Ad Widget

Collapse

Announcement

Collapse
No announcement yet.

Best way to reduce noise when reducing gain, triode pre-amp stage?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Best way to reduce noise when reducing gain, triode pre-amp stage?

    Hi all,

    This is a very general question, but I am interested in any feedback, because I have this situation -

    I am building a reverb recovery section using the 12AX7 side of a 12DW7 (actually a JJ ECC832) and feeding the signal into the usually grounded side of a 12AX7 LTPI.

    The recovery stage is set up as follows -
    2.25K reverb tank output impedance
    22k grid stopper
    1M grid leak
    4.7n cap from grid to cathode (high frequency shunt)
    100K plate resistor
    47n coupling cap
    1M pot (reverb control), one side of track to coupling cap output, other side of track grounded, wiper to LTPI input coupling cap
    LED cathode biasing

    All resistors are metal film. It works very well, except that it gives quite a bit of noise when the reverb is turned up to a high level. The noise is general frequency, but with a high frequency roll-off due to the high frequency shunt cap. It is not hum, more white noise.
    Luckily the noise only becomes easily audible when the reverb effect is already strong enough, at only ~ 2 o'clock on the reverb control.

    So the gain of the stage is more than it needs to be, and I want to reduce the gain from ~ 55 where it is now, down to ~25.

    But what is the best way to reduce the gain in order to give the greatest reduction in noise?????

    Options -
    Reduce grid leak resistance at stage input (probably worst of these options for S/N ratio)
    Reduce plate resistance (lowers voltage gain but increases current through plate resistor and tube, therefore higher noise?)
    Use local NFB on stage (strong contender)
    Reduce value of reverb control pot to AC load stage more heavily (how advisable is it to have highly differing DC and AC load lines for stage performance?)
    Use a resistor before reverb pot to form voltage divider (470K resistor, 500K pot, ~ same total load resistance as 1M pot used at present)

    Thanks for any advice

  • #2
    All of the options should have the same effect on noise level, if you assume that the noise source is at the tube's grid.

    But the NFB will also reduce distortion, and won't reduce the maximum output level along with the gain.
    "Enzo, I see that you replied parasitic oscillations. Is that a hypothesis? Or is that your amazing metal band I should check out?"

    Comment


    • #3
      I wouldn't expect much noise, being as there are no other gain stages between it and the power amp. Have you tried a different tube?

      '22k grid stopper'
      Series resistance will add noise, should be able to do without, or at least reduce the value.

      '4.7n cap from grid to cathode (high frequency shunt)'
      I still haven't got around to LED biasing so don't know the pitfalls, but I would take that cap to ground rather than the cathode.

      And I'd try a 100k pot, rather than 1M, to lose a little gain and stop the impedance at the phase splitter input varying so much.
      My band:- http://www.youtube.com/user/RedwingBand

      Comment


      • #4
        Thanks for the help Steve and Pete.
        I think that you're both right, the noise is mostly coming from the source, and not the stage.
        I actually tried most of the options, testing the 1KHz gain on an oscilloscope, and using my ears to assess the noise component. In all cases the noise seemed proportional to the gain!

        The step I am up to now is using a 12AU7, it gives enough gain on the recovery side to get a deep enough reverb effect. The only ones I have are old used Brimars and Mullards from an old tube computer, and I don't think that they are spec'd for low noise, so there is a bit of hiss present at the output. I have ordered a new JJ ECC82 to sub in to see if that solves the hiss, if so then the circuit is fixed! I am currently using a 47K Rp and a 500K reverb pot.

        Thanks again

        Comment


        • #5
          Oh, just out of the blue: try blocking the ambient light from the LED for a moment. ANy difference? Piece of electrical tape or just throw a rag over it for a moment.

          Isolate the problem. tack a wire from the grid to ground right at the tube. Does that kill the noise or does it remain?
          Education is what you're left with after you have forgotten what you have learned.

          Comment


          • #6
            The majority of amplifier noise is going to be generated by the first gain stage... and no, blocking ambient light would have no effect....

            Also, the grid leak is too high. Most builders use 1 meg because the volume pots in guitars are either 250k or 500k.
            The proper grid leak on a reverb recovery stage ; in the range of ; 220K...
            ______________________________________
            Gary Moore
            Moore Amplifiication
            mooreamps@hotmail.com

            Comment


            • #7
              I think it is understood the first stage would be the source, since there is only the one stage.
              Education is what you're left with after you have forgotten what you have learned.

              Comment


              • #8
                Originally posted by jimboyogi View Post

                I am building a reverb recovery section using the 12AX7 side of a 12DW7 (actually a JJ ECC832) and feeding the signal into the usually grounded side of a 12AX7 LTPI.
                .....
                ______________________________________
                Gary Moore
                Moore Amplifiication
                mooreamps@hotmail.com

                Comment


                • #9
                  Hi,

                  Enzo, I'm sorry but I don't understand the reference to blocking the ambient light from the LED?

                  mooreamps, I see your point about the grid leak resistance being too high, I was just treating it as if it were a guitar input, but the impedance of the tank is very different to that of a guitar!

                  To select the best grid leak value, am I correct in thinking that the grid leak forms a voltage divider with the output impedance of the reverb tank coil, in this case ~2.6K at 1KHz? If this is correct then the grid leak wants to be say 10 to 20 times the coil impedance, so 27K to 47K should be high enough?

                  Comment


                  • #10
                    What Gary said about the grid leak. Change it to 220k for a start and see how much that changes the white noise. It shouldn't make much difference to gain and BW with the level of impedance bridging (from the pan output source to the grid load) that you would still have (AFAICT impedance bridging is more important than impedance matching for keeping good bandwidth - since we are dealing primarily with voltage amplification at that point. Bite me if I'm wrong).

                    The 22k grid stopper could even go higher, 33k (or 47k or maybe even up to 68k?) to further reduce any surplus RF interference from the reverb cables/pan, without losing bandwidth. This resistor functions like an input stage series input resistance (as opposed to a grid stopper) on the recovery stage.

                    Then see if you can get away with removing the 4n7 shunt cap - because in doing that you might be able to improve the wetness factor (or not).

                    If you want to reduce the gain more, you could also try a (say between 220k and 1M) tapering resistor from the 1M pot wiper to the pot ground. This will change the pot taper in a way that changes the way you dial in the verb, as well as reducing the gain by increasing the overall load on the recovery stage. JM2CW
                    Last edited by tubeswell; 12-17-2010, 03:50 PM.
                    Building a better world (one tube amp at a time)

                    "I have never had to invoke a formula to fight oscillation in a guitar amp."- Enzo

                    Comment


                    • #11
                      If you're sure that the noise is coming from before the stage (for example, by grounding the input as suggested), you can use a voltage divider at the input to lower the input signal level and provide grid leak at the same time. Try 220k/220k, which will provide 6dB of attenuation.

                      Comment


                      • #12
                        Thanks for the extra ideas.

                        Enzo I tried grounding the grid, all noise disapeared. So the noise (or most of it) is coming from the source. I just got up, but sometime today I'll wire up a new shielded cable to the tank and see if that helps.

                        Other developements. I took on board the grid leak idea, and lowered it's value to 22K, which is 8.5 times the tanks output impedance (~2.6K at 1KHz). This in theory should reduce resistor noise.

                        But another advantage to lowering the grid leak is the effect on input impedance. The primary reason I have the 4n7 treble shunt cap is that the frequency response is too treble heavy, and this cap works to eq the treble back to a better level. Now the ouput transducer on the reverb unit is inductive, and it's output impedance rises with increasing frequency. So i'm thinking I can take advantage of this with a lower value of grid leak, where higher frequency higher impedance signals are attenuated more by the voltage divider formed between tank output and tube input impedance. Hopefully do away with the treble shunt cap this way. Might also need to increase the input grid stopper as tubeswell suggests.
                        I should be able to do some testing today.

                        Comment


                        • #13
                          Sorry for being such a troglodite, but *why* Led biasing?
                          I see no special advantage, and it has certainly created some social trouble in the last days.
                          1) I would use an old reliable time tested combat proven 1k5 resistor to ground, with or without a 10uF cap in parallel depending on gain needed.
                          2) Instead of wasting it on the grid side, I would pull the 4n7F and use a 1nF or 1n5F plate to ground .
                          3) If anything, I would pull the series grid resistor and wire a 470 to 2200 pF ceramic across the input 1M (or 220K) resistor, to create a resonant peak together with the pickup coil inductance.
                          Adjust it using the patented Ear-O-Meter.
                          Just my 2 cents.
                          Juan Manuel Fahey

                          Comment


                          • #14
                            A couple thoughts. We can get all mathy and figure figure figure, but really, look at a few fender schematics. Most have a 220k grid resistor, no grid stopper. Fender reverbs sound fine, and don't as a rule have a hiss problem. And looking at the fender reissue stand-alone reverb unit, it has a 1M grid resistor, and is not known for hissy either.

                            You have that grid stopper to prevent RF problems, but were we even having RF problems? I don;t see such grid stoppers on reverb circuits in most amps. Try getting rid of the grid stopper.


                            And why did I want to cover the LED? LEDs of course make light when current flows through them. But semiconductor junctions are sensitive to light. And LEDs can also act as a photosensor. it was conceivable if you had a bright flourescent fixture over the bench, the LED could pick it up and add some noise. Now I am sure someone might hop in and explain why this is totally impossible, but I have run into light sensitive circuits before, and since the test literally takes two seconds to cover the LED for a moment, why not try it?
                            Education is what you're left with after you have forgotten what you have learned.

                            Comment


                            • #15
                              Originally posted by Enzo View Post
                              And why did I want to cover the LED? LEDs of course make light when current flows through them. But semiconductor junctions are sensitive to light. And LEDs can also act as a photosensor. it was conceivable if you had a bright flourescent fixture over the bench, the LED could pick it up and add some noise. Now I am sure someone might hop in and explain why this is totally impossible, but I have run into light sensitive circuits before, and since the test literally takes two seconds to cover the LED for a moment, why not try it?
                              Because when you play a gig on stage ; or do a demo cut in the studio ; are you really going to pull the chassis out of the box and stand a big bright fluorescent lamp on the top of it ??

                              -g
                              ______________________________________
                              Gary Moore
                              Moore Amplifiication
                              mooreamps@hotmail.com

                              Comment

                              Working...
                              X