MOSFET source follower for output stage of AB763 blackface

[SixStringBender]

Heyboer will indeed wind it to any spec. Here is what I specified and the quote was $130 plus shipping. I haven't received the payment invoice yet.

122.5V primary
315V-0-315V at 180mA
115V-0-115V at 50mA
50V-0-50V at 50mA
3.15V-0-3.15V at 5.0A
No 5V winding

2-1/2" by 3-1/8" mounting centers.

Internal hum shield on a lead for grounding
Paper stick-wound
M-6 grain oriented lamination steel
Copper flux bands
Uni-strand pretinned 12 inch lead
Stainless steel through bolts

I went with 50-0-50 bias supply so I can do a full wave bias supply with two diodes.
No 5V winding because I will never use a rectifier tube in it again. I have too much main filtering for a 5AR4 anyway.

It'll be 5-6 weeks before it ships. Plenty time for me to plan more and lighten my Mouser order. I was going to be out $95 for the Allen TP35 plus KY sales tax (I'm in KY) and shipping. In addition $70, or more, for the transformers, and other parts, from Mouser to do the bipolar power supply and a mains voltage reducer to drop some B+. Now I don't need any of the parts for the mains voltage reducer. I actually saved money and work drilling and mounting a second PT that would likely cause hum anyway. And work building the voltage reducer.

Do any of you see anything wrong before I pay? I can still change the order. Should I increase the bias tap mA to provide more current?

[Kagliostro]

Just to remember

have you considered that using SS rectify B+ voltage will increase if is used a PT with the same AC voltage of the previous PT ?

Franco

[SixStringBender]

Hi Kagliostro, yes I took that into consideration. I think the stock PT is 330V-0-330V with a 115V primary. I don't have a specs sheet on it, but when I test the AC filament voltage on the output tubes I get a total of 6.78VAC (adding pins 2 and 7). My main line voltage is 122.4VAC (when I tested it). So I think the primary is 115V based on this. The 330V-0-330V is a spec I remember reading on the net about the DRRI PT.

So I reduced the HV voltage a little as well as increased the primary voltage spec. Both will reduce the B+ in an area where it should be good with the SS rectifier. I'm at 476VDC, I think this should put the B+ around 415-420VDC.

The 122.5V primary should put the 115V-0-115V in spec as planned for the bipolar power supply for the MOSFETs and it should put the filament right in spec for the tubes too.

My only concern is should I increase the current for the MOSFETs and/or the 50V-0-50V bias supply. Since I read the Fender bias supply is wimpy I figure I should beef it up a bit. Full wave rectification will be a good thing for smoother negative DC, IMO. I could have done a 0-50V isolated winding, but I would have had to install a bridge rectifier to get full wave. It was the same price for 50V-0-50V, so I decided to do that. They told me they've never done a 50V-0-50V for the Fender bias supply before. So unique I guess. I know I'll probably have to go with a larger resistor than the DRRI's 22R. That is where I'll need some help determining what value resistor to put after the full wave rectifier.

I also specified for all the secondary windings to NOT share any taps from each other. I wasn't sure if the bipolar power supply would be effected if it was tapped off the main HV supply, so it is a totally isolated winding all it's own. As is the 50V-0-50V winding too. And of course the 6.3V CT winding is as well.

I just wonder about the current rating for the bias and MOSFETs. I assume there is no need to beef it up.

[SixStringBender]

New bias supply plan with the dual bias and center tapped winding. Two questions below to help me finalize my PT order.

With the long SPST tremolo intensity pot removed and a dummy pot installed I'll have room for another CTS 10K trim pot where the original bias pot was located to simplify my install and fill the hole.

The original negative DC supply wire to the original single bias pot will be removed from the pot and board.
The wire that goes from the outer pot lug to R59 will be removed.
C29 and R59 will be clipped and removed. Nothing will go back in their place.
The wire to the wiper from the PCB will be left in place.
We will cut the trace that joins R56 and R57 together. Witch ever one that is no longer going to the wiper of the original bias pot will be put to the wiper of the new bias pot by soldering a wire to it's lead on top of the board.

I will mount a small turret board with stand offs in front of the PT (inside the chassis) with two 1N4007 diodes' anodes joined and the cathodes separated.
Each 50VAC supply lead will connect to the cathodes of the diodes for full wave rectification to negative VDC of the 50V-0-50V supply from the new Heyboer PT. The center tap goes to ground.
A 10uF 100V electrolytic cap will have it's positive lead to ground and it's negative lead to the anode of the full wave rectifier.
A divider resistor (?K) will be installed to isolate the rectifier and first filter cap from the next filter caps and bias pots.
Each bias pot will get a 10uF 100V cap with it's positive lead to the pot shell and the negative lead to the input lug with the wire from the divider resistor.
We will put a 10K 1/2W resistor from the wiper lug of each pot to the pot shell (I think that is correct, the text on Hoffman is unclear where the resistor goes???).
The other outer lug will be left empty so the pots will operate as a variable resistor.

A couple questions before I finalize the Heyboer PT order with payment.

Should I leave the bias supply voltage at 50V-0-50V or go higher with this new better rectified and filtered bias supply with divider resistor? Consider the divider resistor with the voltage supply too. Marshalls use a 100V bias supply. I don't know how much voltage is knocked off by the divider resistor. I found this bias supply plan from Hoffman and implemented the dual adjustable bias to it.
Also the mA rating for this bias circuit?

[Kagliostro]

I've seen bias supply that uses 50uF or 100uF as first cap value

and 10uF as second cap value, may be the use of a 100uF as first cap will give

you better performance if compared with a 10uF first cap

--

In an old Geloso PA amp I've seen also a 500uF as first cap and a 100uF as second cap, but they feed the bias circuit and the filament of V1 and V2 from the same 4 diode bridge and may be a so high value of the capacitors is due also to this reason

K

[SixStringBender]

The reason for using the smaller caps is for faster charging. Plus with the divider resistor allowing for dual filtering there is less need for large filter caps. And with full wave rectification there is less need for larger caps because there is less ripple in the current. This is a fast charging low hum bias supply design.

The idea for full wave rectification for the bias supply was my own, but the knowledge of using smaller caps came from Aiken. I was going to do as you say and use a 47uF at 100V (as I read on Hoffman), but I began studying why the Marshall supply is better. It is better due to the divider cap allowing for two stages of filtering. Plus the ability to use two smaller caps for faster charging. And again, Hoffman's "better bias supply" schematic is the source for my new bias supply description. It is a two filter bias supply circuit.

I'm needing to know if I need to bump up the voltage to take better use of this bias design?

Reason I am asking is I believe the divider resistor will drop the supply voltage.

Also, since the bipolar power supply for the MOSFETs is part of the bias circuit I think the charging speed of the caps should be considered. I believe smaller caps could be used here too. I don't know about using a divider resistor for two stages of filtering, but as I understand it I don't see why not. Again, the voltage may need reconsidered because I assume the divider resistor will knock off some voltage. Also, we again have full wave rectification to work with here too, So the need for large filter caps is not as critical.

Some discussion on this topic.
http://www.thegearpage.net/board/archiv ... 30418.html

See what Aiken has to say.

It'll be nice to have R.G.'s and martin's knowledge on this too.

[martin manning]

I don't see anything wrong with the dual bias supply I diagramed earlier. The '65 DR already has two filter stages in its bias supply, so all you have to do is break the trace from R56 or R57 from C29 and add a new 22uF cap (a duplicate of C29), going from whichever of those resistors was disconnected to ground. You could probably find a place to drill two small holes through the PCB to insert a radial cap, and bend the leads over to the appropriate pads: positive lead goes to the positive lead of C29 (ground), and the negative lead goes to the disconnected resistor. That junction would be the connection point for the wiper of the new bias pot.

The first filter is C36, which is fed by R69. That will be shared by both branches of the dual supply. You could just replace bias rectifier CR1 with a jumper, and connect the anodes of your FW rectifier to CP10. The last word in the long post to TGP by Aiken is that a FW bias supply will have lower ripple for the same capacitance, which is of course true. If you want more negative bias voltage, increase your bias winding to 60VAC, otherwise you're fine with 50VAC.

The 100uF caps I specified for the bipolar supply will charge very quickly, so no worries there. They are going to have to supply some current (that's why you're going to all this trouble!), so you don't want them to be too small.

[SixStringBender]

Hi martin, I didn't know C36 was the first filter. Since I can't see the schematic and my wife is short on time I hadn't went over the complete DRRI bias supply. I was going on the descriptions of the original blackface bias supply and there is only mention of a rectifier diode, then the load resistor, then the filter cap, then the pot. So no two stage filtering in that description. I was trying to improve something on the DRRI that I didn't know it already had.

100uF seems a bit high for C36 doesn't it? I guess considering it is half wave rectified, that is probably the reason Fender went with it, but it still seems high based on the description of how large caps slow the response on the bias circuit.

You said join the cathodes, I think you mean join the anodes?

I can't see well enough to use the existing circuit and terminals and I'm afraid my wife may short something out when drilling the PCB without picture instruction. This is why I want to do it all on a small turret board. I can lay the parts out on the board, explain to her how they go and have her make it neat, and she can mark the places to drill for the turrets. We can do it all like that easily.

I have been doing more thinking and reading and I had decided earlier to dump the Hoffman variable resistor bias pots idea and put the filter caps from the wiper to ground. I finally understood the filter caps after the pots would give me my divided resistance and give me my second filter stage. I was unsure if 22 ohms would be enough division in the method I stated earlier.

If I am calculating it correctly and learning it correctly, the voltage with a full wave rectifier at 50V will be 70V. I believe this is higher than with half wave rectification. So 22 ohms probably won't put the bias voltage in range.

My latest thought was to put a 2K trim pot in series with a 22R resistor so I can adjust the voltage to the bias pots to ensure I can find the range for biasing any tubes I may install.

Here is my latest plan. Using a small turret board.

Center tap to ground, 50V taps to the cathodes of the diodes. Anodes together with one lead of a 22R resistor.
Other end of 22R resistor to negative of 10uF 100V filter cap. Positive to ground.
2K trim pot as variable resistor between the negative lead of the filter cap and the input lug on the bias pots.

Wiper of new bias pot, with negative lead of a 10uF 100V cap, to R56 (or R57, depending on trace layout when the cut is made). Positive lead of cap to ground.
The existing bias pot's wiper to witchever resistor is still connected to it's wiper after the other resistor's trace is cut. For the original pot I will just replace C29 with a 10uF 100V filter cap (keeping polarity equal).
R59 will stay in place for the original pot and the new bias pot will get a 10K 1/2W resistor from it's outer lug to ground.

That keeps the layout the same as you described for the pots, but with the rectifier on a turret board and the filter caps lowered due to the full wave rectification. According to Aiken, 100uF slows the charge time tremendously.

[SixStringBender]

About the bipolar supply for the MOSFETs and the filter cap size.
I didn't think the cap size determined current, rather the smoothing of the ripple in the current. Shouldn't the speed of the charge be considered since this is part of the bias supply? Not arguing, rather asking to learn. Couldn't the ripple be smoothed using a two stage filter to obtain a nice smooth DC and achieve a fast response time using smaller filter caps? Randall did say to use a two stage filter any time you can in a bias supply.

Aiken said this about bias supplies, "For those of you interested in really learning how bias supplies work and the tradeoffs (without any "magic" BS), here is a plot of the difference in bias supply risetime in a Marshall 1987 50W circuit when replacing the stock 8uF bias filter capacitors with 47uF capacitors:

http://www.aikenamps.com/bias_caps1.pdf

Note that with the stock 8uF capacitors (the green trace), the bias supply comes up to nominal voltage in around 4 seconds, and in a "safe" area in about 1 1/2 to 2 seconds. With the 47uF caps (the red trace), however, the supply takes about 20 seconds to get up to nominal, and about 8-10 seconds before it gets in the "safe" area. This is why you don't want very large bias supply capacitors.

Note that the measurable bias supply ripple in either case in negligible, you can't even see it on the scale of the plot. Now take a look at the blue trace, which is the "raw" bias supply at the first filter cap. Notice how much greater the 60Hz ripple voltage is? This is why you want to use a 2-stage bias filter whenever possible."

He also said this that is interesting.
"On a side note, electrolytic capacitors have a rising ESR at higher frequencies, and they can look pretty lousy to fast-rising transitions of square waves. It is always a good idea to bypass electrolytics with smaller value capacitors to maintain bypassing at higher frequencies. Typically, a 0.1uF works fine."

He didn't say what type of cap to use to bypass.

[martin manning]

Yes, anodes to the first filter.

Ripple voltage is proportional to 1/(Rload x C). A bias circuit sees a high load impedance so even with a small capacitance the ripple will be low. Use of half wave or full wave rectification will have very little effect on the bias voltage, and I think the circuit I described will be fine with a 50VAC supply. You should have around -65VDC available.

The MOSFET bipolar supply will potentially have to supply some significant current, so larger caps will be needed to keep the ripple down. The charging rate isn't a problem; I wouldn't hesitate to use 220uF caps there.

The link to Aiken's graphs doesn't work for me, but I'm betting he is looking at a bias supply that sources from the HT, and has a significant resistance in front of the rectifier. That is not the case here, and the bias voltage will be established in a half a second or less.

[SixStringBender]

Hi martin, thank you. You said, "A bias circuit sees a high load impedance so even with a small capacitance the ripple will be low."

So you too agree the 100uF C36 is much too large since the ripple will be mild with small filter caps? Based on Aiken's description it is and it sounds like you agree.

So what do you think about my description of how I would do the full wave rectifier, then a 22R resistor, then the 10uF filter cap, a 2K trim pot as a variable resistor, bias pot inputs, then a second 10uF filter on the wiper of each bias pot? And of course the other connections.

The 22R will behave as an inrush limiting resistor by being before the first filter cap. It will also set the voltage range by being before the bias pots.

What do you think about the 2K trim pot? turned down to no resistance it will be invisible and tweaked with some resistance it will allow for range adjustment of the bias pots. Thinking of other output tubes in the future.

And do you agree 10uF is a good value for the three filter caps in the Fender bias supply since a 100uF cap will make little difference to the ripple? I see no point in having a cap that large when it is not required.

Yes, I believe Aiken is using both HV taps for the full wave bias supply in the description. I didn't even read the article in the link, just the quote from TGP. In another message on that thread he says he installs a full wave rectifier in all his amps using the HV supply, so I assume the example is using the same configuration.

I didn't even think of the effect of the resistor to drop the HV supply in the Marshall/Aiken amps. The only thing I figured it did was sag the supply down. Maybe that is why the filter caps must be small in the Marshall supply? the caps must be small to charge quickly due to the sag? If so, why don't Marshall builders have the proper voltage winding put in their PT? It would make sense to improve on a design any time one can.

About the bipolar supply for the MOSFETs.
I will install two 100uF 250V filter caps in the bipolar supply as you suggest. I think I'm understanding how the drop resistor is sagging the supply by limiting current in Aiken's examples. I think I understand how that is a none issue when using a proper voltage secondary winding. Or maybe I'm wrong... lol

[martin manning]

You said, "A bias circuit sees a high load impedance so even with a small capacitance the ripple will be low."

So you too agree the 100uF C36 is much too large since the ripple will be mild with small filter caps? Based on Aiken's description it is and it sounds like you agree.

The charging time is not an issue here. Fender's bias circuit is half-wave rectified, so their thought may have been to throw more uF's at it to keep the ripple low across the whole bias range. It's not necessary to have a really low ripple either since it's common mode noise. There are several ways to get an acceptable result, and Fender is certainly considering cost and part count in their solution.

So what do you think about my description of how I would do the full wave rectifier, then a 22R resistor, then the 10uF filter cap, a 2K trim pot as a variable resistor, bias pot inputs, then a second 10uF filter on the wiper of each bias pot? And of course the other connections.

That's fine, but I wouldn't bother with the 2k trimmer. Just put a fixed resistor there (1k, say) for each branch so you always have some series resistance before the second filter cap. The range will be quite wide, with voltage from the low 30's to the mid 60's. I'd also make the first filter bigger, double or more, since it has the load of two adjustable networks.

My preference would be to use as much of the existing bias supply, and do the minimum modification to get a dual supply in there (as described above). That would include keeping and using the 100uF first filter.

[SixStringBender]

Hi martin, you suggested a 1K fixed resistor, but I'm afraid that will put the bias pots out of range. Even on the larger amps they only use a 470R resistor to bias 6L6GCs. A 1K trim pot would probably be good enough, but I was originally worried about the voltage increasing with the full wave. I know you said it won't increase, but everything else I read about full wave rectification says it will. The ripple will be reduced by half because there are twice the negative charges on the filter caps. So it seems the voltage will rise some. That is one reason I wanted to use a 2K trim pot. So I can make sure I can find the proper range for the bias pots without having to swap resistors in and out. Plus it will serve as a divider resistor. Again, a 1K trim pot would likely be good enough.

Half the ripple is also why I wanted to reduce the first filter. Plus if 10uF is good for a half wave rectified Marshall with a sagging supply I still figure it will be plenty good enough on a Fender without a saggy supply and full wave rectification. I will do 22uF at 100V there as you recommend.

I don't think I can use the PCB for the bias supply. I just don't think I can figure placement out without a lot of time studying it. The anodes would go to CP10, but what about the cathodes and the two 50V leads? I would need more help to use the PCB.

Or do you mean, remove and jumper the existing diode, put my rectifier diodes on a small board, join there anodes there and run a wire to CP10? Then use the PCB for the rest of the supply? I might can do that if I can find a place to glue down the trim pot. It just hit me that this is what you probably mean.

I can cut the wire to the input of the original bias pot and run the wires to the 1K or 2K trim pot from there. The other bias pot will just have it's input lug jumpered to the original bias pot's input lug to share the supply.

And I just remembered something else. My new PT has a 122.5V primary instead of a 115V primary. That will almost cancel out the voltage increase I would gain with the full wave rectifier. About 4V I figure.

[martin manning]

Regarding half-wave vs. full-wave voltage and ripple, if the peak voltage delivered is the same either way (same VAC input), and the supply has a low impedance (that is the case here with only a 22 ohm resistor between the transformer and the reservoir), and you have enough capacitance in the reservoir to get the same ripple, then the DC voltage at the output will be nearly the same.

The 1k resistor I suggested adding is from the reservoir (first filter) to the input of each bias pot, so that's two 1k resistors, actually. It adds 1k to the 20k going to ground, so it would only reduce the maximum negative voltage available by about 5%. It would improve the effectiveness of the second filter when the pots are set for maximum negative voltage, IF you need to do that. The problem is finding a place for these resistors, and without them the connections to the bias trimmers is much easier. May be more trouble than it's worth.

Yes, above I suggested replacing the existing rectifier with a jumper and connecting the output of your FW rectifier to CP10. Then you can use the 22 ohm resistor and the 100uF cap that are already there and leave them in place.

I still like the idea of adding the new second filter on the board. If you could get a picture of that area with a light behind it to make the traces visible, or just the bottom side, that would help in figuring out how to do that.

[SixStringBender]

Okay, I got it now.

I understand wanting to put the cap and resistor to the same ground source. I'll try to see what we can do.

About the 1K resistors. So the pots can't share the resistor?

I could cut and run the negative supply wire that comes out of the PCB, that was going to the original bias pot, to the end of two joined 1K resistors and the other separated ends to each bias pots input. I could insulate them with heat shrink and the heat shrink would reinforce their leads to the wire at each end via the body of the resistors (3 pieces of heat shrink, 2 small and one large). Would that be okay?

I think you mean the 1K will be in series with each 10K resistor, correct (R59 on the original pot and the other 10K on the new pot)?