totem pole main filter caps Fender Blackface

[sbirkenstock]

Hi everybody,
just working on an ´64 BF Vibroverb where somebody changed quite a few parts.
It does have a totem pole for the main filter caps = 2 * 100uf 350v in series instead of 2 * 20uf 525v parallel.
The schematics of mine (AA763) shows the parallel "version" though.
I have a schematic for the Super Reverb AB763 showing the series version with 2 * 70uF 350v.
So I guess Fender did not keep tightly to the schematics for that?

However, here is my main question:
Fender used two 220k 1W resistors to balance the two caps.
What is the reason for 220K?
Would 100K work as well or 1Meg?

thanks in advance,
Stephan

[pompeiisneaks]

The 220k is I think a common minimum value to allow balanced capacitor charging, and provide a good path to ground for discharging but high enough to not waste too much power. I've seen 220k,270k I don't know how well 100k would work, as it would allow quite a bit more power leakage. 1M may not provide enough leakage to do the job, but I'm not sure. that range is pretty common.

~Phil

[sluckey]

AB763 Vibroverb schematic shows two series connected 70µF with 220K balance resistors.

[gingertube]

The balance resistor thing is one of those buried tech things.
Because the caps are in series and have the same charge and discharge currents through them, they will share voltage according to the cap values. 2 equal value caps will share the voltage equally.

There are 2 things which can screw up the voltage sharing. First is the caps being not exactly equal due to tolerances.
the 2nd is leakage current in the caps.

The balance resistors address both those issues and they should be sized to conduct 3 to 5 times the cap leakage current.
You have to go to the cap datasheets to find a formulae to calculate those leakage currents.

I've actually done that on a few occasions (EE design Eng in the day job) and can assure you that 220k balance resistors are fine for 47uF or 100uF caps.

For cap values like 470uF you would need 100K balance resistors. (The C appears in the leakage current formulae).


Cheers,
Ian

[pompeiisneaks]

The balance resistor thing is one of those buried tech things.
Because the caps are in series and have the same charge and discharge currents through them, they will share voltage according to the cap values. 2 equal value caps will share the voltage equally.

There are 2 things which can screw up the voltage sharing. First is the caps being not exactly equal due to tolerances.
the 2nd is leakage current in the caps.

The balance resistors address both those issues and they should be sized to conduct 3 to 5 times the cap leakage current.
You have to go to the cap datasheets to find a formulae to calculate those leakage currents.

I've actually done that on a few occasions (EE design Eng in the day job) and can assure you that 220k balance resistors are fine for 47uF or 100uF caps.

For cap values like 470uF you would need 100K balance resistors. (The C appears in the leakage current formulae).


Cheers,
Ian

Now that is super cool to learn, thanks!

~Phil

[R.G.]

Sorry to be leaping in to another thread with a long polemic. These things just happened to hit some areas I've worked with a lot. Let me start by saying the previous replies offer solid advice, and if I hit one of them again, it's not me trying to appropriate the idea, just that it fits in there in my meandering.

Two electros in series can be modeled as two perfect caps with tiny resistors in series with each one (that's the ESR for the cap) and a large value resistor in parallel, representing the leakage resistance. Manufacturers won't specify the leakage resistance, as they try to make it zero, but know they'll fail at that. Instead they specify that it won't be any worse than X current when new and possibly Y after its expected lifetime. The point of swamping resistors is to make the voltage share between the two caps at DC conditions be no worse than the caps can withstand.

For some idea of what the numbers are, I looked up the specs on Illinois Capacitor (CDE) at Mouser. Their specs on their TTA line say that the leakage won't be any worse for caps from 160V to 450V than
CV * 0.04 +100uA for caps with CV (the product of capacitance and voltage) less than 1000 and
CV * 0.1 +40uA for caps with CV>1000.

ACK! What's all that mean? Well, you pick a cap - say, a 47uF/450V unit, then calculate CV. In this case that's (47E-6)*450, or 0.02115. Well under 1000 then. So the leakage is guaranteed to be less than 0.02115*0.1 +40uA = 0.00215A +40uA = 0.00255A. Call it 2.5ma. That's the maximum, and 40uA is the minimum.

So we can now look at the stacked case. This is two 100uF caps rated at 350V. The CV is then 100E-6 * 350 = 0.035. Leakage is then at least 0.0004A and at most 0.0004 + 0.0035A, or 0.0039A. The worst case voltage division is if one cap has the minimum leakage of 40uA and the other the maximum of 3.9ma.

This happens at the max rated voltage, so let's convert that to an ersatz resistance. For 350V and 40uA, the equivalent resistance is 350/40u = 8.75M. For 350V and 3.9ma, the equivalent resistance is 89.7K. If we don't do anything to even the voltages out, the total current from the power supply will be 450V/ (8.7M and 89.7K), or 51uA. The low resistance cap will have 4.6V across it and the high resistance cap will have 445V across it. The high resistance cap fails first, then the whole voltage is applied across the low resistance cap and it then dies from the bigger voltage. So we have to parallel resistors with each cap to keep this serial failure from happening.

We want the parallel resistors to force no more than 350V to happen across the high resistance (40uA) cap when there is a low resistance cap in the other half. We'd like it to be exactly split voltages, but that ain't gonna happen all the time. Each cap can have 350V max and not die, so we want to add equal resistors across both caps that make neither one have more than 350 out of 450V. Let's call the parallel resistance of the lower cap with one of these resistors R1 and the parallel resistance of the other cap with a swamping resistor R2.

So let's calculate two resistors that divide 450V into 100 and 350. By long familiarity with Ohm's law, we know that 150/450 = R2/(R1+R2)
or R1+ R2 = R2* 450/150. R1 + R2 is what sets the total current through the caps and resistors from the power supply. We have to pick our resistors so we don't eat too much power. We know this will be a bigger current than 51uA.

What happens if we use 220K? That makes the low resistor cap and 220K in parallel be 63.7K, and the high resistance one act like 8.75M || 220K, or 214.6K. The voltage across the high resistance cap is then 450 * (214.6/(63.7+214.6))=347V. The low resistance cap gets 450*(63.7/(63.7+214.6))= 103V. By blind luck, 220K will barely work, even for worst case caps.

Notice that one of the 220K resistors dissipates 0.56W, so a 1W is needed for both of them.

That's awfully thin margins. I'd probably pick a 180K or 150K value for the swamping resistors, in a belt-and-suspenders kind of move. That would account for caps not made by the one company I looked up that might have worse tolerances, and harder use. Lower resistors would be needed in case you used lower voltage caps too. If you were counting on two 100uF 250V caps like might seem clever, you'd have to dramatically lower the swamping resistors to avoid chain failures if one of the caps started going leaky.

[tictac]

There are a few unusual examples that seem to fit into this conversation.....

The re-issue Deluxe Reverb uses an odd combo of 220uF/100V - 100k network in series with a 47uF/500V - 470k for the first filter after the 5AR4 rectifier.

Strange brew but evidently it works, or maybe it's a way to use up those 220uF/100V caps laying around doing nothing!

TT

[sbirkenstock]

R.G., thank you for that very interesting answer!
So do you think it would make sense to put two 220K resistors in, measure the result and change the resistors that both caps get close to the same voltage?
Had my amp still open, it has two 270K carbon comp resistors and it is actually putting the voltage to half almost exactly.
I´ll certainly check on that on differnt amps in the future!
Stephan

[Ten Over]

Two electros in series can be modeled as two perfect caps with tiny resistors in series with each one (that's the ESR for the cap) and a large value resistor in parallel, representing the leakage resistance. Manufacturers won't specify the leakage resistance, as they try to make it zero, but know they'll fail at that. Instead they specify that it won't be any worse than X current when new and possibly Y after its expected lifetime. The point of swamping resistors is to make the voltage share between the two caps at DC conditions be no worse than the caps can withstand.

For some idea of what the numbers are, I looked up the specs on Illinois Capacitor (CDE) at Mouser. Their specs on their TTA line say that the leakage won't be any worse for caps from 160V to 450V than
CV * 0.04 +100uA for caps with CV (the product of capacitance and voltage) less than 1000 and
CV * 0.1 +40uA for caps with CV>1000.

ACK! What's all that mean? Well, you pick a cap - say, a 47uF/450V unit, then calculate CV. In this case that's (47E-6)*450, or 0.02115. Well under 1000 then. So the leakage is guaranteed to be less than 0.02115*0.1 +40uA = 0.00215A +40uA = 0.00255A. Call it 2.5ma. That's the maximum, and 40uA is the minimum.

So we can now look at the stacked case. This is two 100uF caps rated at 350V. The CV is then 100E-6 * 350 = 0.035. Leakage is then at least 0.0004A and at most 0.0004 + 0.0035A, or 0.0039A. The worst case voltage division is if one cap has the minimum leakage of 40uA and the other the maximum of 3.9ma.

Manufacturers try to make the leakage resistance as large as possible, not zero.

In general, when the manufacturer states the leakage current as 0.1CV +40uA, they mean "C" is the nominal capacitance in uF, "V" is the rated voltage in V, "40" is a number, and "uA" is the units when you are done with the math. So for the 47/450, the CV is 21,150 and you would use the the equation CV * 0.1 + 40 which gives 2,155uA. Interestingly, you used the correct equation even though your CV product indicated the other equation should be used.

I see no reason to suppose that the minimum leakage is 40uA.

40uA is 0.00004A, not 0.0004A.

[R.G.]

The re-issue Deluxe Reverb uses an odd combo of 220uF/100V - 100k network in series with a 47uF/500V - 470k for the first filter after the 5AR4 rectifier.
Strange brew but evidently it works, or maybe it's a way to use up those 220uF/100V caps laying around doing nothing!

Very strange. Perhaps they just believe the leakages will add up correctly, or maybe they have some testing done on in-house stock. It's hard for me to trust anything but worst case spec limits when all I have is the specification.

Then there's the real world where worst-case stuff doesn't bite you, as happens most often:

So do you think it would make sense to put two 220K resistors in, measure the result and change the resistors that both caps get close to the same voltage?
Had my amp still open, it has two 270K carbon comp resistors and it is actually putting the voltage to half almost exactly.

If the two caps are nearly balanced, you don't need to mess with it. You have two caps which are close enough together that they almost match in leakage anyway, so the nominal values work OK. For the case where you can actually measure and balance the real caps, you only need to worry about leakage drift over time. If one cap gets leaky nearer end of life, or is constantly cooked by being too close to hot tubes, you can run into this.

Manufacturers try to make the leakage resistance as large as possible, not zero.

Actually, I suspect that if they could make it zero, they would. It would be a tremendous advantage over their competitors with non-zero leakage. But I agree that the MBA-school position would be to make it as large as they can get away with. It's far too easy to make leakage so huge that they're not capacitors any more. They definitely would make their process as sloppy as they can get away with - but they will always try to just edge out the competitors. It's one of those delicate business-school races that happen.

In general, when the manufacturer states the leakage current as 0.1CV +40uA, they mean "C" is the nominal capacitance in uF, "V" is the rated voltage in V, "40" is a number, and "uA" is the units when you are done with the math.

Yep, I know C is the rated capacitance, V is the rated voltage. The physics position is that C is always farads. That did confuse me, and I did start to call C in uF, but I went with the physics definition. None of the makers' spec sheets I consulted said that C was in uF, or if they did, I missed it. I was mystified that I could not fathom why they called for CVs over 1000, other than if it was uF. In any case, a worst case scenario you'd just pick their worst units anyway.

So for the 47/450, the CV is 21,150 and you would use the the equation CV * 0.1 + 40 which gives 2,155uA. Interestingly, you used the correct equation even though your CV product indicated the other equation should be used.

Yeah, getting the units wrong in equations from manufacturer's datasheets has been an occupational hazard for me for at least 30 years. This gives the result that my outcomes were unnecessarily pessimistic, but certainly not fatal.

I see no reason to suppose that the minimum leakage is 40uA.

I do. It's right there in the manufacturer's equation: leakage is equal to mumble, mumble plus 40uA. Admittedly, many caps might beat that number, but the datasheet is what they will admit to, and they won't admit to leakages less than 40uA. Well, in addition, as you just said, they would work to make leakage bigger, so the 40uA they'll admit to is a reasonable place to say it is likely not any smaller. Right?

40uA is 0.00004A, not 0.0004A.[/quote
You are correct.

I'll run through the equations again and refine the results some more.

But tell me this - do you agree with the method for worst casing balancing? And if not, where are the holes in the method?

[Ten Over]

Yep, I know C is the rated capacitance, V is the rated voltage. The physics position is that C is always farads. That did confuse me, and I did start to call C in uF, but I went with the physics definition. None of the makers' spec sheets I consulted said that C was in uF, or if they did, I missed it.

Somewhere, sometime I've seen the units called out. Even if I had never seen it, "C" would have to be in uF because the largest capacitor/voltage combination in the 160-450V range is 47uF/450V. I'm certain they would not specify a range that none of their capacitors are in.

I see no reason to suppose that the minimum leakage is 40uA.

I do. It's right there in the manufacturer's equation: leakage is equal to mumble, mumble plus 40uA. Admittedly, many caps might beat that number, but the datasheet is what they will admit to, and they won't admit to leakages less than 40uA. Well, in addition, as you just said, they would work to make leakage bigger, so the 40uA they'll admit to is a reasonable place to say it is likely not any smaller. Right?

It doesn't actually say "40uA". 40 is a constant with no units. The "uA" belongs to the whole equation. There is nothing there to give any reason to think that the constant "40" is derived from the fact that the minimum leakage is 40uA. There are cases when the leakage is specified as the greater of a set value or the result of an equation. Here they are specifying a minimum and a maximum. But the set value is not even close to the constant in the equation, so it is clear that the constant has no relationship to the minimum leakage.

I have three different varieties of brand new 100uF/350V capacitors here, so I tested a couple of each. After a couple of minutes, the ic TTA's went down to 20uA, the Nichicons went down to under 10uA, and the generics around 50uA. The ic's consistently running well below the alleged 40uA's suggests that 40uA is not the stated minimum.

There seems to be some sort of confusion about resistance and current and bigger/smaller leakage current. And what it is that I said.

But tell me this - do you agree with the method for worst casing balancing? And if not, where are the holes in the method?

I agree with the worst-case explanation, but I think that the worst case has an even bigger spread than you think. I also think you may as well consider one capacitor to have no leakage and the other the maximum leakage to simplify the calculations because the results are going to be about the same and there will be no arguments about the minimum leakage.

[R.G.]

I agree with the worst-case explanation, but I think that the worst case has an even bigger spread than you think. I also think you may as well consider one capacitor to have no leakage and the other the maximum leakage to simplify the calculations because the results are going to be about the same and there will be no arguments about the minimum leakage.

I'd like to see your numbers on the worst case spread. Maybe I can learn something, which I love to do.

Can you include in there where you'd put a mimimum, if manufacturers try to get leakage higher, and what minimum you'd estimate, as well as why? I'd like to see what I missed.

[Ten Over]

I'd like to see your numbers on the worst case spread. Maybe I can learn something, which I love to do.

Can you include in there where you'd put a mimimum, if manufacturers try to get leakage higher, and what minimum you'd estimate, as well as why? I'd like to see what I missed.

0.1 * 100 * 350 + 40 = 3540uA
350V / 0.00354 = 98.87K
(98.87K * 220K) / (98.87K + 220K) = 68.21K
450V / (68.21K + 220K) = 1.56mA
1.56mA * 68.2K = 107V when using all significant figurers
1.56mA * 220K = 343V
Therefore, 343V across one cap and 107V across the other at the worst case.

Using 40uA instead of 0.00uA as the minimum leakage and correcting your math:
341V across one cap and 109V across the other cap.

So whether using an errant minimum leakage current or none at all results in nearly identical results.

Where I'd put a minimum: The manufacturer doesn't specify a minimum because nobody cares. Everybody is interested in the maximum. I have determined experimentally that these things can have very low leakage current. So low that zero is a good estimate for the purposes at hand as my numbers have borne out.

If manufacturers try to get leakage higher: Of course not and I said nothing of the sort.

What minimum I'd estimate: I don't estimate any minimum because I'm using zero for my worst case scenario.

You'd like to see what you missed: Frankly, I think that you are patronizing me, but I went ahead and responded to your post, anyway. You don't know who I am or what I know, but that's okay because I don't know who you are or what you know and just like you I don't care. It just isn't pertinent to the discussion.

[martin manning]

Here is a calculation made following the recommendation in a Nichicon app note, same arrangement as RG's, 2x 100u, 350V caps:

C1, C2 100 uF
V rated 350 V
Tamb 60 Deg C
Ileak Temp Coeff 2 f(Tamb)
Voltage Balance Rate 10%
Ileak Variation Coeff 1.4

Voltage Imbalance 35 V1 - V2
Ileak Range (1) 157 Imax - Imin (uA)
Rbal (2) 223 kΩ
Pdiss Rbal 0.67 W

(1) 3/10*(CV)*TempCo*IleakVarCo
(2) (V1-V2)/(Ileak Range)

As you can see, the balance resistor is more than twice as big. I added the power calc for Rbal, assuming the voltage could be V/2 plus the 35V voltage imbalance.

What if there was no way that there could be 700V across this stack? Can I just reduce the voltage assumption to 500V, and assume each cap nominally sees 250V? Then Rbal becomes 188k at 0.4W, and a 180k 1W resistor would seem to be adequate.

[Ten Over]

What if there was no way that there could be 700V across this stack? Can I just reduce the voltage assumption to 500V, and assume each cap nominally sees 250V? Then Rbal becomes 188k at 0.4W, and a 180k 1W resistor would seem to be adequate.

No.

Both the leakage current and the leakage current range are functions of the CV product. The applied voltage does affect both the leakage and the range, but it doesn't alter the CV product. So you cannot substitute 250V for 350V in equation 2-6 when determining the leakage range.

The numerator in the R=V/I equation is the amount of voltage imbalance that we are willing to tolerate, presumably at the worst-case scenario. They set it at 10% of the capacitor's voltage rating. If the applied voltage is only 500V, we can tolerate a higher imbalance and still not exceed the 350V rating. This increase in the numerator will result in a value higher than 222.7K for the balancing resistors when all else is equal.

We know that the leakage current decreases with decreasing applied voltage, but we don't know how the leakage range changes and it is the range that determines the necessary balance resistors. Whatever the variation is, they feel a factor of 1.4 will cover it. There is a tendency to assume that the range will decrease as the applied voltage decreases, but it ain't necessarily so. If it does, that would decrease the denominator and again increase the value of the necessary balance resistors.