nodes, such as ECC83/12AX7 anodes or cascodes sharing envelopes between channels. It is unfortunate that the condition for low triode distortion is also the condition for maximum crosstalk.
However, bear in mind that LF crosstalk comes from inadequate shared supply rail decoupling. In some circuits that could be a worse problem, although perhaps less noticeable audibly.
To get high HF crosstalk by this method you need high impedance nodes, such as ECC83/12AX7 anodes or cascodes sharing envelopes between channels. It is unfortunate that the condition for low triode distortion is also the condition for maximum crosstalk.
However, bear in mind that LF crosstalk comes from inadequate shared supply rail decoupling. In some circuits that could be a worse problem, although perhaps less noticeable audibly.
Last edited:
If a 12AX7 can manage better than 50dB separation across most of the audio band, crosstalk isn't going to be a problem, even in an LTP. And most triodes are better than the 12AX7 (in terms of crosstalk).
i did not record specifically which of the old stock 6SN7 tubes i tested other than what i have described in the previous messages and information.
Based on the testing and the comments of others, i chose to use the triodes within the 6SN7 for similar signals for the right and left amplifiers. therefore, each triode had the same signal amplitude, therefore, the crosstalk to each triode would be similar.
The original thought was to use the two triodes for two stages of amplification of the same channel, but that would have meant that the second stage of amplification would have been stronger than the first stage causing more crosstalk in to the first stage and at 180 degrees phase shift.
Of course, if my logic was flawed, i would like to have those with more experience correct my thinking.
Based on the testing and the comments of others, i chose to use the triodes within the 6SN7 for similar signals for the right and left amplifiers. therefore, each triode had the same signal amplitude, therefore, the crosstalk to each triode would be similar.
The original thought was to use the two triodes for two stages of amplification of the same channel, but that would have meant that the second stage of amplification would have been stronger than the first stage causing more crosstalk in to the first stage and at 180 degrees phase shift.
Of course, if my logic was flawed, i would like to have those with more experience correct my thinking.
Thanks for the clarification. I went back just now to your initial post, and see pictured standard large parallel plate 6SN7s, but I can't tell if they're the "staggered" plate type. Then there's the Sylvania construction with the two plates angled with respect to each other, my guess is that would be lower capacitance between the plates.
** Inter-Channel Crosstalk Distortion **
Thank-you for the great measurements and discussions in this thread. A related question has tickled my curiosity for many years:
Start with a stereo amplifier containing channel 1 and channel 2.
If we have a crosstalk signal of -50dB in channel 2 of the channel 1 monaural signal, and if that crosstalk signal in channel 2 is distorted 10% due to the perhaps nonlinear inter-electrode coupling, should we be able to measure this distortion in channel 2?
My TEK Scope FFT does not have the resolving power for this experiment, but we could hunt for a distortion signal below -70dB. This may require some differencing measurement techniques...
Could this inter-channel crosstalk distortion help explain the preference some have for a particular construction style within a given tube type?
David
Thank-you for the great measurements and discussions in this thread. A related question has tickled my curiosity for many years:
Start with a stereo amplifier containing channel 1 and channel 2.
If we have a crosstalk signal of -50dB in channel 2 of the channel 1 monaural signal, and if that crosstalk signal in channel 2 is distorted 10% due to the perhaps nonlinear inter-electrode coupling, should we be able to measure this distortion in channel 2?
My TEK Scope FFT does not have the resolving power for this experiment, but we could hunt for a distortion signal below -70dB. This may require some differencing measurement techniques...
Could this inter-channel crosstalk distortion help explain the preference some have for a particular construction style within a given tube type?
David
The usual crosstalk mechanism (stray capacitance - the one being considered in this thread) is linear, so it won't create distortion.
There is another mechanism - common resistance/inductance - which is also linear, but sometimes the currents through it may contain distortion. For example, the tail current of a well-driven push-pull output stage will contain even order distortion. Even so, it is likely that any problem here will affect the relevant channel more than cross-talk in the other channel. It may be easier to measure the distorted cross-talk in an otherwise quiet channel.
Good or bad amps can be built using almost any construction style.
There is another mechanism - common resistance/inductance - which is also linear, but sometimes the currents through it may contain distortion. For example, the tail current of a well-driven push-pull output stage will contain even order distortion. Even so, it is likely that any problem here will affect the relevant channel more than cross-talk in the other channel. It may be easier to measure the distorted cross-talk in an otherwise quiet channel.
Good or bad amps can be built using almost any construction style.
Thanks, DF. So you are saying the primary crosstalk mechanism is capacitive and linear, and so we should not see crosstalk distortion per se. That is good news and makes sense, using a classic model of two metal parts. My familiarity is with the world of non-linear conductors, where starting with an estimate of some non-linearity is usually prudent.
My continuing curiosity and confusion comes from the oscillograms in post #26: "phase shift of crosstalk freq" (esp the first and second trace), and post #40: "cross talk frequency response of tests". Most of the crosstalk waveforms seem to be visibly distorted to me...? Perhaps this apparent "distortion" I see is simply noise?
David
My continuing curiosity and confusion comes from the oscillograms in post #26: "phase shift of crosstalk freq" (esp the first and second trace), and post #40: "cross talk frequency response of tests". Most of the crosstalk waveforms seem to be visibly distorted to me...? Perhaps this apparent "distortion" I see is simply noise?
David
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.