Capacitive Divider Bias Supply Help

[Wilder Amplification]

Hey guys. I'm having issues with a test circuit I'm breadboarding of a capacitive divider bias supply. Here is the test circuit -



Symptoms are -

I'm ending up with double the voltage at the output of the B+ bridge rec

The cap in the divider is getting hot

I have no AC voltage at the output of the capacitive divider as well as no voltage anywhere in the bias supply. Even with the bias supply disconnected I have no AC voltage at the output of the divider.

I've checked...and checked...and checked...the circuit is correct. The transformer is a 370VAC and the 220nF is rated at 630V.

Any ideas?

[defaced]

I've simulated a circuit similar to this less the 56k resistor to ground and it's worked. I can't unfortunately post a schematic because the simulation program I used is broken. This is the post I used to design my circuit. http://music-electronics-forum.com/t12693/#post100408

Seems to me the difference is changing the resistor with a diode, which would stop half of the AC cycle which is probably going to ground and causing the hot cap.

[g1]

Oops, my file correction was incorrect as pointed out by Merlinb.
Attached are 2 examples of existing circuits in Ampeg & Marshall.

[Merlinb]

The schem is correct as you have drawn it (the 56k to ground is required, or it won't work) except for one thing: You need a load on the high voltage output (100k say) or the coupling cap will act like a charge pump and your output voltage will keep rising!

It is probably getting hot because it has now gone over its voltage rating and has died.

[Wilder Amplification]

The schem is correct as you have drawn it (the 56k to ground is required, or it won't work) except for one thing: You need a load on the high voltage output (100k say) or the coupling cap will act like a charge pump and your output voltage will keep rising!

It is probably getting hot because it has now gone over its voltage rating and has died.

Thanks a bunch Merlin! Worked like a charm! Used a 330K as the DC load on the BR.

Which leads me to another question...on the JCM900 that uses the capacitive divider bias supply, there's nothing in the circuit that provides a DC load on the BR if the tubes aren't installed. Does that mean that you want at least 1 preamp tube installed on these to provide a load when testing with the tubes removed? Or does the slight amount of leakage current in the filter caps provide enough of a load to keep anything bad from happening and allow this circuit to work in that scenario?

Another thing, it seems that as I change the bias supply resistor values in an attempt to get more B- I get diminishing returns. Is the 56K at the coupling cap the range resistor for this type of setup? If so, do I wanna increase its value to get more B-?

[Merlinb]

on the JCM900 that uses the capacitive divider bias supply, there's nothing in the circuit that provides a DC load on the BR if the tubes aren't installed. Does that mean that you want at least 1 preamp tube installed on these to provide a load when testing with the tubes removed?

Yes it certainly looks that way to me, unless there's a bleeder resistor hidden in there somewhere (it's not a nice schem to read!).

Another thing, it seems that as I change the bias supply resistor values in an attempt to get more B- I get diminishing returns. Is the 56K at the coupling cap the range resistor for this type of setup? If so, do I wanna increase its value to get more B-?

The cap coupled bias is a tricksy circuit. For any given coupling cap there is an optimum value of shunt resistor that will maximise the negative voltage, after which you can't make it any more negative. However, the load on the HT must not be greater than 5 times the shunt resistor, or you get the voltage multiplication problem (that shouldn't be a problem with all the valves running though).
Another requirement is that the total load resistance on the bias supply (bias pot plus any other dropping resistors) must be greater than the shunt resistor.

Anyway, the easiest way to get more negative voltage is to make the coupling cap bigger. Since it only has to be rated for the peak AC voltage (not peak-to-peak) that should be easy. A 100nF cap and 47k resistor should get you negative voltage that is about 12% of the HT in magnitude, which should be enough for anyone. Since you already have 220nF in there I would suggest that you redesign the rest of the bias circuit so you don't have so much dropping resistance. I would just sick the bias pot directly on the first smoothing cap and then add one more cap directly at the output. More than enough smoothing that way.

[Wilder Amplification]

Yes it certainly looks that way to me, unless there's a bleeder resistor hidden in there somewhere (it's not a nice schem to read!).


The cap coupled bias is a tricksy circuit. For any given coupling cap there is an optimum value of shunt resistor that will maximise the negative voltage, after which you can't make it any more negative. However, the load on the HT must not be greater than 5 times the shunt resistor, or you get the voltage multiplication problem (that shouldn't be a problem with all the valves running though).
Another requirement is that the total load resistance on the bias supply (bias pot plus any other dropping resistors) must be greater than the shunt resistor.

Anyway, the easiest way to get more negative voltage is to make the coupling cap bigger. Since it only has to be rated for the peak AC voltage (not peak-to-peak) that should be easy. A 100nF cap and 47k resistor should get you negative voltage that is about 12% of the HT in magnitude, which should be enough for anyone. Since you already have 220nF in there I would suggest that you redesign the rest of the bias circuit so you don't have so much dropping resistance. I would just sick the bias pot directly on the first smoothing cap and then add one more cap directly at the output. More than enough smoothing that way.

Which I'm seeing about 11% so that's pretty close (HT voltage = 367/Max B- = -40V).

Now of course this is with no filtering...the load I have is a 330K and a 100K in parallel...both rated for 2 watts...just to make the supply work. So the next question I have is...if this were on a fully filtered supply, would the bias voltage I'm currently seeing increase to 10-12% of the filtered HT Voltage?

Also, I'm only able to get an adjustment range of -5VDC between the extremes. I'm using a 25K pot along with the typical 47K/15K divider circuit seen in most Marshalls. Any way to increase this range to a 10 - 12 volt range? Or will this require a bigger value pot?

[Merlinb]

if this were on a fully filtered supply, would the bias voltage I'm currently seeing increase to 10-12% of the filtered HT Voltage?

This kind of bias supply is almost completely invariant with loading of the HT, so your negative voltage will probably remain the same whatever you do with high voltage filtering, but I wouldn't have thought you'd need more than 10% of the HT anyway...

Also, I'm only able to get an adjustment range of -5VDC between the extremes. I'm using a 25K pot along with the typical 47K/15K divider circuit seen in most Marshalls. Any way to increase this range to a 10 - 12 volt range? Or will this require a bigger value pot?

Yeah, use a better circuit!
You are limited in that you must keep the bias loading above 56k, so a bigger pot is needed.

EDIT: I just did a quick sim. Looks like 15k (1W) is the optimum value for the shunt resistor and should get you about 18% of the HT magnitude)

Like I say, I wouldn't use the Marshall circuit, but the arrangement below, as this allows you to vary the bias between 100% and 17% of the raw negative voltage. (You could always remove the 10k to give you full range, or add another resistor at the other end of the pot to limit the range to less than 100%)

[Wilder Amplification]

This kind of bias supply is almost completely invariant with loading of the HT, so your negative voltage will probably remain the same whatever you do with high voltage filtering, but I wouldn't have thought you'd need more than 10% of the HT anyway...

Yeah, use a better circuit!
You are limited in that you must keep the bias loading above 56k, so a bigger pot is needed.

EDIT: I just did a quick sim. Looks like 15k (1W) is the optimum value for the shunt resistor and should get you about 18% of the HT magnitude)

Like I say, I wouldn't use the Marshall circuit, but the arrangement below, as this allows you to vary the bias between 100% and 17% of the raw negative voltage. (You could always remove the 10k to give you full range, or add another resistor at the other end of the pot to limit the range to less than 100%)

Thanks a bunch Merlin! Prior to you mentioning the need for a load on the main BR I was trying to figure out how the thing could possibly work as there was really no way for current to return to the winding opposite the winding that the divider is connected to. Now that I've worked with it some, I see now that it uses the load on the main BR as its "return".

I probably won't ever use this circuit if I can avoid it to be honest. But it's a good circuit to know in case you ever get forced into using an off the shelf PT that doesn't have either a bias tap, bias windings or a center tapped HT winding.

In regards to my original question about the bias voltage increasing...what I'd meant to ask was if the max negative voltage value was dependent on what the HT voltage happens to be?

[Merlinb]

I'd meant to ask was if the max negative voltage value was dependent on what the HT voltage happens to be?

No, the bias voltage is pretty much fixed regardless of HT current demands and sag. It should remain a fixed percentage of the AC voltage, not the positive DC voltage (yes, even though the two are interelated!)

An alternative circuit for getting a bigger and more powerful negative supply is a Capacitative Charge Transfer Circuit. Something for you to play with in spice.

[Wilder Amplification]

No, the bias voltage is pretty much fixed regardless of HT current demands and sag. It should remain a fixed percentage of the AC voltage, not the positive DC voltage (yes, even though the two are interelated!)

An alternative circuit for getting a bigger and more powerful negative supply is a Capacitative Charge Transfer Circuit. Something for you to play with in spice.

That's pretty slick! So basically the "charge pump" capacitor charges up on the negative cycle of the HT winding, then discharges its negative charge into the bias supply circuit on the positive cycle? Also, from looking at the schematic it looks like you'll also need some sort of a load on the HT BR for this to work as well?

[Merlinb]

Also, from looking at the schematic it looks like you'll also need some sort of a load on the HT BR for this to work as well?

Yeah, actually I hadn't got around to really examining this circuit yet, but you're right. Without a load the HT will be multiplied by 1.5 times, and the negative voltage will be 1/2 the peak AC in magnitude.
When you attach a load to the HT it pulls down on both rails, so the HT falls to its normal value and the negative supply falls by the same amount, so you can potentially achieve a bipolar supply.

This one needs looking into.

[Merlinb]

Ok, doing some sims, under normal working conditions a 220nF charge pump cap will get you a bias that is 17% the magnitude of the HT. Plus, it sags in response to the HT too.
A 470nF cap gets you 31% of the HT magnitude!! (Assuming a 20k load on the bias supply itself in both cases.)

The problem is (and it's a big problem), if the load on the HT drops below about 10mA the HT starts its inexorable rise! And at 400V, 10mA is a lot of current to dump with just a bleeder.

There must be an easy way to tame this beast....

[Wilder Amplification]

And this is exactly why this method is even in question lol.

It just makes me wonder...why do some companies make power trannys without a separate bias winding?

I was just turned onto Antek toroidal PTs, which I thought would be an excellent candidate for a dual rail rackmount power amp. They all come with split HT windings, which is perfect for the stacked BR doubler, but...only one of them has a 70 volt tap! And this one is rated for a VA of 400 and both the windings are 400/70VAC, which is WAY too high for anything that I'd build.

They make others but none of them have a bias tap or a separate bias winding.

To make matters worse, none of the windings are center tapped.

So on these, the only thing you can do is either use the capacitive divider, run a separate BR off of the screen winding, or use a separate transformer to derive the bias.

I'm starting to lean toward running a separate BR off the screen winding, which I think would be much easier than trying to get the capacitive divider to work.

[defaced]

I use the Antek transformers and for bias you can take a small (<6VA) 120v/6v PCB mount transformer and run it backwards from the heaters - or you could use a small isolation transformer - I've done both. I think they're like 3 bucks and about 1" square or something. I thought about going down the road you're going down and was really put off by the hoops you had to jump through to make it work. Seems so much easier to just a small transformer just for the bias in these sorts of situations.