12BH7 Long Tailed Pair Clipping Early

Discussion in 'Shock Brother's DIY Amps' started by andrewRneumann, Nov 20, 2021.

  1. Ten Over

    Ten Over Tele-Holic

    Posts:
    943
    Joined:
    May 13, 2015
    Location:
    Central California
    Spurious result? I can put the black probe on the cathode and the red probe on the grid with no problem. You didn't try to measure to ground, did you?
    LTPI 12AT7 with meters.png
     
  2. andrewRneumann

    andrewRneumann Tele-Afflicted

    Posts:
    1,078
    Joined:
    Mar 22, 2020
    Location:
    Cincinnati, OH, USA
    I just measured it several ways. The number I assumed was actual bias was the voltage drop over the bias resistor. I got -9.3V over the bias resistor. If I measure across the grid leak resistor by itself, I get 0V, which makes sense. But if I measure from grid to cathode I get -7.8V, which I thought was suspect. Even if that's the more accurate measurement, it's not closer to cut-off.

    Grid to ground was way off. I never trusted that number. (Expected 48.6V, got 32.8V.)

    I played with the grid leaks and they seem to be having an unexpected affect on the bias. Then I checked the datasheet again and I think I misread what the maximum grid leak was and it's either 1M or 2.2M. I should have used 1M in this situation.

    Anyway, I used a substitution box and tried all sizes of grid leak resistors and found, crazy enough that a grid leak of only 110K minimized early cutoff clipping and still retained enough signal to fully drive the tube.

    Finally, I used the substitution box on the bias resistor and found that only about 900 ohms was appropriate to get grid current and cut-off clipping at approximately the same time.

    I did all of this on the scope--will have to play it tomorrow and see if it actually sounds good.
     
    sds1 likes this.
  3. andrewRneumann

    andrewRneumann Tele-Afflicted

    Posts:
    1,078
    Joined:
    Mar 22, 2020
    Location:
    Cincinnati, OH, USA
    Not your fault. I just come to understanding in a roundabout way usually. You've clearly done a lot more analysis on these than I have. Thanks for helping me understand this stuff at a deeper level.
     
  4. NTC

    NTC Tele-Holic

    Posts:
    909
    Joined:
    Mar 17, 2003
    Good thread here, gentlemen. I look forward to the results. It woukd be nice to see a revised schematic with new voltages after the changes (assuming it sounds good).
     
  5. David Barnett

    David Barnett Doctor of Teleocity Ad Free Member

    Age:
    65
    Posts:
    16,202
    Joined:
    Mar 17, 2003
    Location:
    The Far-Flung Isles of Langerhans
    The 1.6K cathode resistor seems large to me.
    Yeah that 1.6K would have been lean even in a lower-current 12AX7 circuit. With the 12BH7 I bet you could go lower than 900.
     
  6. andrewRneumann

    andrewRneumann Tele-Afflicted

    Posts:
    1,078
    Joined:
    Mar 22, 2020
    Location:
    Cincinnati, OH, USA
    I planned on 6mA of current and wanted a 10V drop for the bias. It came out to 1K6. I don't know why 900Ω comes out better just on theory alone.
     
  7. andrewRneumann

    andrewRneumann Tele-Afflicted

    Posts:
    1,078
    Joined:
    Mar 22, 2020
    Location:
    Cincinnati, OH, USA
    Latest design and voltages. I changed the grid leaks and the bias resistor. Measured voltages to chassis shown in blue.

    upload_2021-11-22_12-16-10.png

    Here are the maximum clean traces:

    IMG_0369.jpg

    Input headroom is only 30Vpp now, but output headroom is up to 196Vpp.

    Here are the overdrive traces:
    IMG_0370.jpg

    Looks a lot more symmetrical. I'm sure the troughs would be more flat if I wasn't driving this with a CF.

    So, I'm pretty happy with it. Reality is, that I don't understand why my original design ended up so far off. A bias of -6.2V seems too warm to me to be center bias, but it seems to work out on the scope.

    I did play guitar through it this way and it seemed fine.
     
    sds1 and D'tar like this.
  8. 2L man

    2L man Tele-Holic

    Age:
    61
    Posts:
    889
    Joined:
    Nov 23, 2020
    Location:
    Finland
    I think on first the bias was colder and outout did clip when anode voltage "collided" to Vss.

    Connect oscilloscope to anode and use DC-mode and see if that cause that clipping with this bias as well?
     
  9. David Barnett

    David Barnett Doctor of Teleocity Ad Free Member

    Age:
    65
    Posts:
    16,202
    Joined:
    Mar 17, 2003
    Location:
    The Far-Flung Isles of Langerhans
    What would happen if your long tail was "longer"? Like change the 8.2K to 22K-ish?
     
  10. andrewRneumann

    andrewRneumann Tele-Afflicted

    Posts:
    1,078
    Joined:
    Mar 22, 2020
    Location:
    Cincinnati, OH, USA
    If I had more HT I could consider a longer tail. I need as much output swing out of this as I can get with HT at 310V. If I go with a tail that large I’ll be looking at something like 130Vpp instead of the 200Vpp I have now.


    That’s the reason I selected the 12BH7 to begin with. So I have to compromise and deal with some asymmetry in the gain to keep that output swing relatively high. I’m happy with 200Vpp and don’t want to lose any from there.
     
  11. Tom Kamphuys

    Tom Kamphuys Tele-Holic

    Age:
    42
    Posts:
    870
    Joined:
    Sep 18, 2018
    Location:
    The Netherlands
    Last edited: Nov 24, 2021
    andrewRneumann likes this.
  12. andrewRneumann

    andrewRneumann Tele-Afflicted

    Posts:
    1,078
    Joined:
    Mar 22, 2020
    Location:
    Cincinnati, OH, USA
    +1

    Thank you Tom! I think it made a difference whether I probed on the HT side of the coupling cap, or the grounded side of the coupling cap. Does it make a difference on LTSpice?
     
  13. andrewRneumann

    andrewRneumann Tele-Afflicted

    Posts:
    1,078
    Joined:
    Mar 22, 2020
    Location:
    Cincinnati, OH, USA
    So I had an *aha* moment tonight fooling around with this LTP. I now believe the reason for the center bias calculations being off is because I’m driving this with a cathode follower. The cathode follower is able to drive the grid positive and extend the input headroom on the positive side. This made it look like the bias was colder than it actually was. And this is why a “warm” bias is actually more center bias in this circuit.

    Lesson—if you are driving a tube with something with a low source impedance like a cathode follower, center bias will be shifted to the warm side.

    As far as the early clipping goes, my hypothesis is that too much grid leak resistance was leading to an accumulation of electrons on the grids which was shifting the bias cold. Sort of like grid leak bias on top of cathode bias. This is still tentative—I want to do more experiments with different grid leak values to determine what is the optimum size. This probably is interrelated with the cathode follower driver.

    As for the sloping overdrive segments. I think it was indeed the coupling caps charging. I’m only using 1nF caps right now, so they charge quickly—thus the steep slope. More experimentation needed to confirm this.

    That’s all for now.
     
    sds1 likes this.
  14. 2L man

    2L man Tele-Holic

    Age:
    61
    Posts:
    889
    Joined:
    Nov 23, 2020
    Location:
    Finland
    Using a DC-mode on oscilloscope it is easier to see where signal "travels" between B- (reference/ground) and B+ and signal clips, when it can not climp "higher" than B+3 in phase inverter.
     
    sds1 likes this.
  15. Ten Over

    Ten Over Tele-Holic

    Posts:
    943
    Joined:
    May 13, 2015
    Location:
    Central California
    Yeah, bias shift, but the electrons accumulate on the 22nF input capacitor giving its grid side an overall more negative charge which makes the bias colder.

    The grid acts like a diode when grid current occurs. This makes the input circuit a half-wave rectified power supply. The average charge on the capacitor depends on the charging time and the discharging time when the AC is held constant. These times depend on the charging path and the discharging path. The grid stopper and the output impedance of the CF are in the charging path and have a relatively small value of 3.3k combined. The discharge path is the grid leak with a relatively huge value.

    Reducing the size of the discharge path by reducing the value of the grid leak resistor will allow the capacitor to discharge faster which will reduce the average charge on the input capacitor. That means that the bias shift becomes less when you reduce the grid leak.

    Driving the LTP with a 12AX7 gain stage with an output impedance of around 40K instead of an output impedance of 0.6K from the CF will increase the charging time and decrease the average charge on the input capacitor. So the CF makes the unwanted bias shift even worse.
     
  16. Ten Over

    Ten Over Tele-Holic

    Posts:
    943
    Joined:
    May 13, 2015
    Location:
    Central California
    The input signal doesn't exhibit any appreciable clipping because there isn't enough resistance to create enough voltage drop to notice. Change the grid stopper to 220K and you will see some nasty clipping.
     
    sds1 likes this.
IMPORTANT: Treat everyone here with respect, no matter how difficult!
No sex, drug, political, religion or hate discussion permitted here.