BTW. i will take a look at the suggestions for additional testing and better notes on things like what was grounded or not and publish a second set of test results with the additions probably sometime next week. dont mind doing a binge of testing once more for the community and my own edification.
I did testing like this some years ago. I can't find exact results, but remember that the miniature dual triodes were not bad, but there was still coupling - mostly capacitive. This is a problem in a preamp where the plate impedances are high, such as in stages with 220K to 270K plate resistors.
I remember reading an interview from the 1970s with (I think) Sid Smith of Marantz where he complained that "modern" 12AX7/ECC83 tubes were no good because they had too much coupling between sections. If you look at the schematic of the Marantz 7 preamp, every 12AX7 is shared between channels.
The octal dual triodes had much more variation, and in particular I remember some 6SL7GT WW2-vintage tubes had very high cross-talk that was non-linear - indicating crosstalk via electron flow. These tubes had rather large gaps at the top and bottom of the plate structures. To get definitive test results, it is important to test as many structural variants as possible.
An interesting test (which I didn't do) would be to compare crosstalk between a 6CG7 and 6FQ7. The 6CG7 was released by RCA in 1954 as a 9-pin miniature version of the 6SN7GT. It had a shield between the two sections, connected to pin 9. RCA later came out with the 6FQ7 in 1961 which was identical to the 6CG7 but minus the shield. This must have been done for cost-cutting reasons. In any case, comparison of these two types would be interesting.
- John Atwood
I remember reading an interview from the 1970s with (I think) Sid Smith of Marantz where he complained that "modern" 12AX7/ECC83 tubes were no good because they had too much coupling between sections. If you look at the schematic of the Marantz 7 preamp, every 12AX7 is shared between channels.
The octal dual triodes had much more variation, and in particular I remember some 6SL7GT WW2-vintage tubes had very high cross-talk that was non-linear - indicating crosstalk via electron flow. These tubes had rather large gaps at the top and bottom of the plate structures. To get definitive test results, it is important to test as many structural variants as possible.
An interesting test (which I didn't do) would be to compare crosstalk between a 6CG7 and 6FQ7. The 6CG7 was released by RCA in 1954 as a 9-pin miniature version of the 6SN7GT. It had a shield between the two sections, connected to pin 9. RCA later came out with the 6FQ7 in 1961 which was identical to the 6CG7 but minus the shield. This must have been done for cost-cutting reasons. In any case, comparison of these two types would be interesting.
- John Atwood
I tested this amp: http://www.diyaudio.com/forums/tubes-valves/125488-loftin-white-801-amp.html#post1551490
Actually using the local version 7n7 for the input tube. I get 60dB at 2kHz both channels, and at 20kHz, 56 on one channel and 42 dB on the other. That's input to output with the test channel unloaded. Loaded, it's a little less. Almost identical result with a second input tube. Small sample size, but something inherent in tube construction?
I should note that initially the results were worse. I noticed that I had twisted the inputs together (dumb), thanks for the heads up.
Sheldon
Actually using the local version 7n7 for the input tube. I get 60dB at 2kHz both channels, and at 20kHz, 56 on one channel and 42 dB on the other. That's input to output with the test channel unloaded. Loaded, it's a little less. Almost identical result with a second input tube. Small sample size, but something inherent in tube construction?
I should note that initially the results were worse. I noticed that I had twisted the inputs together (dumb), thanks for the heads up.
Sheldon
Last edited:
I think that if you test with square waves, you'll find some interesting results, particularly in the polarity of the crosstalk.
Testing the phase shift of crosstalk versus frequency
I did a few more tests and will post the remainder of my results tomorrow; however, one test result I wish to share now. I checked to see what happens to the phase between the output of the triode under load (frequency generator input) and the triode under test for crosstalk (input loaded with a 47kohm resistor). This test was done on a 6SN7. I have posted the screen prints of the oscilloscope in this post. Note that the phase difference is 180 degrees at around 100 kHz. The phase difference is less as the test frequency is reduced and the phase difference is more as the test frequency is increased.
Can anybody explain why? I am unsure if this has any bearing on the ‘sound’ of the amplifier, it would be more an academic curiosity.
I did a few more tests and will post the remainder of my results tomorrow; however, one test result I wish to share now. I checked to see what happens to the phase between the output of the triode under load (frequency generator input) and the triode under test for crosstalk (input loaded with a 47kohm resistor). This test was done on a 6SN7. I have posted the screen prints of the oscilloscope in this post. Note that the phase difference is 180 degrees at around 100 kHz. The phase difference is less as the test frequency is reduced and the phase difference is more as the test frequency is increased.
Can anybody explain why? I am unsure if this has any bearing on the ‘sound’ of the amplifier, it would be more an academic curiosity.
It might be useful to be able to short the input to the second triode, to isolate the effects.
the ground is common for both amps as they are on the same circuit boards, therefore the ground is not an issue. do you meant tie the two inputs together? that would negate the test for crosstalk and what i understand regarding typical crosstalk measurements for stereo amplifiers, the test setup i have is correct.
No, he means short the grid of the second triode directly to ground instead of terminating it with a 47k resistor. Take measurement both with the 47k resistor and with the direct short, and plot them side by side. That way you can identify how much crosstalk is being induced via the grid, and how much elsewhere.
It allows you to figure out how much of the crosstalk couple across via the grid, and how much via other paths.
One would expect anode-to-anode capacitance to induce cross talk that is 90 degrees out of phase, but if some is also coupled to the grid and then amplified and inverted, it would add to the crosstalk at the anode, and hence lead to a change in the apparent phase shift measured at that point.
Last edited:
Some of them are. But as they are in reality distributed between each and every electrode, you never quite know how things will manifest in reality. Adding in the socket and wiring capacitance is also a bit of guesswork. Simulation is not to be trusted without real world data to confirm its plausibility.
This very interesting...or why the tubes sound like
My suggestion is to feed on grid 1 and 2 a different, but closed frequency. To have symmetric conditions.
Let's say 1. Grid 1 Khz & 2. Grid 1.1 Khz and record it with a FFT Analyzer to have the cross talk & the IM products..
just my 2 cents
Hp
Tested another amp, this time under typical operating conditions. Input tube is 6N7, which has the cathodes connected to a common tube pin. It's fixed bias, no cathode resistor. With 100R resistor shorting undriven input, and 8R on outputs, channel separation is greater than 50dB (2mA or less with 1V output on driven channel) at 20k Hz.
Sheldon
Sheldon
Test to determine makeup of phase shift in crosstalk
Following is the test description and results for determining the reason for the phase shift between the crosstalk and input signals.
Purpose of the test:
The purpose of the test to determine why there is difference in frequency phase shift between the input signal into one triode of a dual triode tube (Driver Triode) and the resultant crosstalk on the output of the other triode in a dual triode tube (Test Triode).
Hypothesis:
The hypothesis is that the two crosstalk signals to the grid circuit and the plate circuit of the Test Triode add together causing the shift in phase.
The crosstalk signal at the plate of the Test Triode caused by the crosstalk to the Test Triode grid from the Driver Triode plate should be in phase with the Driver Triode input signal as the signal is amplified by both triodes.
The crosstalk signal at the plate of the Test Triode caused by the crosstalk directly from the Driver Triode plate should be 180 degrees out of phase with the Driver Triode input signal as it would be amplified only by the Driver Triode.
Test setup:
Test was done on a 12ax7 tube. Plan was to do it on a 6SN7 but this test setup has been dismantled. Test frequencies were 1, 10, 100, and 1000 kHz. To measure the maximum signal transfer to the grid of the Test Triode, the Test Triode had an input of 1Mohm. To measure the signal to the plate of the Test Triode with no interference from the grid circuit, the grid of the Test Triode had an input of 0 ohm (shorted to ground).
Schematic of the test circuit and labels showing which triode is the Driver Triode and which is the Test Triode is attached.
Resultant waveforms shown on attached document.
In addition to the above tests, the supply voltage to the plate of the Test Triode was set to 0 volts and with either 0ohms or 1Mohm impedances on the Test Triode grid, similar results to having the Test Triode at 0 ohms and Vp=180VDC was recorded.
Discussion of the results:
The crosstalk signal received by the Test Triode grid is several times larger than the signal received by the Test Triode plate. The phase shift between the input signal and the crosstalk signal is very similar between the Test Triode input of 1Mohm (maximum crosstalk received by the Test Triode grid) and the input of 47kOhm (chosen normal Test Triode input impedance).
Conclusion:
The crosstalk signal received by the Test Triode is mostly from the grid circuit of the Test Triode. The crosstalk signal received by the Test Triode plate directly from the Driver Triode plate can be ignored.
The crosstalk signal phase shift between the input signal of the Driver Triode and the output of the Test Triode would be due to the transfer function equating these two signals. It appears that the two signals are in phase at about 10kHz.
The similar results between Test Triode grid at 0 ohms and Vp=180VDC and Test Triode grid at any input impedance and Vp=0V would be expected as both circuits would have the equivalent Thevenin circuit.
Following is the test description and results for determining the reason for the phase shift between the crosstalk and input signals.
Purpose of the test:
The purpose of the test to determine why there is difference in frequency phase shift between the input signal into one triode of a dual triode tube (Driver Triode) and the resultant crosstalk on the output of the other triode in a dual triode tube (Test Triode).
Hypothesis:
The hypothesis is that the two crosstalk signals to the grid circuit and the plate circuit of the Test Triode add together causing the shift in phase.
The crosstalk signal at the plate of the Test Triode caused by the crosstalk to the Test Triode grid from the Driver Triode plate should be in phase with the Driver Triode input signal as the signal is amplified by both triodes.
The crosstalk signal at the plate of the Test Triode caused by the crosstalk directly from the Driver Triode plate should be 180 degrees out of phase with the Driver Triode input signal as it would be amplified only by the Driver Triode.
Test setup:
Test was done on a 12ax7 tube. Plan was to do it on a 6SN7 but this test setup has been dismantled. Test frequencies were 1, 10, 100, and 1000 kHz. To measure the maximum signal transfer to the grid of the Test Triode, the Test Triode had an input of 1Mohm. To measure the signal to the plate of the Test Triode with no interference from the grid circuit, the grid of the Test Triode had an input of 0 ohm (shorted to ground).
Schematic of the test circuit and labels showing which triode is the Driver Triode and which is the Test Triode is attached.
Resultant waveforms shown on attached document.
In addition to the above tests, the supply voltage to the plate of the Test Triode was set to 0 volts and with either 0ohms or 1Mohm impedances on the Test Triode grid, similar results to having the Test Triode at 0 ohms and Vp=180VDC was recorded.
Discussion of the results:
The crosstalk signal received by the Test Triode grid is several times larger than the signal received by the Test Triode plate. The phase shift between the input signal and the crosstalk signal is very similar between the Test Triode input of 1Mohm (maximum crosstalk received by the Test Triode grid) and the input of 47kOhm (chosen normal Test Triode input impedance).
Conclusion:
The crosstalk signal received by the Test Triode is mostly from the grid circuit of the Test Triode. The crosstalk signal received by the Test Triode plate directly from the Driver Triode plate can be ignored.
The crosstalk signal phase shift between the input signal of the Driver Triode and the output of the Test Triode would be due to the transfer function equating these two signals. It appears that the two signals are in phase at about 10kHz.
The similar results between Test Triode grid at 0 ohms and Vp=180VDC and Test Triode grid at any input impedance and Vp=0V would be expected as both circuits would have the equivalent Thevenin circuit.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.