Smoothing the bias supply?

[Uncle Daddy]

This just came up on a forum. Is there any value in using a bypass cap on the bias adjustment pot?
Thx.

 

[Jon Snell]

In my experience, calculating the tank capacitor value to allow a small amount of ripple can be advantageous, (exactly as the first capacitor on a C L C HT supply is used). It beats against any 100HZ hum on the main HT and has the effect of reducing any remaining hum in the output transformer.
Adding more capacitance, is of no consequance. (A bit like firing a load of parts at a car that won't start to find the issue was actually no petrol).
For decades, main manufacturers have chosen 10 or 22uF as the only smoothing, for that reason.

 

[YellowBoots]

This just came up on a forum. Is there any value in using a bypass cap on the bias adjustment pot?
Thx.
View attachment 1223632

Merlin seems to think so. Not only does it further smooth the bias output, but it also decouples the grid leaks from one another (I assume this means more gain is had from the preceding PI).

https://www.valvewizard.co.uk/bias.html

It would be interesting to determine if this extra cap had any effect on the perceived “sag” characteristics of the amp under heavy duty. I would hypothesize that the extra cap would reduce bias sag and have the effect of stiffening up the whole amp. How much, I have no idea. By extension of less bias sag, one may be able to bias the tubes at a higher current with less risk—mainly to avoid crossover distortion. All untested hypotheses on my part! Mainly I’m just try to jumpstart an interesting thread.

 

[Pete Farrington]

it also decouples the grid leaks from one another (I assume this means more gain is had from the preceding PI).

I think there would be a naturally formed signal V AC null point at the junction of the grid leaks, because their values are the same and the V AC at either end is equal but opposite.
Hence connecting a decoupling cap there kinda seals the deal but doesn't actually have any material effect on the AC signal analysis.
So as I see it, connecting a decoupling cap there can't reduce the loading on the preceding stage (phase splitter), so no increase in gain, sorry

It's an analogous scenario to the cathode bypass cap on a cathode resistor that shared between push pull output valves, or other balanced stages.
ie whilst the balanced stages are both in their class A area of operation, the V AC at the shared cathodes pretty much cancels out. The bypass cap can be removed without affecting anything much.
The bypass cap only decouples anything useful once the signal level increases sufficiently to push the output valves into their class B area of operation.

It would be interesting to determine if this extra cap had any effect on the perceived “sag” characteristics of the amp under heavy duty. I would hypothesize that the extra cap would reduce bias sag

I'm not exactly sure what's meant by bias sag?
My understanding is that grid current clipping at or near typical overdrive signal level causes signal rectification and pushes the grid more negative.

Grid Bias Excursion Calculator

www.ampbooks.com

In fixed bias, that occurs at the other end of the grid leak too, though it's somewhat reduced by the potential divider effect of the bias supply's effective output resistance and the grid leak resistance.
Hence if the voltage at the bias supply output and grid leak node is monitored, compared to its level at idle, it will become more negative when grid current clipping overdrive becomes prolonged.
So in fixed bias, that contributes to the bias excursion effect.
Increasing the RC time constants at the bias supply output, eg by adding / increasing smoothing capacitance, will tend to prolong the effect of bias excursion after the signal level subsides.

 

[2L man]

I think nowadays when electrolytes are not expensive two filter capacitor bias circuit is good choice!

When I use HV AC-0-AC for bias supply I always install one diode to each output which initially make DC "double smoother" than half wave rectifier does.

 

[King Fan]

Good info here, and easy to implement, as shown in the short lucid Merlin chapter linked by @YellowBoots :

"C2 is added to decouple the grid leak circuits from one another (when more than one valve share the same circuit) and simultaneously acts as an RC smoothing filter with P1. Since the bias supply draws very little current we only need one stage of smoothing in addition to the reservoir capacitor. Both capacitors are typically 10u to 100u -- they’re not critical."

 

[YellowBoots]

Hence connecting a decoupling cap there kinda seals the deal but doesn't actually have any material effect on the AC signal analysis.
So as I see it, connecting a decoupling cap there can't reduce the loading on the preceding stage (phase splitter), so no increase in gain, sorry

Yeah good point. This is assuming perfectly matched signals off the PI. Maybe a hi-fi builder is more interested in the decoupling due to complexities getting a perfectly matched signal. Also any hum induced by uneven power tube matching should be squashed right?

I'm not exactly sure what's meant by bias sag?
My understanding is that grid current clipping at or near typical overdrive signal level causes signal rectification and pushes the grid more negative.

I’m talking about the natural sag of the bias supply due to loading of the PT as the amp gets into Class AB up to maximum clean power. Is that not a thing? I thought that was one of the reasons we don’t bias up to 100% on fixed bias.

I wasn’t thinking pushing the amp all the way into grid current, but you are correct about bias shifts under those conditions.

 

[King Fan]

BTW, while I'm agreeing with Merlin and the brain trust here about that second cap, let me also say in smaller Fenders up to at least Princeton Reverb level, Leo's 1-cap solution is almost certainly adequate. At least in simulation, the residual ripple in a proper Fender board is very low, and if you upgrade your single cap to Doug Hoffman's (say) 47uF, logic, experience, and Merlin's 'very little' bias current also pitch in to say bias ripple should be a non-problem.

 

[GotA24Fretter]

You can model this in PSUD. Without a cap off the wiper and a 470 ohm resistor with a 100uF cap you get a sawtooth wave with about 550mV swing peak to peak.

This drops to less than 10mV when a 22uF cap is added to the wiper.

This makes an audible difference.

Across the bottom leg of the voltage divider like is shown in the picture is less valuable. When connected to the wiper it is directly at the bias leak resistor junction, electrically speaking. Look at the image above from Merlin's page or the DRRI schematic for inspiration.

 

[Lowerleftcoast]

A secondary reason to use a cap near here. Some post PI master volume designs (frondelli) will not let the volume drop to *0* when there is not a path for AC signal to easily find ground. The cap(s) allow the signal to drop out.

 

[Pete Farrington]

Also any hum induced by uneven power tube matching should be squashed right?

I'm a bit stuck trying to work out a mechanism by which that might be the case. Could you explain your thinking?

I’m talking about the natural sag of the bias supply due to loading of the PT as the amp gets into Class AB up to maximum clean power. Is that not a thing?

Ok, compared to HT sag, any sag of the bias supply will tend to be a related but secondary, lesser effect. ie bias sag will invariably tend to be a lower % reduction than HT sag, especially so at the screen grid (anode current being determined by the relationship between control and screen grid voltages). So even without grid current clipping, the combined result of sag (HT and bias) will generally tend to be that the net instantaneous bias level will become a bit colder.

I thought that was one of the reasons we don’t bias up to 100% on fixed bias.

No, with signal, with most AB loadlines, anode dissipation will almost invariably tend to increase above its idle level. Aiken demonstrates how this takes place here
It is feasible to tweak parameters so that it doesn't *, but I can't think of any valve guitar amps that follow that model.
* see conditions bottom right p2 of https://frank.pocnet.net/sheets/093/6/6L6GC.pdf
and its accompanying operation characteristics charts at bottom of p9

Without a cap off the wiper and a 470 ohm resistor with a 100uF cap you get a sawtooth wave with about 550mV swing peak to peak.

This drops to less than 10mV when a 22uF cap is added to the wiper.

This makes an audible difference.

Indeed, adding a 2nd RC filter stage, even if the values seem fairly low, generally makes for a much more effective result than eg just making the reservoir cap much larger.

1-cap solution is almost certainly adequate

It's kinda stating the obvious, but there's little benefit to be gained from improving something beyond 'good enough'

 

[sds1]

According to LTspice that cap isn't doing too much there. Probe is at junction of R2/R3 with 10kΩ pot wiper at 50% (green trace=no capacitance, blue trace=68uF)

But if you move the 68uF cap to the wiper you get much less ripple:

 

[Pete Farrington]

According to LTspice

The traces look the same?

 

[sds1]

The traces look the same?

The green traces are without any capacitor (actually just a really small capacitance for simulation purposes), the blue traces are with the capacitor in circuit.

You can see the blue trace in the 2nd graph shows a very smooth DC output.

 

[Mr Ridesglide]

This is akin to something like the bias rollout on say a 5E6 A Bassman?

 

[King Fan]

You can model this in PSUD.

According to LTspice...

Thanks, gents. I like sims for questions like this. And great point about wanting the cap hanging off the wiper (as in Merlin, I think) rather than the 'tail' lug as in the web pic we saw first.

I'll note that Doug Hoffman, way back in the Dot Matrix age (late Pleistocene), ended his article about Fender's "typical" (1-cap) bias circuit by showing the "better" bias board he put in amps like the Plexi or 5F6a.

Trying to learn more about simulation, I set out to draw Merlin's simple bias circuit (post #5) in Falstad. I'm a noob, and *I musta done something wrong* -- my ripple is a half wave, not a sawtooth. Don't hesitate to point out my mistakes.

Notes: For simplicity, I used his 10K pot and minimal 10uF caps, but left out his 'emergency pulldown resistor' (R3). I made my max cold target about –50V; he notes this would be enough even for something like 6L6s or KT88s, and I figured a big max negative voltage would generate a big ripple. But I get only 66mV ripple (on my suspect half-wave) and can't get it above 100mV at max neg voltage.

But let's ignore my training-wheels sim. I wonder how much bias ripple it takes to be audible? And audible how -- does it automatically sound bad?

 

[GotA24Fretter]

Extraneous background noise (2x mains, but I often hear hum and hiss too). There's some degree of common mode cancelation, but it's imperfect. In my opinion better to not fully rely on the cancellation if you can reduce the issue with a single small cap.

 

[Uncle Daddy]

Last night I was looking at some images of the bias supply on the Chris Stapleton Princeton when I noticed this extra cap on lug1 of the intensity pot, where the wiper from the bias pot connects. Can I assume it does a similar job in smoothing the bias supply? Is there an advantage in connecting it from the bias feed to ground, as opposed to bridgeing the tail resistor on the bias pot (original post)? No schematic unfortunately.

 

[Pete Farrington]

Can I assume it does a similar job in smoothing the bias supply?

I think so.

Is there an advantage in connecting it from the bias feed to ground, as opposed to bridgeing the tail resistor on the bias pot (original post)?

I think there will be.
In the post 1 arrangement, the full '2 stages of filtering' effect will only be available at the hottest bias setting. As the pot is rotated from cold to hot, the filtering will probably become worse, at ghe colest setting the extra cap won't be providing much benefit.

Whereas connected as per post 18, the extra cap is always providing the most benefit that it can.
The filtering will still be best at the hottest setting though. At the coldest setting, it'll revert to being single stage, but at least the extra cap will then be in parallel with reservoir capacitance, thereby still reducing ripple.

 

[sds1]

Is there an advantage in connecting it from the bias feed to ground, as opposed to bridgeing the tail resistor on the bias pot (original post)?

The simulation outputs in post #12 answer this question.