Mysterious problem with fixed bias Variable Voltage Regulator (VVR)

Discussion in 'Analog & Mixed-Signal Design' started by PeterGreenFan, Sep 21, 2017.

  1. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    Hello All,

    This is my first ever forum post anywhere, so forgive me if this is not the correct place for this question.

    Despite no real education in electronics, I have tinkered around with tube amps for some time and have a general idea about how they function. Over the summer I took it upon myself to build my own 5F6-A style amp of my dreams. One of those dreams was to install a VVR to reduce power output and allow power tube saturation at low volumes. I was going to buy a pre-made PCB, but they are no longer produced. As such, I cribbed the circuit from an image of the unpopulated PCB from a now defunct installation manual (music-electronics-forum.com/attachments/28470d1397674879-vvr2.pdf). I've seen multiple discussions on the cathode biased version of the circuit, which typically devolve into arguments about the cost of the NTE transistors, but nothing regarding problems with the fixed bias version.

    Here is a schematic of what I believe the circuit to be: VVR Circuit (Original Power FET).png

    Here is my problem:

    When I have the dual ganged pot set to max B+ (VVR "off"), my idle values using a pair of standard 6L6 tubes are the following: (Plate Voltage: 445 VDC, Bias Voltage: -58 VDC, Plate Current [per tube]: 28 mA --> Plate Dissipation [per tube] ~ 12.5 W or ~ 60% max plate dissipation of 20 W). So at idle, my amp is working perfectly and sounds wonderful.

    From reading discussions about the thermal requirements of the power MOSFET (http://ampgarage.com/forum/viewtopic.php?f=6&t=14261), I know that at a 1/2 turn of the dual ganged pot I should be at ~ 1/2 max B+ and 1/2 max plate current resulting in 1/4 power output. This reduction in the power output is the whole point of the circuit. Unfortunately, as I turn the pot almost to 1/2 way I get: (Plate Voltage: 405 VDC, Bias Voltage: - 32 VDC, Plate Current [per tube] 80 mA --> Plate Dissipation [per tube]: 32 W!!!). If I turn the pot to exactly 1/2 my rectifier tube will begin to arc, so needless to say I don't do that anymore.

    Has anyone seen this before? Looking at my schematic can anyone see a glaring error that I am not?

    Also, for anyone who has used this type of circuit, when you adjust your dual ganged pot, is your bias voltage getting less negative or more negative?

    Logically, I would think it should be getting more negative (to reduce the current), but looking at the PCB and my schematic I simply cannot see how that would be possible.

    I have come here in desperation because I simply have no idea what the problem is. Hopefully one of you engineer types will be able to see where I've gone wrong.

    Thanks,
    Kyle

    FYI: I have a mica insulator between the power FET and the chassis. The thermal compound I am using is Thermalcote and has a dielectric strength of ~ 1,200 V/mm, so it should not be shorting to ground. I have checked my FETs and diodes and none of them are showing continuity where they should not be.

    Here are a few pics of the prototype in case anyone was curious. Front View (6L6).jpg Actual Power Layout (2).jpg Actual Preamp Layout (2).jpg
     
  2. wayneh

    Expert

    Sep 9, 2010
    14,708
    5,226
    My first idea is to double and triple-check the pinout of the MOSFETs. It's very easy to place them wrong, there's no particular standard that they all follow, and it's a frequent source of problems.
     
  3. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    What I see is a bad design. The p-fet isn't doing anything and the 1M pot has too much effect, therefore R54 is too small or 1M is too much. Designing this properly requires knowledge of tube characteristics at various B+ voltages and actual grid current (which is theoretically zero). That isn't fun, so I would just experiment by increasing R54 until the negative bias does not fall so much as to let the tubes over-power.
     
  4. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    Always good to double check, so I looked up the data sheets for my two transistors:
    NTE2973: www.nteinc.com/specs/2900to2999/pdf/nte2973.pdf
    IRF9620PbF: https://www.vishay.com/docs/91082/91082.pdf

    Both are left to right GDS according to the documents. Assessment of diode continuity confirmed this.

    Thanks,
    Kyle
     
    wayneh likes this.
  5. wayneh

    Expert

    Sep 9, 2010
    14,708
    5,226
    Do you mean you verified the orientation of the body diodes?

    Hopefully an expert here will come around soon. This circuit is well beyond my pay grade but there are folks around here that will probably have some ideas for you. But in the meanwhile you might get more expertise at one of those specialized forums.

    Is this a guitar amp? You might put that keyword into the title. I think that would attract more skilled eyeballs.
     
  6. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    It's funny that you say that about the p-channel FET. An amp repair guy I went to see puzzled over that as well.

    I agree about adjusting R54. As it is described in the instruction manual, the 100K resistors for both the power and bias sides of the pot allow you to scale your voltage down to 10% of whatever you started with (100K being 10% of 1M). So by that logic, if I replaced the 100K with a 800K then I would only have a range on the bias end of 100% down to 80%. I think this would work; however, the instructions are quite adamant that the resistors must be the same value for the amp to scale proportionately and properly.

    Whether this is true or not I don't know, but I know these circuits were used to great effect in a few different boutique amps, so the circuit must work. I feel like must have interpreted something from the PBC incorrectly which is hindering the plate voltage limiting side of the circuit.

    What that is, remains a mystery to me.

    Thanks for your post,
    Kyle
     
  7. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    Yep. I used the diode mode on my multimeter to find where the body diode was and the direction, then based on the diagram in the data sheet deduced which pin was which.

    I seriously debated about posting to the ampgarage forum, but decided not to because even though the guy who made and sold the PCB version of this circuit is on there, any questions about the circuit always seem to devolve into arguments about how so and so makes a better version, this or that part is overkill, the price of the NTE transistors are too expensive, and it just burnt the poor guy out. As such, I thought I would try a fresh location and see if I couldn't find some electronics engineer types who would look at my schematic and immediately see a glaring error and tell me "You can't do that!"

    Heck, they they could even tell me I'm an idiot for screwing it up. As long as their notion gets this thing up and running I'm perfectly fine with it. :)

    Can I change the title of the thread without posting a new thread? The guitar amplifier addition is a good idea. How would I go about that?

    Thanks,
    Kyle
     
  8. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    Yes, you do know. The arcing tubes are telling you the circuit is wrong, and electrons can't lie!
    Repeat that until you understand it. It doesn't matter what you designed or how you built it, if it doesn't work, you are wrong because electrons can't lie!
    Click the button to, "report" yourself and a Moderator will assist you. We don't have a, "Guitar Amp" section. I think you are looking for, "Technical Repair", but, "Chat" or "Analog" seems fine to me. The local denizens check each Forum they are good at. You got found. I don't see a problem.
     
  9. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    Consider this:
     
  10. wayneh

    Expert

    Sep 9, 2010
    14,708
    5,226
    I think more folks here might chime in on a thread with "guitar" in the title, compared to the original. Maybe not.
     
    #12 likes this.
  11. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    True. "Guitar Amp" would have several people latch on to this like a hungry dog.
    I'm thinking, "Guitar Amp Adjustable Bias Failure".
     
  12. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    True, the electrons can't lie, but the schematic is merely my interpretation of a circuit that was produced, sold, and used in boutique amps long before I came along and tried to decode it from an old installation manual. While it is certainly possible that there is an inherent flaw with the original circuit, I believe it is much more likely that the improper behavior is the result of some failing on my part to properly replicate the circuit.

    Have you had a chance to look at the link to the old instruction manual I referenced up above?
    Here it is again: music-electronics-forum.com/attachments/28470d1397674879-vvr2.pdf

    Perhaps you'll spot something I missed (other than the 1M trim pot which I omitted due to its redundancy with my 10 turn trim pot--the manual makes a specific note about such a circumstance).

    So you think I should move the pot to the other side of the P-Channel MOSFET like it is on the power side? I suppose I could give that a go, but if simply increasing the resistor to ground from 100K to 470K or 820K will reduce the range of my bias to (470K --> -58max, -28 min; 820K --> -58max, -48min) leaving things where they are, what will moving the pot to the gate side of the FET do for me? That's a serious question. I know next to nothing about transistor function, so any information is appreciated.

    Thanks,
    Kyle
     
  13. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    True. There are several reasons why a person might interpret or assemble wrong. I said it HARD because I wanted you to take it as Gospel, not some nebulous, half baked theory that I tossed out while passing by.;)
    Maybe yes, maybe no, but I am working on the parts I CAN work on.
    In the drawing you posted, the mosfet does nothing except subtract about 3 volts from the available bias voltage, and it is not adjustable. You could replace it with a 3.3V zener.o_O No matter how well you followed instructions, you couldn't win because the circuit you posted here was crap before you started. While I'm busy slandering, you are accustomed to a Forum with dubious authorities. I have designed and built more than one guitar amp from scratch. If I say something wrong, there are a dozen people on here who will correct me, and they are usually right.:oops: Now, down to business:
    If that works, do it and forget about cleaning up a bad circuit. I'm a pragmatist.
    Moving the pot will make the mosfet adjustable. The mosfet has a low output impedance which is better for a circuit like this than high impedance. Now that I've downloaded the datasheet for the IRF, I estimate that 1.5M to 2.2M is the value for the source resistor. If you go too high, the bias will tend to drift to the safe side when the mosfet gets hot and leaks more current. If you go too low, you will just waste current from the bias supply and begin to warm up the mosfet. Right now, 0.0005 amps of leakage times 48 volts is 0.024 watts. No struggle at all for a TO-220 mosfet.

    Now that I have downloaded the datasheet for the KT66 I can see that the bias has only a small influence when the B+ changes and the voltages and currents you measured do not resemble the datasheet very much. That's why figuring the resistor under the 1M pot is a guessing game. Just drop the B+ from 445V to maybe 100V (with full negative bias) THEN find out which resistance provides acceptable current flow, then double check the operation over its full range.

    ps, those pots with independent wipers are a bad idea. Wipers get dirty. The suffer from being, "open" when they fail. You need the wiper circuits designed so they fail in the, "safe" direction when the wiper fails to contact the internal resistance in the pots.

    Any questions?
     
    Last edited: Sep 23, 2017
  14. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    Hey #12,

    Your responses and ideas are much appreciated. I haven't had time to reply as I was out fishing this weekend, but I'm planning on tinkering with the amp again tomorrow.

    You brought up a point which has been puzzling me since I first encountered this problem. Given the tendency for tube current to decrease as plate voltage is decreased (when bias voltage is held constant), why would I ever want to make my bias voltage less negative as I decrease my B+? Looking at the KT66 operation curve, if I'm at 400 V, with a bias of -30 V, my plate current per tube would be about 50 mA, resulting in a total dissipation of 40 W--which is about 75% max dissipation for the KT66. If I simply leave the bias voltage as is, and turn my plate voltage down to 100 V, my plate current will now have dropped to about 30 mA per tube leaving me with a dissipation of only 6W. My infantile understanding of electronics makes me think that simply leaving the bias fixed would solve my problems, so why are these fixed bias VVR circuits wired up to make the bias voltage less negative as you turn the dual ganged pot? Is there some sort of current clamping effect by the power MOSFET that needs to be compensated for by decreasing the bias voltage?

    I posted my game plan for tomorrow on the ampgarage forum (http://ampgarage.com/forum/viewtopic.php?f=6&t=30612&p=377861#p377861). In case you're curious, I will definitely be trying out your circuit, but as it requires the most rewiring I'm going to save it for last.

    How did you calculate the source resistor requirement and leakage? It would be nice to know how to do that for future builds.

    Also, is your comment about the independent wiper pot referencing the 1M trimmer I omitted? Because I don't really like those either. They've always seemed flimsy to me.

    Anyhow, thanks again for all your help and I'll let you know what I find tomorrow,
    Kyle
     
  15. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    A bad day fishing is still better than a good day at work.
    I assumed it was intended to keep the KT66 at some specification while reducing the plate voltage. Less B+ and less negative bias = same current? Same plate dissipation? This could get silly, fast! You don't know what your goal is, and neither do I! We both know you found a circuit that would melt a tube.:D
    I looked at the datasheet for the IRF. Leakage at room temperature is 100 ua, then divided by your reduced voltage across the mosfet (1/4th max) by assuming you aren't going to get it very hot at 0.024 watts. (Transistors usually exceed the datasheet specs for quality.) Using the numbers you gave today, if the maximum voltage is -48V and the bias applied is -30V you only have 18 volts across the transistor. 18V/200V x 0.0001A datasheet leakage = 9 ua. 30V to ground / 9 ua = 3.3 meg. Today my estimate is 3.3 meg. That number assumes there is never any grid current and there is never more than 18V across the mosfet. We know you intended to dial down the bias voltage, so 18V across the mosfet is not the final answer. Therefore, I guessed again and said, 1.5 to 2.2 Meg. Better to go low on the leakage resistor and waste some power than go high and see the tube bias drift back to -30V or -40V because of bad guessing. In the end, you can build the circuit and see how it works in a closed cabinet with the tubes driving the speaker.
    I keep looking at the red square in post #1 and seeing two pots which will go, "open wiper" when they get old or dirty.

    If your B+ wiper goes, "open", the gate will float and the maximum B+ will be applied to the tubes. That won't hurt anything if you don't have the bias turned down...but you intend to turn the bias down.:confused: That suggests that the B+ wiper needs a few megs to ground to make the mosfet shut down if the wiper goes bad.

    (You don't even need the 50K bias pot if your mosfet circuit works.)

    If the pot for the bias mosfet goes "open" your post #1 circuit won't apply any bias voltage: Instant tube melt.
    If the circuit I drew goes "open wiper" the gate will float toward the drain and the bias will max out. No melting.
    Still, I would add a couple of meg from drain to gate to make sure the gate of the bias pot floats high, and that would cause a re-think on R55.

    I assumed you knew about these things when I designed that bias circuit. I was wrong.:oops: I said, "consider this" because I did not believe the circuit was finished, and you could finish the design. Wrong again.:(

    This whole bias section needs to be re-thunk. Do you want to keep the original bias pot in place? Do you want constant plate current or constant plate dissipation or neither? Did you know the bias voltage changes with the wall outlet voltage? The wall outlet at my place is 125 VAC. I set the bias for my house, the amp goes to California, and the negative bias isn't strong enough! That's why I use a Variac to turn down my wall voltage to 115 VAC and then set the bias. Been there, done that, but your modification started out as crap and it isn't finished yet.

    I do things very differently. Observe:
     
  16. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    I see in your ampgarage post some discussion about two adjustments being independent. In a good design, they are, but this design gangs two pots together and so, interaction is intentional. I want to emphasize that 820K was just a guess. Whatever size resistor is required is what you need to install. I don't care what size it turns out to be. The one that works is the right size.

    As for the accelerator grids: When you decrease the B+ those grids also have lowered voltage. In fact, your phase splitter and your input tubes have less B+! (Some people only lower the plate voltage on the input tubes to get the, "brown sound".) If you want the input and phase splitter to act as they originally did, you could isolate the power tube voltage from the input and splitter stages, but that's a whole 'nuther animal.

    I want you to forgive how I use words like, "accelerator grid". I am an analog designer, but I started from a physics point of view. Imagine when I got my state certification to design air conditioners and I didn't know what to call the grilles that blow air into each room!:D I call 'em what they do (from a physics point of view). To this day, I don't remember what the trade name of the second grid is. I just call it what it does.:rolleyes: Control, accelerator, suppressor...that's what they do.
     
  17. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    Well...I finally found the problem. It certainly wasn't what I expected, but made the repair very simple.

    To determine if my problem was due to improper operation of the bias scaling circuit, I bypassed the bias scaling circuit all together so that no matter what I did to the B+ my bias would stay at max. As I turned to pot leading to the power MOSFET toward 1/2 way, I saw virtually no change in my plate voltage, but this time my plate current was staying the same.

    As a result, I knew my problem had to be in the B+ side of the circuit. I tested the pot to make sure that it was providing the correct resistance throughout the range and it was. Next, as it was a simple thing, I began to prep myself for an experiment assessing the voltage from the wiper to either side of the 100k gate resistor. As I looked at the leads from the pot to the VVR board my jaw dropped. Despite painstakingly hand drawing, to scale, my preamp and power layouts correctly, when I actually wired up the pot, I had accidentally wired the 47uf electrolytic, drain, and WIPER in parallel--with pin 3 leading to the 100k gate resistor as seen here: How not to wire up VVR....png

    Instead of how it should have been:
    What it should look like....png

    Once I wired up the B+ leads correctly, I fired up the amp again with the bias still fixed at max. Low and behold, my B+ values began to scale as I turned the pot, but my current dropped even faster. It was basically 0 at a 1/2 turn. Apparently this is a result of the simultaneous decrease in screen grid voltage with the plate voltage, so the current behavior is much different than observed in the fixed screen grid condition reported on the tube data sheets.

    With the B+ scaling as it should, I rewired up the bias leads to the dual ganged pot, and just like that the behavior of the scaling circuit was exactly as described in the old ampgarage discussion on the MOSFET thermal requirements (http://ampgarage.com/forum/viewtopic.php?f=6&t=14261). At a 1/2 turn of the pot, I was getting about 1/2 max B+ and about 1/2 max plate current for an output of about 1/4 max.

    With the scaling circuit functioning satisfactorily at the moment, I'm going to hold of on any modifications to the design until I complete all the actual life things I put on hold while I tried to figure this thing out. However, when I get the chance to tinker with this thing some more, I'll post my findings and eventually some pictures of the amp (with cabinet) when the project is completed.

    Thanks again for all your help and explanation of your calculations. It was much appreciated,
    Kyle
     
    #12 likes this.
  18. PeterGreenFan

    Thread Starter New Member

    Sep 21, 2017
    8
    3
    I really wish you had asked me to double and triple-check the leads going to my dual ganged pot... :)
     
    wayneh likes this.
  19. #12

    Expert

    Nov 30, 2010
    18,078
    9,616
    I can't fix a mis-wire from here!:D
    But I want you to check the voltage on the source of the bias mosfet and see if it changes. If it doesn't, the mosfet isn't doing anything useful.
     
Loading...