Screen current - harder to understand than I thought

Discussion in 'Tube Audio' started by kward, Mar 20, 2017.

  1. kward

    kward AK Subscriber Subscriber

    Messages:
    3,050
    Location:
    Utah
    I'm building a new amp with a 6L6 push-pull output stage. The output stage parameters are:
    • Push pull, into a 6.6K primary load (plate-to-plate)
    • 8 ohm dummy load resistor on 8 ohm secondary winding
    • 300V screens (regulated through my bench power supply)
    • Fixed bias (but with 10 ohm current sense resistors in the cathode circuit)
    • -30 VDC grid bias
    • Plate voltage is 440V at full power, and 460V at idle.
    • Brand new Tung Sol 6L6-GC STR tubes.
    I have a installed a 150 ohm resistor on each screen so as to be able to sense screen current, plus for screen stability/protection.

    I drive the amp to just under clipping (35 watts) from my signal generator with a 1 KHz sine wave. At -30V grid bias, I *should* see right around a 30V peak drive signal on each of the control grids at max power output, which indeed I do as verified on my scope. So in other words everything is working as expected.

    Now the question:
    I measure 8.5 mA DC screen current draw across each 150 ohm screen resistor and 8.33 mA AC current draw across each screen resistor. Current draw in both cases is calculated by measuring either the AC or DC voltage drop across the resistor, and then dividing by the resistor size of 150 Ohms. AC voltage drop is measured from my digital volt meter, which approximates RMS readings.

    Double checking the screen current draw with the 6L6-GC data sheet, and looking at the "average transfer characteristics" graph where it plots grid #1 voltage on the X axis and screen current on the Y axis, and with a 400V plate voltage (a little lower than I'm using at 440V, but within the ball park), I see that at 0V grid, I should be getting right around 17 mA or 18 mA screen current draw per tube for a 300V screen voltage. But in actuality I am getting pretty much half that screen current draw for either the DC current or the AC current.

    See this graph from the 1959 GE 6L6-GC data sheet. The red circled area is what I am focused on:

    6L6 transfer characteristics.png

    Here you can see that the graph says I should be getting about 18 mA screen current per tube at 0V grid for 300V screen and 400V plate.

    I am trying to figure out why the graph shows roughly double the screen current of what I am actually measuring. One thought as to why is perhaps the graph above is assuming DC conditions, whereas my test is measuring AC conditions.

    I'm not sure if this helps decipher things, but here is the voltage wave as seen on the scope under full power conditions (with AC coupling on the scope probe), with a 10X scope probe placed on the screen pin of one of the output tubes. Vertical scale is 0.1V per division (thus in reality 1V per division).

    Screen voltage 1KHz drive signal max power.JPG

    Anyway, I am trying to make sense of the transfer characteristics graph with my actual measured screen current. At the moment they don't seem to coincide. I am wondering what is wrong with my mental model of screen current?
     
    Last edited: Mar 20, 2017
  2. thorpej

    thorpej Super Member

    Messages:
    1,452
    Location:
    San Francisco, CA
    Couple of things:
    • You're actually not that far off from the "Maximum-Signal Screen Current" (for two tubes) for the data sheet's 450V plate / 400V screen example operating point, so what you're measuring seems correct (or at least in the ballpark).
    • How much current the screen soaks up depends on where you load line crosses the 0V-grid plate curve, right? If it's below the knee, you're going to see the screen current go WAY up very quickly as you cross the 0V-grid line, whereas if it crosses through or above the knee, the plate is still going to be sinking most of the current.
    This is where having one's own curve tracer to make a set of plate curve graphs specifically for the chosen operating point would be super-handy.
     
  3. dcgillespie

    dcgillespie Fisher SA-100 Clone Subscriber

    Messages:
    8,942
    Location:
    Ball Ground, GA
    Remember that the AC screen current is actually riding on top of the DC screen current, so their effects are cumulative. They are separate elements that together make up the total current drawn by the screen.

    Dave
     
    Last edited: Mar 20, 2017
  4. primosounds

    primosounds AK Subscriber Subscriber

    Messages:
    1,471
    In the last graph from the same GE tube manual entitled Operation Characteristics, their output trans has a load resistance of 5600 ohms. Maybe your own opt. which is 6600 ohms, would account for some of the differences you measured since the 2 opts are 1000 ohms apart.
     
    Last edited: Mar 20, 2017
  5. kward

    kward AK Subscriber Subscriber

    Messages:
    3,050
    Location:
    Utah
    Dave--Ah yes, how simplistic now that it was said out loud. The answer was right there in front of me. 8.5 mA DC screen current + 8.3 mA AC screen curernt = 16.8 mA total current, which is nearly exactly what the data sheet says I should be getting. Sheesh. Sometimes the easiest things are the hardest.

    Thorpe--the load line (class b) should be crossing at the center of the knee of the 0V grid curve, but I actually don't know how to measure that on the actual build. What I do know is the distortion analyzer shows lowest distortion output at 298V screen, -30.2V grid, and at whatever plate voltage drop the load dropped it to. So wherever it actually crosses, it's the point that delivers lowest overall THD.

    Primo--yes it seems to match closely now the data sheet values whether viewed from the transfer characteristics graph or the tabular section of the sheet.

    Any hows, thank you to all for your quick replies.

    P.S. This is for an EFB installation, where (for the first time for me), I am attempting to find those golden EFB parameters on my own, without copying anyone else's schematic. And my bench power supply that I finished building earlier this year is already being put to some very good use!
     
    thorpej likes this.
  6. thorpej

    thorpej Super Member

    Messages:
    1,452
    Location:
    San Francisco, CA
    Cool. The lowest THD is actually going to be slightly above the knee, because it avoids the non-linearity of the knee itself. You sacrifice a small amount of power to achieve this (because you don't allow the output to swing into the knee). If you draw the load line where it passes though the middle of the knee at Vg2=300, and then set your actual Vg2=298, then the load line would pass just above the knee.
     
  7. kward

    kward AK Subscriber Subscriber

    Messages:
    3,050
    Location:
    Utah
    Makes sense. So avoiding the knee portion of the 0V curve is important. So if the load line passed slightly below the knee or slightly above the knee, that is better (distortion wise) than passing smack dab through the center of the knee?

    Wait, I think I answered my own question from simply looking at the data sheet graph. Looks like distortion wise it is similar either way (load line passing above or below the knee) but screen current has the potential of going up substantially if the load line is crosses below the knee.

    Also, since I have the test bench all setup for this at the moment, I guess I could just try it and see what the amp meter and distortion meters report.
     
    Last edited: Mar 20, 2017
  8. thorpej

    thorpej Super Member

    Messages:
    1,452
    Location:
    San Francisco, CA
    Passing below the knee is going to lead to even more distortion, and worse, possible destruction of the screen because the screen starts quickly soaking up all the current as the plate current drops off rather rapidly. Guitar amps put large-ish screen stoppers in the circuit in designs where they're going for output tube distortion in order to squash down the screen voltage as the screen current goes up. It lends a compressing effect to the sound.

    Passing smack dab through the center of the knee is the textbook way to maximize output power, but at the cost of a bit of distortion a maximum output.
     
  9. kward

    kward AK Subscriber Subscriber

    Messages:
    3,050
    Location:
    Utah
    Well that's interesting--by itself (changing nothing else) adjusting the screen voltage higher (by +10V), does indeed produce more distortion and higher screen current (and with a slight increase in max power). But then at that given screen voltage, if I also decrease (make more negative) the bias voltage, the distortion goes back down again to the null I found previously, while the screen current is incrementally higher also.

    Now, I'm not sure what is happening to the load line when I adjust the bias voltage more negative (and changing nothing else). I thought the bias voltage only came into play during the class A portion of the load line, not in the class B portion, but it's clearly the class B load line that would be passing through the 0V grid curve.

    What I really want is 35 watts output power, so that is what is driving where I set the screen voltage, and then adjust the grid bias voltage for lowest distortion given that particular screen voltage selected. This seems to produce very good distortion results (at 1 KHz anyway).

    Anyway, all this is new interesting territory that is helping me gain new appreciation and understanding now that I have a way to adjust these various dials and watch what happens in real time.
     
    Last edited: Mar 20, 2017
  10. drtool

    drtool It might get loud In Houston Subscriber

    Messages:
    2,074
    Location:
    Houston Texas
    Fascinating thread, it is helping me.
     
  11. BinaryMike

    BinaryMike Pelagic EE Subscriber

    My collection of high-voltage power supplies now stands at five, for this very reason: accelerated learning. Do be careful to turn them on and off in the correct sequence!
     
  12. thorpej

    thorpej Super Member

    Messages:
    1,452
    Location:
    San Francisco, CA
    It makes sense that the screen current would stay roughly-ish the same (although I am a little surprised that it went up a bit). I am going to make a guess that the extra (negative) bias is keeping your input signal away from the 0V line[*] (you said you changed nothing else, so I assume that includes the amount of drive signal applied to the amp); when you increase the screen voltage, the 0V line moves higher up the graph, as did your original bias point... When you increase the screen voltage like that, to take full advantage of it you're going to want to either use use a different output load (specifically, a lower impedance) that gives you a "steeper" class B load line that gets you back through or above the knee at the same plate voltage, or a higher plate voltage that slides the class B line to the right to get you above the knee with the same slope.

    [*] IIRC, most EL34 amps are biased so that the grid never actually hits 0V when driven to full rated power. 6L6s, however, were designed to be able to draw grid current, so it's possible to drive them right to the 0V line (and beyond into AB2, if your driver stage can manage it).
     
    Last edited: Mar 21, 2017
  13. kward

    kward AK Subscriber Subscriber

    Messages:
    3,050
    Location:
    Utah
    The way I determined that power went up or down given a slightly different screen voltage was to adjust the drive voltage ever so minutely to just a smidge under max power. I determined max power based on the needle on the distortion meter--when it shot up, that's a bit too high drive voltage so I backed off just a titch. So in fact I did adjust drive voltage ever so vanishingly small given the different screen voltages. Interesting that distortion stays really low right up until it starts to clip--it does not rise gradually with increased power output (at 1KHz anyway). It's like it hits a wall at max power where distortion is flat and then shoots up dramatically.

    298V screen produced 32 watts output (with -30V grid), while 302V screen produced 35 watts output (with -31V grid). I won't be able to distinguish 32 vs. 35 watts output, so this whole exercise is an interesting academic experiment to learn how things change when tweaking screen and bias voltage. Both cases produced nearly identical THD at 0.075% to 0.08%. (single channel driven, for now, until I get the EFB board installed).
     
  14. kward

    kward AK Subscriber Subscriber

    Messages:
    3,050
    Location:
    Utah
    I've already identified a need for a second power supply myself! And man do I wish I would have put a stand by switch on the HV line! It's killing me to need to pull the wire out of the banana jack over and over and over. If I build a second one, I'm going to put analog meters on it though.
     
  15. thorpej

    thorpej Super Member

    Messages:
    1,452
    Location:
    San Francisco, CA
    That's really cool! It's like you can see the effects in the real world of the grid signal passing the point where the plate current falls off the cliff. And it also makes sense if you look at the plate curves ... you're keeping the load line in that nicely spaced as-linear-as-a-tube-gets region.

    My plan has always been to build a smaller, self-contained supply that takes B+ from the regulated bench supply and regulates down for screen. Simple enough to build on a small circuit board, and doesn't take up 9U of rack / bench space!
     
  16. thorpej

    thorpej Super Member

    Messages:
    1,452
    Location:
    San Francisco, CA
    I'm still guessing that you're still not hitting exactly 0V when you clip at the higher screen voltage. A test instrument that latches the maximum voltage seen at that test point would be really handy here!
     
  17. dcgillespie

    dcgillespie Fisher SA-100 Clone Subscriber

    Messages:
    8,942
    Location:
    Ball Ground, GA
    What you're likely witnessing is the results of using regulated power supplies, that will in fact allow distortion to remain quite low until clipping is reached. If you're primarily used to the "real world", where power supply voltages droop with increasing power output and tube element voltages fall away well out of lockstep with each other from those established under quiescent conditions (the very thing that EFB(tm) addresses), then no doubt seeing the distortion performance that regulated power supplies can produce is quite an eye opener!

    Dave
     

Share This Page