Help with VVR on a AC30

[Mostro]

Hi people, It´s been a long time since I posted here but I read from time to time when I have some doubts, the amount of knowledge here is great.

So now I have a problem, maybe you can help.

I´m building a modified AC30, only the normal and top boost channels, I added a dumblelator FX loop in it, master volume either post PI or pre PI (to be decided), and a VVR to control the beast. Some minor details to be defined once the thing is running OK. Here´s the schem:

[IMG597]http://i68.photobucket.com/albums/i5/Morocotuco/00---AC30-schem-12_zps5be23570.jpg[/img]

and a pic of the build:

[IMG1042]http://i68.photobucket.com/albums/i5/Morocotuco/20130810_173456_zps8b404184.jpg[/img]

For debugging reasons, I made it first without connecting the VVR at all. fired it up, some problems, oscillations, hum, shielded some wires, solved those things, the amp was running OK save for some minor stuff. So onto the VVR then. I connected it and it didn´t work.

The symptoms are:

At maximum voltage the amp works OK, voltages stable and reasonable. As soon as I start to turn it down, at about 1/8 of the pot´s turn the thing starts to make a loud buzzing noise, the voltages become unstable and readings go crazy (1000V for example) which I assume is ripple confusing the meter. So I could only go from 335V to approx. 215V. At some point in the debugging I managed to get it to 32V at the end of the pot, so the thing kinda worked in reducing the voltage.

I tried many many things to make it work:
- Different values of the gate R from 10K to 100K
- Different values of the zener from 6V2 to 15V
- Different values for the current sensing resistor from 3R9 to 15R
- Different MOSfets, three IRFP460 and one 2SK2837
- A 22uF reservoir cap between the rectifier and the dividing diodes D1 and D2

The behaviour is always the same with slight variations but no solution. I tried the circuit inside the amp and then outside of it, connected with wires. The MOSfet is mounted on a heatsink NOT attached to the chassis, it´s currently mounted on a piece of wood, so no possibilities of shorts to ground from it.

So I´m out of ideas as to what is wrong. The last thing I can think of is some kind of loop through the ground, or positive feedback or something to do with the grounding scheme. In the schem I numbered the ground connections according to how they are grouped. I have a "bus" wire that runs across the board,with the preamp in one extreme and the power supply in the other, and the different connections to it along the bus. The ground is connected to chassis at the preamp end only, all jacks are isolated plastic ones. Here´s a drawing of the grounding:

[IMG830]http://i68.photobucket.com/albums/i5/Morocotuco/AC30-HEAD-layout-GROUNDS_zps9dbf55b3.jpg[/img]

Well, that´s the problem. I have installed the VVR in another amp and it works OK, with the same MOsfet (IRFP460), only difference is the zener is a 6V one. Of course it´s a different amp, a push pull 6V6 Fender-like thing.

What do you think? is it possible to get oscillation/feedback/problems through ground? Since there´s two "branches" for power, should there be two branches for ground also? I´m out of things to try and I really want to make this work because without the VVR getting OD from this amp is out of the question unless I want to blow the roof out of the house (and my eardrums), and the OD is one of the main reasons I made this amp in the first place.

Any help is welcome!

P.d: I can provide more detailed photos / diagrams of the layout, recording of the noise, etc.

[Colossal]

Taking one variable at a time, just looking at your schematic (very nice work by the way), you need a cap IN FRONT OF the VVR. The VVR is trying to source current directly from the rectifier and this is likely the issue once you turn down the power. I would put a 10-33uF cap in front of the VVR circuit.

[Mostro]

Colossal, thanks for the answer. Although it´s not on the schem I mentioned it on the post, I have a 22 uF cap in front of the VVR. Can´t go much higher because in the preamp power branch, after the D2, I have another 22 uF in parallell, and I don´t want to exceed the GZ34´s max cap, that is 60 uF. Might have to try a 33 uF...

[Colossal]

Colossal, thanks for the answer. Although it´s not on the schem I mentioned it on the post, I have a 22 uF cap in front of the VVR. Can´t go much higher because in the preamp power branch, after the D2, I have another 22 uF in parallell, and I don´t want to exceed the GZ34´s max cap, that is 60 uF. Might have to try a 33 uF...

Ok. It's not necessary to make that first cap (that the VVR sees, coming off the rectifier tube) that large. Yes, keep the values down but not so low that the amp produces ghost notes. Do you have an oscilloscope? Before moving on to other areas of the amp to troubleshoot (since you say your problems seem to arise from turning down the power), to ensure that the VVR is truly working as intended, can you look at the output ripple voltage and measure voltages at different settings of the Power control pot. Perhaps, as you said, you have a ground loop. Although your grounding method follows Merlin Blencowe's and AX84's scheme, you may need to consider separating your power amp and screen grounds from the preamp ground. I had an issue with a Vox type build once where I had to do this. Despite logic in this common ground rail that terminates at a single ground point at the input, once I separated the preamp and power amp grounds, it resolved the issue.

Another thing, although others have been able to make a VVR work with having the control pot wired remotely (on the back panel in your case), it is generally best to have that run as short as possible. Perhaps there is oscillation in the MOSFET at lower power settings with your application. I am just basing my speculation off of what I see on your drawing and from the chassis photo. Do you have any NTE2973? This is the MOSFET Dana Hall uses and has a massive track record of successful installations.

[10thTx]

There are some schematics, layouts and photos regarding cathode biased VVR's here:

http://www.el34world.com/Forum/index.php?topic=6899.0

Maybe something in that thread will be useful in providing a solution.

With respect, 10thtx

[Mostro]

There are some schematics, layouts and photos regarding cathode biased VVR's here:

http://www.el34world.com/Forum/index.php?topic=6899.0

Maybe something in that thread will be useful in providing a solution.

With respect, 10thtx

10thTx, thanks, I´ve read that thread at Hoffman, but no answers to my problem there.

Ok. It's not necessary to make that first cap (that the VVR sees, coming off the rectifier tube) that large. Yes, keep the values down but not so low that the amp produces ghost notes. Do you have an oscilloscope? Before moving on to other areas of the amp to troubleshoot (since you say your problems seem to arise from turning down the power), to ensure that the VVR is truly working as intended, can you look at the output ripple voltage and measure voltages at different settings of the Power control pot. Perhaps, as you said, you have a ground loop. Although your grounding method follows Merlin Blencowe's and AX84's scheme, you may need to consider separating your power amp and screen grounds from the preamp ground. I had an issue with a Vox type build once where I had to do this. Despite logic in this common ground rail that terminates at a single ground point at the input, once I separated the preamp and power amp grounds, it resolved the issue.

Another thing, although others have been able to make a VVR work with having the control pot wired remotely (on the back panel in your case), it is generally best to have that run as short as possible. Perhaps there is oscillation in the MOSFET at lower power settings with your application. I am just basing my speculation off of what I see on your drawing and from the chassis photo. Do you have any NTE2973? This is the MOSFET Dana Hall uses and has a massive track record of successful installations.

I´ll keep the reservoir cap at 22uF then.

I do have a borrowed oscilloscope. i refrained from taking measurements because the probe input says "300V pk MAX", I don´t know if that is absolute AC or AC riding on top of DC voltage, or if it´s positive OR negative peak OR peak to peak... don´t want to break the scope. The amp has more than 300V. Can I use it?

Yes, the pot is remotely placed. Haven´t thought of that. Because the MOSfet is mounted in a heat sink that in turn is mounted to the board and "peeks" outside of the chassis through a hole (you can see it as an orange thing in the layout) I can´t mount the pot near the MOSfet, and I really don´t want the MOSfet attached to the chassis for safety reasons. Could I shield the wires to the pot? I might try it with the pot closer to at least see if it´s causing trouble.

Regarding the NTE2973, that is not available here in Argentina sadly.

I´l see if I can try separating the grounds, it´s a bit hard because the wiring is mostly solid core 1 mm. copper wire. Time to get out the big cutter.

[John_P_WI]

I have not read the whole thread so I apologize IF you have already done this.

1) twist all wire pairs and wires going to the pot and to the mosfet.
2) make sure the zener is mounted on the legs of the mosfet to prevent oscillations.

[Mostro]

John, thanks for the suggestions. The zener is right at the gate of the MOSFET. the wires are kind of twisted, not tightly, but... I´ll try tighter twist.

Also, before any more experimentation, I´ll draw a schem with the grounds exactly as they are in the board, to be able to better see if there´s a feedback path or something there.

[Tillydog]

I could see that there could be oscillation due to the interaction between the choke, the VVR and supply sag - have you tried it without D3?

Do you still get the problem with the MV wound to zero? (If not, then the problem could be the amount of filtering between the normal and regulated supplies).

If neither of these, then I second the advice to twist the wires to the pot. I would also put a resistor in series with the gate of the MOSFET (like a grid-stopper resistor) - anything 100R - 1K, but this might be difficult with your layout.

[Mostro]

Tillydog, you´re THE MAN!!!!!

I´ve been trying many many things, scaled the whole amp, removed diodes, twisted cables, removed unused caps, lit some candles, swore in two different languages, and so on and so on. I was getting quite desperate, started to contemplate ditching the VVR and building a power soak, or just take a hammer and make thousand of little AC30 pieces. As a last resort I tried a 1K at the gate, and boom, the thing started working right!!!!!

Good Lord, I almost gave up, I´ve been running in circles for weeks with this.

Now I have to test it thoroughly but I have the feeling I´ve found the answer. Now, I´ve read every post / article / page / schem I could find on this circuit on the net, and nowhere I found any mention of anyone using a gate/grid stopper R. My guess is that the mosfet was oscillating, in many places I read that it´s a common problem with power mosfets, the gate has a lot of capacitance, cables /PCB traces have inductance, it´s real easy to trigger oscillations...
So, i´m the only one on Earth that had this problem??? Seems it should be quite common...
I feel that I have discovered something here. Should everyone start using gate stoppers to prevent this?

[Tillydog]

Well, I hope it turns out to have solved the problem with your amp...

Should everyone start using gate stoppers to prevent this?

Yes

(I have been brought up to believe that it is good practice when designing with MOSFETS.)

[Mostro]

Tilly, why do you think nobody uses gate stoppers? And, what is your background with the mosfets? I mean, do you use them just in guitar amps or some other application? I ask this to learn as much as I can on this critters.

[Tillydog]

I don't really know - laziness? Ignorance? Or that nine times out of ten it doesn't cause any problems?

I come at this from an electronics background, and the 'text book' designs all seem to advise a small resistor in the grid lead for FETs and MOSFETs.

I suppose that the 100k resistor from the pot wiper to the gate would fulfil the purpose if it was mounted close to the gate connection.

[Mostro]

While trying to resolve this, I´ve read many manufacturer´s papers about mosfets. Granted, they talk about using them as high frequency switchers, I believe for switching power supplies or electric motor control, but they ALL talk about gate capacitance and oscillations, and how to deal with that.

The 100K R should be a gate stopper, but I think the difference is this: in all the VVR schems, the zener is connected directly to the gate, and somewhere on those papers I read that a zener directly to the gate can cause oscillations. What I did was put a R BETWEEN the zener and the gate. do you think that could be the difference that made it work?

Of course I´m just guessing, I don´t have the deep knowledge to back this idea up with theory.

Also, thinking about why nobody had this problem, I see most used the NTE2973. That has a Ciss of 2900pF, I used the IRFP460 that has 4900pF and the 2SK2837 that has 3700PF. The test conditions are different in each datasheet, but could that be a factor?

Also, I have a VVR in another amp that uses the IRFP460, same VVR schem, and that never oscillated. The only thing i can spot as different is that here I used a 5V zener, in the other amp it´s a 1V zener. Maybe the 5V one triggers oscillation easier? Has more inductance or something?

[Tillydog]

I doubt that the zener has too much to do with it (but it might) - more likely the length of the connection from the gate to the resistor / zener and where that wiring is routed. Until the zener is conducting (either way) I don't think it does much more than add a bit of capacitance.

To be 'text book' the 'stopper' resistor should be right on the gate, and ahead of any other connections.

I think you need a certain amount of bad luck for the lack of it it to cause a problem.