Selecting Power Supply Dropping Resistors for Interstage Isolation

[joelap]

I've searched a few forums and cannot seem to find the technical basis behind selecting power supply dropping resistors for the sake of avoiding interstage loops.  I am aware of other reasons for selecting certain values, namely to create a LP filter below 60/120Hz, but I often see it mentioned that the resistor value must be sufficient enough to avoid certain issues with loops but I have yet to find the theory behind where the breakpoint exists between sufficient/insufficient.  I'd imagine it has to do with the current draw of the various stages of the amplifier, no?  I've been working on a bass preamp that I want to have a very tight power supply but also high headroom, I am confined to the highest DC voltage out of my PT after the rectifier to about 310V with ripple, after two stages of filtering I'm down to about 295V or so using 390R dropping resistors and 100uf capacitors throughout.  I have subsequent filtering stages to this, and I'm trying to balance keeping the voltage high for headroom reasons and ensuring I have a large enough value for each dropping resistor.  I would be willing to sacrifice headroom for less noise if I had to make a choice, but only to a certain point.

Thanks,

Joe

[FYL]

You should set RC time constants at least a decade below the lowest frequency to be reproduced, say 7 Hz or below for a guitar amp.
The formula is Fc= 1 / (2*Pi*R*C), common values are 4K7 and 22µ for a nice 1.5 Hz Fc and app. 10V voltage drop for a typical 12AX7 twin triode. Don't worry about headroom, plate voltage is usually more than adequate.

[Fresh_Start]

Just looking at the formula, you can make a tradeoff between capacitor value and resistance value to get the same cutoff frequency.  However, doesn't the absolute capacitor value matter in terms of bass response and/or sag?

IOW Marshall seemed to go with BIG (for its time) capacitors and smaller dropping resistors compared to Fenders.  That difference in the power rail design is part of the difference in sound too, isn't it?

Cheers,

Chip

[FYL]

Just looking at the formula, you can make a tradeoff between capacitor value and resistance value to get the same cutoff frequency.

 

Yes, but larger Rs = higher voltage drop, larger Cs = higher cost.

However, doesn't the absolute capacitor value matter in terms of bass response and/or sag?

The TC conditions response, so a small cap and a big R will give the same result as a big cap with a small R. There's no sag when it comes to small signal tubes working in class A...

IOW Marshall seemed to go with BIG (for its time) capacitors and smaller dropping resistors compared to Fenders.  That difference in the power rail design is part of the difference in sound too, isn't it?

Could be, but the main voicing differences are caused by other components such as the slope R in the TS, smaller cathode decoupling caps, etc.

[joelap]

All, thanks - When I do my power supply calculations, I assume worst case scenario for the values of the 20% tolerance caps.  So with 390R and 100uF caps, using 1% metal film resistors I calculate the TC at 80uf just to be overly conservative.  1/(2*pi*390*(80e-6)) = 5.1Hz.  Advice given is to set the TC a decade below the lowest frequency, in this case that would require 4Hz for a standard tuned 4string bass, but since I am going low I will use 2Hz.  1/(2*pi*2Hz*(80e-6)) = 994.7R, so 1k.  At 1k, I have the following supply voltages going to the valves according to PSUD, assuming 5mA per triode stage:
V1=237.58
V2=249.70
V3=271.86
From here, I have the anode resistor also dropping the voltage further to the plates.  As I have it built currently utilizing 2 tubes only (no FX loop currently) I am supplying 280V and 295V to V1 and V2, respectively.  

So perhaps there is some value in setting my time constant lower, but I am seriously dropping the voltage I am supplying to the tubes.  Maybe it is acceptable?  FWIW, as it is currently built I have a nearly flat response down to 10Hz measured on an oscilloscope, and I am only using 100R dropping resistors, not the 390 I am intending on eventually using.  I do have a VERY SLIGHT hum I am looking to eliminate, maybe this is the issue?

Still looking for info on where the break point is in resistance value for the purpose of avoiding loops.  Any ideas?

[joelap]

There's no sag when it comes to small signal tubes working in class A...

Understood, and if all tubes are operating in class A would this negate any benefits of a choke on the power supply?  Understood I don't need one and can use a purely capacitive power supply, but will a choke only benefit amps running at higher currents?  Typical placement is before the screens, and with the exception of the screen supply its all preamp tubes being fed from the choke.  Cost, for me, is not an issue in determining benefit.  I planned to include one and intend to install one just to see its effects, if any, but am curious as to what I should expect as I wait for my Mouser order to arrive.

[tubesornothing]

You concern about headroom in the preamp stages is not a worry.  All you need to do is work backwards from the power stage.  Tpical power stages will want anywhere from 20-50V p-p.  You should be able to do that with 150V B+ on the stage preceeding the power stage.

So tune your preamp B+ for max quiet, then adjust gains to get the gain you want at the final stage before the power stage.

[joelap]

TON, thanks for the response.

[Fresh_Start]

There may not be any sag inherent to the preamp portion of the amp.  I understand that preamp tubes run in Class A and have very little current draw anyway.  However, there may be significant sag in the power amp which draws on the power rail before the preamp.  My uneducated guess was that larger caps in the preamp section of the power rail might reduce the voltage swings in the preamp resulting from big current draw(s) from the power amp.

Again, just an uneducated guess.  I don't have an amp to test this theory on at the moment.

Chip

[eleventeen]

If we are talking about wringing the very last bit of perceivable hum out of a build, what about the argument that physically placing preamp tube power supply caps closer to (or right at) the preamp tubes (instead of at a central location) might achieve the same/more than scientifically selecting the PS dropping R's?

I myself have not built anything that used that method versus a more conventional "filter cap central" method. My intuition is that it would produce detectable improvement.

[HotBluePlates]

... I have the following supply voltages going to the valves according to PSUD, assuming 5mA per triode stage ...

It would probably help to see your amp schematic, since designing a power supply in isolation is a futile exercise.

I say that because I am quite surprised by the 5mA draw per triode. Typical guitar amps might be more like 0.5mA to 1.5mA, and usually on the lower side. You might not be building "typical" but unfortunately, we can't tell if this is an error without more info.

From here, I have the anode resistor also dropping the voltage further to the plates.  As I have it built currently utilizing 2 tubes only (no FX loop currently) I am supplying 280V and 295V to V1 and V2, respectively.  

So perhaps there is some value in setting my time constant lower, but I am seriously dropping the voltage I am supplying to the tubes.  Maybe it is acceptable?

ToN hit it on the head: There are preamps with 300v on the plates of a triode, and others with 100v or less. Which is "best" or correct depends on a lot of other factors, but most importantly how big the signal must eventually be to drive the output tube(s). So we need to know more about the output section to know how to configure the preamp.

As an aside, it is common for a preamp triode to have a plate voltage that is near 1/2 to 2/3 of the supply voltage at the filter cap which feeds it. So significant voltage drop across a plate load resistor is fine, and gives a class A triode room to swing the plate both positively and negatively.

I am only using 100R dropping resistors, not the 390 I am intending on eventually using.

The reactance of the 80uF cap equals 100 ohms at 19Hz. The fits our rule if your lowest frquency is 190Hz, so hum is likely getting through. With the draw of two complete 12A_7's at your high-current estimate, that's 5*6 = 30mA -> 30mA * 100 ohms = 3v of drop for an entire preamp. That's probably typical in solid-state, but very unusual in tube circuits. A drop from 450v output tube plate to 440v output tube screen to 380v phase inverter supply to 310v preamp driver supply to 260v preamp supply would be boringly normal. Big voltage drops trough the supply are common in tube circuits for a variety of reasons, but often because you feed big output tubes and small preamp tubes from 1 supply, and each likes to eat a different supply voltage.

... if all tubes are operating in class A would this negate any benefits of a choke on the power supply?  Understood I don't need one and can use a purely capacitive power supply, but will a choke only benefit amps running at higher currents?  Typical placement is before the screens, and with the exception of the screen supply its all preamp tubes being fed from the choke.  Cost, for me, is not an issue in determining benefit.

Chokes help knock down ripple without dropping a lot of d.c. voltage. For big power from moderately-sized output tubes (6L6, EL34) you want to run class AB; that raises plate voltage and peak signal plate current, but also needs the idle current knocked down to keep the tubes from melting at idle. Often, the screen current varies greatly between its idle and full-power level. A chooke feeding the output tube screens keep the screen voltage from dropping during full-power conditions, which would then limit plate current and output power available from the output tubes.

Chokes have their greatest economic advantage (benefit/cost) when used in this situation. You could tack on more chokes to your power supply, but other points in the supply don't necessarily need minimal d.c. voltage drop, and chokes can radiate hum into other circuits. So fewer chokes usually has a cost and hum benefit, and series resistors often provide a needed d.c. voltage dropping function.

But again, I can't give specific recommendations without seeing a specific circuit.

[FYL]

I am seriously dropping the voltage I am supplying to the tubes

You may either use a ladder (serial nodes), a tree (parallel nodes) or serial + parallel approach if you feel that your B+ is too low, but any 12A*7 with more than 120 V will be fine for preamp duties, and, say, 200 V for the PI with big bad power tubes.

I do have a VERY SLIGHT hum I am looking to eliminate, maybe this is the issue?

Layout, cabling, hardware, tube selection, etc. could be the culprits.

Still looking for info on where the break point is in resistance value for the purpose of avoiding loops.  Any ideas?

The TCs condition everything. Don't use exactly the same values node to node and you'll be fine.

More than 80 years ago: http://www.clarisonus.com/Archives/Amp_Design/Motorboating.pdf

[joelap]

HBP, that was 5mA per 12AX7 not per triode, I typed incorrectly... when I measured current from the first filter cap I forgot when I started typing that I have a voltage divider in there.  In total I am drawing (with only 2 12AX7 in there right now) 20mA or so because of that voltage divider of 27k+2k2 drawing 280V/29200R = 9.59mA.  I am using that as a reference voltage for the AC heaters / some shielded wires.  Also, 2 of the stages are cathode followers, one of which has a lower Rk than typical, another common cathode stage has a low value of Ra contributing to the higher than typical current draw.  Also, input stage is bootstrapped, if that plays a part on the current draw of a common cathode stage I am unsure.

Also, this is purely a preamp so I'm guessing the benefits of the choke will be minimal at the end of the day... save myself 16 bucks and maybe leave it out.

FYL thanks for the link.  Next question based on that: Is it better to have earlier time constants "earlier" in the power supply?  IE first filter stage at 1Hz, 2nd at 2Hz, 3rd etc... or can they be staggered throughout with no ill effects?

[HotBluePlates]

In total I am drawing (with only 2 12AX7 in there right now) 20mA or so because of that voltage divider of 27k+2k2 drawing 280V/29200R = 9.59mA.  I am using that as a reference voltage for the AC heaters / some shielded wires.

That certainly works, but you could reduce current draw by scaling up those resistors to 270k/22k. Then current drawn through the divider is down to about 1mA. One would expect little current draw from this divider to the heater itself, so having high divider current for stable voltage doesn't seem to apply in this case.

I'm also guessing that your other threads are about the same circuit, i.e. a bass preamp. A 4v RMS output is still less than 6v peak, so you don't need a lot of supply voltage for any preamp tube to swing the voltage. You may need a low output impedance to drive the following source impedance and/or cable capacitance.

[FYL]

I am using that as a reference voltage for the AC heaters / some shielded wires.  Also, 2 of the stages are cathode followers

If you're using two CFs, the residual hum could be caused by poor cathode to heater isolation - most russky 12AX7s from the Sovtek stable are very poor here. Try to raise heater ref voltage by 70 to 100V.

Also, input stage is bootstrapped

A bootstrapped first stage can be quite noisy because of it's gain

Next question based on that: Is it better to have earlier time constants "earlier" in the power supply?  IE first filter stage at 1Hz, 2nd at 2Hz, 3rd etc... or can they be staggered throughout with no ill effects?

The first stage sees residual ripple, so you should use a big cap here, with proper grounding - this is the noisiest terminal in your preamp. Use 100 µ and up, the more the merrier. Other nodes are in a CRC network with lower ripple, so you can use smaller caps.

Seeing your schemo would be useful if you want more specific info.

[PRR]

The hum/buzz filtering should be computed for each stage. In general the tolerable hum gets lower as you go toward the input jack, and you can stand more 120Hz than 360Hz or 600Hz, so sequential (not spider) filtering is most economic. Figure your signal level at the plate. You should figure your PSRR but for tube triodes the PSRR is small and can be neglected. Pick a signal/hum ratio. 60dB or 1000:1 will be audible. 80dB or 10,000:1 is very clean on stage or in studio.

Small voltage drops are not the way to go. If you start from 300V, the "headroom" difference between 299V, 290V, and 270V is quite negligible, while the filtering/$ improves 10 or 30 times. If a 10% drop "matters", you should consider lower working levels or a much higher raw supply. Except when pushing over-large over-volted power tubes, there's hardly ever a reason to have signal over 10V. Any half-decent tube stage can do 10V at 5%THD with 100V, or 10V at 1%THD at 350V. 5% pure tube THD is hardly audible; if you expect point-something THD then you gotta use non-guitar hi-fi NFB schemes.

All stages after your Volume control, the signal level is at the designer's discretion. If you need 20V at the last stage to clip your power tubes, you don't care much what happens at higher levels: any softness is masked by power amp clipping.

The first stage, before any volume control, is not controlled by the designer but by the user. Normal guitar levels run 20mV to 500mV, and a 12AX7 will boost to maybe 25V. However a hard-plucked active bass through a booster pedal "could" ram 3V in, which should give 150V after the first stage, but will clip around 60V. Should the designer cater to "any" signal level"? (Even 70V PA outputs??!) Or is the user eXpecting "bad" things when over-plucking/boosting, and will be annoyed by lack of dirt?

Good 12AX7 git-amp design can use 300V-400V for the final driver, 300V-350V for the input stage, quite a lot less in most other stages, but extra-clean for input, volume-, tone-, and reverb-recovery.

Motorboating.... first leave out all the filter caps. All stages inject signal into the rail. Usually the last stage has the largest signal and does the most injecting. The rail signal leaks into earlier stages, is amplified, and encourages the last stage to inject more.



V3 has gain of 50 across 100K ohms. Without a rail cap and with a 10K dropper from the raw supply, about 1/10th of that appears across the 10K.

That signal sneaks via R2 back into V3, mildly attenuated by V2's plate resistance. However V3 inverts so the sneakback is NEGative feedback. Without some major complication, this is always stable. TIP: put your stages in pairs.

That signal also sneaks via R1 into V2 into V3, mildly attenuated by V1's and V2's plate resistances. Neglecting the loss at V1, we have gain of 50 * 5 or 250. And the inverter-inverter path is POSITIVE feedback. This is UNstable.

We must break the power rail, perhaps at "X", and insert enough signal-sneak loss to eliminate the forward gain throuh V2 and V3.

If all stages were DC-coupled, we'd be in a bind. To get 250:1 loss all the way to DC, we'd end up about 1 Volt. DC amplifiers do other things.

We are Audio men. The world stops at 20Hz and we don't pay to go lower. Therefore if "X" has 250:1 loss at-and-above 20Hz, we will be stable.

Actually the PSRR at V1 may be 3:1. So "X" only needs about 100:1 reduction at 20Hz.

And the full V1 V2 V3 string has main-path gain of maybe 50*50*50= 125,000 !! We will NEVER run three sequential stages wide-open! We have three stages because we have tone-control loss, mixer loss, cliper/filter loss, etc. If there's 10:1 tone-stack loss between V2 and V3, we only need ~~10:1 loss in "X".

I said we audio-guys stop at 20Hz. In fact we swing many ways. Guitar sound may "want" 70Hz or even 100Hz drop-off. Doing this with undersized coupling caps saves pennies. That also means less power-rail filtering which saves more pennies. OTOH, a cascade of many "20Hz" cut-offs adds-up to serious loss by 50Hz, so hi-fi systems often aim each stage below 1Hz. Also some tube coupling networks just don't cut-off much: a self-bias cathodyne has huge input impedance and the smallest (cheapest) cap may still pass to a few Hz. So there's some thought needed.

[PRR]

> voltage divider of 27k+2k2 drawing 280V/29200R = 9.59mA

Yowsa. If a 12AX7 had 1ma heater-cathode leakage it would stop working, 0.1mA would suck so bad you would find and replace that tube. You want reference impedance low enough to suck-out 0.1mA leakage without much drop; 1mA divider current is more than ample.

> Also, input stage is bootstrapped, if that plays a part on the current draw of a common cathode stage

AC-bootstrapped I am sure(?). How would that affect DC current?

But why bootstrap? 12AX7 has no maximum grid resistor. You can use 2Meg or more. The impedance of any wound pickup, or piezo/capacitor pickup without onboard preamp, won't be over 250K. That's 10:1 mismatch which preserves Voltage-Transfer conditions. And as mentioned, bootstrapping opens a world of other troubles.

Here's my advice. PLAGIARIZE!! Find a known-good plan and minimally modify it. All this stuff "can be computed or breadboarded..... but you are not the first to climb this mountain, look at what those before you did.

[joelap]

PRR, thanks for the informative posts.

Why Bootstrap?  Why not?  Hah.  Honestly the hum in the amp is very very slight and only when I really turn the volume up.  raising the power supply resistors will probably fix that.  And its giving me the tone I want... besides, plagiarizing is no fun.  It's more fun to work out the details from design to construction and work out the issues found.  More rewarding that way.  Using the bootstrapped stage in conjunction with a lower gain recovery stage after the tone stack sounds better to me... because I've built it both ways, with varying configurations and component values.  The values I have in there now sounds better than the other configurations I've tried.

And I know 27k/2k2 is not ideal, but its all I had that was >1/4W.  Sometimes you gotta make due with the parts you have on hand.

[HotBluePlates]

This thread may be dead, but a thought occurred to me...

It seems to me that you might have the tail wagging the dog, in that the power supply design is odd due to a desired high idle current through the triodes.

The thought is based on the raw numbers you gave. If the idle current is very high, then given typical plate load resistors, there is a large voltage drop from the supply node to the triode plate (possibly well over half the supply voltage). Two options then become automatic reactions: severe reduction of plate load/cathode load resistor values, and severe reduction of power supply dropping resistors. The first might result in low gain (depending on the value of internal plate resistance at the operating point acting against the external load resistor). The second could result in excessive power supply ripple or low interstage isolation.

There is a balancing act:
Lower tube currents result in higher internal plate resistance, and need higher load resistors to drop supply voltage down to the desired plate voltage. The reduced stage current allows larger power supply dropping resistors without excessively reducing power supply node voltage. "Big" dropping resistors result in a large RC value with modestly-sized caps.

I could be wrong, but I think you have been experiencing a cascade of problems to solve as a result of biasing the tubes to draw very much more idle current than is typical (or needed). Depending on the specific tube type(s), the actual internal resistance at idle and the load resistors you have used, you may be getting much less than optimal performance. It is hard to say for sure without a schematic. I do however wonder if there is a single mistaken assumption that you began with which has driven a number of other design decisions, and made things harder then they had to be.