Need help re-designing power supply for high-end amplifier

Discussion in 'The Projects Forum' started by earth1, Mar 25, 2016.

  1. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0
    Hello everyone..

    I recently acquired two mono-block tube amplifiers. When these were manufactured, they were purported to be very high-end amplifiers. However there are some issues with line noise (120cps full-wave rectifier hum) which ruins any other benefits that this amp might have had for being audiophile quality, and I am trying to address this... I have replaced all electrolytics and so we know it's not a bad cap, so we won't waste our time on that theory..

    I see three major factors that seem to be at the root cause, each factor compounding the other.
    First, the manufacturer should have used a good CLC pi filter, but instead used a single 1650uf capacitor @800v (actually two 3300uf caps in series, rated @ 400v each, with 220k bleeder resistors) which serves as plate supply for the 807 output tubes.
    That B+ rail feeds into additional resistors to drop voltage for feeding the 807 screen grids, and has 330uf/400v caps on the low side of the resistor, which theoretically should further smooth this lower voltage rail, but it's the lack of a pi filter preceeding the 807 B+ supply that I believe to be our first contributing factor.

    The second contributing factor is that this amp has four solid state rectifiers (discrete 1n4007 junction diodes for B+, a separate rectifier for bias supply, a separate transformer and rectifier for 12AT7 B+ supply, and then a large square rectifier for 6.3v filiment supply.) .. These all share the same ground rail, and none of them have snubbers, nor are they followed by CLC pi filters (again just single caps with nothing for the ripple voltage to fall across).

    The third contributing factor and the reason I am reaching out to this forum for help, is that someone who owned these amps, at some point removed the original supply transformers and substituted them with much heavier-duty supply transformers with a lot more available current, which means that they have a considerably lower secondary winding impedance, and this will factor into the re-design of the power supply filters and I want your input.

    My multi-pronged attack for eliminating the hum, is 1) Place 0.22ohm 1 watt resistors in series with each 1n4007, and then install 0.01uf snubbers across each diode/resistor pair.
    2) Remove the two 3300uf/400v series connected electrolytics, and replace them with two 1500uf/600v electrolytics, but this time create a CLC pi filter with one of the 1500uf/600v caps preceeding the inductor and the other 1500uf/600v cap following the inductor.
    I did some calculations and the needed value I came up with for my inductor is 8.5mh. However I looked at some on-line calculators and when I plugged in the variable, I see that the impedance of the supply transformer is a fairly critical factor, or so it would seem.
    3) I am contemplating the removal of the rectifier from the filiment supply and connect unrectified filiment secondary directly to the tube filiments, as this will probably present less noise than the unsurpressed ringing of the diode junctions.

    With these details in mind, I would like your input for choosing the inductor value, including if necessary, figuring the impedance of the B+ supply secondary of the supply transformer...
    Also if anyone thinks I may be missing the boat somewhere for cause/solution to this line-noise problem, please speak up. I want the help.


    Regards,
    Jim Butler
     
  2. KeepItSimpleStupid

    Well-Known Member

    Mar 4, 2014
    1,147
    204
  3. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,292
    1,255
    Do you have a schematic?
    Is it a push pull or single ended amp.
    Assuming you can actually hear the hum have you tried anything?
    Do you have a scope?
     
  4. SLK001

    Well-Known Member

    Nov 29, 2011
    820
    229
    When these amps were made, there would not have been any problem with 120Hz like you are having. I don't believe that you need to rebuild your supplies - as built they should amply reject any ripple present. I suspect a bad part or parts. If you must add your CLC filters, find out the maximum current supplied and put the largest caps at the max voltage and the largest L that will pass twice the max current. Don't worry about the impedance - the desired result is DC, so all filter calculations are meaningless.
     
  5. RichardO

    Well-Known Member

    May 4, 2013
    1,235
    384
    I once helped a friend fix an amp that had bad ac hum. I think he built it from a kit.

    The hum was caused by the charging current for the filter caps flowing through the low level signal ground. As I remember... The fix was to cut the signal ground from the power ground and reconnect to the ground side of the filter capacitors.

    I still find it amazing that the amp could ship with such an obvious fault in the PCB design.
     
  6. blocco a spirale

    AAC Fanatic!

    Jun 18, 2008
    1,440
    368
    As above, before doing anything else draw out the earthing arrangement, It may be as simple as moving a single wire.

    I assume the hum is present when the audio input connector is shorted out?
     
  7. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0
    Thanks for the response.. I'll take a look at this..
     
  8. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
    One of the earlier questions was, Do you have a scope. You need to find out where the noise is coming from, instead of shotgunning parts.

    Oh yea, Welcome to the forum.
     
  9. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
  10. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0
    Hi..
    This is a push/pull design.
    Connected to each leg of the primary of the output transformer, there are two 807 tubes in parallel for a total of 4 807s in each mono-block.
    Yes, I do have a scope. (Tektronix 2235 100mhz).

    Yes, I have tried a number of things, some more effective than others..

    One observation that I made was that when grounding the input of this amplifier to various places along the grounding rail, I get more quieting in some places than others; however, moving the shielding or ground connection of the input coax cable, does not have any impact on the hum; the upshot of this, is that we know two things: 1) the ground buss and it's distance from a 0ohm impedance node or common grounding point is a factor. 2) this hum is being picked up at the front-end of the amplifier, and is being amplified from stage to stage, thus producing enough amplitude at the output to be an annoyance.
    Corroborating this theory, is the fact that this amplifier has an inverse-feedback switch which simply increases/decreases the amount of inverse feedback. So when I change the position of this switch, the hum becomes somewhat quieter, because of the degenerative feedback which dampens the amplitude of the hum.
    From this angle, there are a couple of other things I haven't tried that I may, including moving cathode resistor and other ground connections to the front-end of the amp, to the ground point of the 1650uf 800v capacitor, which is essentially the main ground point of this amp. This can (and should) be explored. One thing I already did try, was adding thick copper ground buss along the thin circuit traces used for grounding, thinking that this might quiet the problem by reducing the buss's impedance and hence the voltage drop across the ground buss. This however did not resolve the hum problem.

    With the foregoing observations, my mind goes in two different directions for root cause and best solution:
    My first inclination is to find a proper spot to ground the frontal-stages of this amplifier so that it finds no hum to be amplified.
    It then quickly occurs to me that the existing ground arrangements of this amp are as the factory engineers designed it, and that the manufacturer wouldn't release a production version of their amp to the audiophile market, knowing it's plagued with um problems. So that leads me to believe that the front end of this amplifier worked noiselessly when it was built and sold. However, I see no evidence that the front-end of this amp was ever modified, and I have a good eye for that sort of thing.
    So then my mind goes in another direction, and very pragmatically observes:
    There would be no line noise present in the voltage rails to be amplified, if the power supply circuit were effectively filtering out the line noise to begin with.. With this premace in mind, the lack of a proper PI filter becomes the elephant in the room.

    I did replace all electrolytics and this did have a mildly positive impact on the problem. Indeed it helped a little, but not to the degree needed. Among the many theories and potential solutions I have going around in my head, I wanted to at least place an inductor before the 1650uf cap, so I dug around in my parts, and found a transformer that had been used as a supply transformer for a Samick digital piano.
    I reasoned that if I leave the secondary (low-voltage) connections completely disconnected, the primary should have enough inductance for an effective test, although I do not know the specific inductance that the primary of this device presents. Anyway, I took some measurements before I got started including oscilloscope readings. I measured the forward-voltage ripple at the positive side of the 1650uf 800v cap, and it showed a sharp saw-tooth (as expected coming right off of a solid state rectifier).I disconnected the connection from the positive side of the HV rectifier, and ran that to one side of my make-shift inductor. The other side of the inductor connected to the 1650uf/800v capacitor, which also connected to the rest of the amp (i.e. 807 plate supply, and successive filtering for preceeding stages, etc.). This brought the amplitude of the hum down by more than 2/3 as observed on my oscilloscope, and also smoothed out those sharp peaks in the sawtooth, to form something closer to a sign-wave..
    With this, the amp was certainly better, but would no doubt benefit from having two 1650uf caps - one at the front of the inductor and one at the back of the inductor.

    Regards,
    Jim
     
  11. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0
    Yes, I too am inclined to believe that these amps should have been quiet enough when then were made, although I do not find any parts that I can identify as "bad". Resistors show their proper values, electrolytics have been replaced. Tube sockets and pins have been treated with Deoxit D100 (my favorite ;-) ), plate caps have been treated, and areas between circuit traces have been cleaned and are free of solder flux and other contaminates that could cause leakage currents.. I haven't changed out any feedthrough capacitor, but I wouldn't think that would be a contributor.

    If I go through with adding a full-fledged PI filter to this amp's power supply, I'd like some help with choosing the proper inductance. I've made my own calculations, but a second opinion can never hurt. If you can help me with the calculation, that would be appreciated.
    As stated earlier, the HV secondary of the supply transformer provides voltage such that when rectified, we get 495vdc. The amplifier currently has two 400v 3300uf capacitors in series, separated by 220k bleeder resistors, for a resulting 1650uf @ 800v.. If we were to use the existing value at the front of the PI filter and another 1650uf 800v cap at the back of the PI filter, what would be the ideal inductance of the inductor? I came up with 8.5mh, but I'm not 100% confident about that value. What say you, and what's the proper way to calculate this?

    Regards,
    Jim
     
  12. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0

    Yes, this possibility crossed my mind as well.
    I do know that the output transformer is electromechanically isolated from the chassis by nilon shoulder washers.
    The supply transformers are not. However, (and this is where it gets interesting) when I got these amps, the chassis were isolated from the rest of the amplifier circuit.I knew this was conspicuous so I experimented with grounding the chassis to the amp. This had no effect on the hum however. Interestingly though, even with the chassis electrically isolated from the rest of the amplifier circuit, I was still able to use the chassis as an effective ground connection for the shielded end of the coax cable. How do I know? Because with nothing plugged into the coax jack, I would get hum at the speakers but when I would remove the nilon washer from the jack and allow it to touch the chassis, most of the un-grounded hum would go away but I was still left with my power supply diode pn-junction ripple hum..
    I wondered for a time if the power supply transformer has an internally connected ground, but I then ruled this out with an ohm meter. Even if it did, there is no way for it to work the way it does, as any ground connection internally would result in a short circuit once the secondary voltage is rectified and the negative side grounded.
    Just so you know, this transformer does have a center tap for it's secondary, but it is not connected to anything. The outer-two legs of the HV secondary are connected to the rectifier ..

    I have worked on a lot of amplifiers in my life, but this one is different in the sense that so many things look normal but I get the hum anyway.
    I do think that those in this thread who alluded to the front end picking up hum because of where the first stage is grounded, are definitely at the heart of the problem or at least where the un-filtered ripple is making it's way into the signal path.

    Regards,
    Jim
     
  13. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0
    You are right on track with your thinking.. Indeed we know that the hum, (for whatever we can say about how clean or unclean it may be at the rails and at the ground at various points that one could connect to it), we know that the hum is entering the signal path at the earliest stage of the amp. I verified it by looking at oscilloscope readings at the plates of the 12AT7 tube at the font of this amp.

    I too thought that because of the low-impedance nature of this amp's power supply and lack of a PI filter, that the sharp edges of the sawtooth waveform of the rectifier's forward-ripple was basically "ringing" it's way down the ground rail. I still think we're both right about this.
    One thing I did try, was to take some 10guage solid copper and solder it along the top of the ground rail, in an attempt to reduce the impedance of the ground rail. That didn't help. I also moved the ground connection of a filter capacitor from the far-end of this ground buss, back over to where the rectifier it's self is and where the large 1650uf/800v capacitor connects.
    Each of these attempts has helped in a small way, but the hum is still there to the point that listening to quiet passages isn't as much fun as it would be on my Sansui QRX8001 and a few other gems that I have around the house..

    Anyway, one thing I haven't done which is basically what you are suggesting, is to move the ground connection of the cathode resistor and a couple of other ground connections of the 1st stage, away from where they are and back to the ground points of the 1650uf cap.

    I do want to know more about the fix you and your friend came up with for his kit amplifier.
    You say you cut the signal ground from the power ground and reconnected it to the ground side of the filter capacitors. Was this just the ground of the coax cable, or did you all move such things as cathode resistor grounding (or equivalent if this were solid state)?
    any other details would be appreciated.

    Regards,
    Jim
     
  14. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0
    Yes.. When input is shorted using a shorting plug or a test-lead, the speaker are at their quietest. Even so, there is always still that hum that we are trying to get rid of.. It's not the kind of hum you would hear when simply removing the ground of a signal connection; it's more solid than that, and resembles a power supply who's filter capacitor just isn't sufficient. It's at 120hz.

    Regards,
    Jim
     
  15. blocco a spirale

    AAC Fanatic!

    Jun 18, 2008
    1,440
    368
    If you believe your first statement is correct then your proposed solution is unnecessary.

    Can you post a diagram of the earthing arrangement showing exactly how things are connected, it's a little difficult to visualise from your description?
     
  16. Lestraveled

    Well-Known Member

    May 19, 2014
    1,957
    1,215
    Jim
    The electrostatic shield is a separate shield internal to the transformer. You would have an extra isolated wire coming out of the transformer. If you read resistance of this line to anything else you would read an open. It has nothing to do with grounding the case of the transformer.

    Do the power transformers have a separate ground wire coming out of them??
     
  17. earth1

    Thread Starter New Member

    Mar 25, 2016
    8
    0

    Thanks for the welcome..

    I do have a scope, and I was able to detect the ripple at the plate of the 1st stage of the amplifier.
    This hum is amplified from stage to stage until it's audibility is undeniable.

    Also I don't have a schematic for this amp.
    The amplifier is a VTL Deluxe 120.
    I emailed the company and was told that they don't have the schematic available for this amp, but they did send me some schematics of their equivalent connections, basically a schematic showing the topology that applies to all of their early amps.
    Examination of the circuit confirms what VTL's example schematic attempts to show. The first stage is a 12AT7 twin triode. The two triodes are ganged together. (that is, the cathodes are tied together, the plates are tied together, and the grids are tied together). Pins 4 and 5 are tied together so that the filament can be supplied with 6.3v instead of 12.6v.
    I will check later to get the values of cathode and plate resistors. There is a resistor also going from the grid to ground, and as explained in other emails, the ground point for this 1st stage, although it's along the same ground rail, it's not at the same exact spot as the 1650uf capacitor is grounded. I'm considering moving the physical point where this 1st stage connects to ground.

    Regards,
    Jim
     
  18. ronv

    AAC Fanatic!

    Nov 12, 2008
    3,292
    1,255
    Sounds like kind of a strange set up. If the 12AX7's are on a separate supply increasing the filters on the big guys probably won't help.
    Since transformers were changed that is the area I would suspect. It seems strange to me that they would rectify the filament supply for all the tubes. Mainly because it would take a huge cap to filter it. Usually it is only the preamp that is filtered. Then 5 or 10,000 Ufd will work. If you decide on a pi filter for the plate supply the inductor needs to be very large or the audio can excite it.
    Not much help I'm afraid, but maybe some ideas.
     
  19. SLK001

    Well-Known Member

    Nov 29, 2011
    820
    229
    What exactly is ground? Is it the chassis? If so, look for corroded joints - rivets, etc. How about a picture of what you are talking about?
     
  20. RichardO

    Well-Known Member

    May 4, 2013
    1,235
    384
    His was a solid state amp on a PCB. The ground was sort of "U" shaped with the low level inputs on one leg of the "U" and the power output on the other leg of the "U". The power supply filter caps connected to the bottom of the "U".

    The flaw in the PCB layout was that the charging current for the caps flowed through part of the bottom of the "U". The effect was that this current was added to the ground reference between the input circuit and the output circuit. The difference voltage caused by the current flowing in the ground trace was amplified by the low level input circuit that spanned this section of the ground trace.
     
Loading...