PS recovery time calculations Q?

[Chuck H]

I'm thinking of some refinements to one of my designs. I've often thought a faster power supply would be a good thing as under heavy OD you can literally hear the voltage pump back up for a half a second after long runs and heavy chords. This design uses a resistor in line with the HV rail for a controlled amount of sag. I don't want to change the sag, just the recovery time. On Duncan PSUII I'm playing with Pi filter arrangement to relieve the main filter charging through the sag resistor and a lower overall uf amount. With a smaller uf filter at the actual plate supply for a bit faster recovery and the same ripple management. But PSUII is only calculating for mains ripple. My concern is heater induced hum. I know that ideally the AB1 circuit will cancel this for me, but it's never perfect and I'd like to be sure the plate filter is decoupling it from the rail before the preamp. One thing would be to know the circuits in play for figuring the lowest practical uf value. Would I parallel the power tubes internal resistance with the OT half primary Z as the impedance for the circuit???

Screen recovery time is a worse offender here too. A smaller uf cap there is working well in the simulation to reduce recovery time. With the Pi filter ahead of it I'm not seeing any offensive ripple as a result in the simulations, but are there other drawbacks to the smaller uf screen filter that should be considered?

And, if you've hung in thus far , keeping the plate and screen filter sizes down, I can use a somewhat larger uf filter for the node following the screens (LTP PI) to keep un decoupled hum or such off the preamp end of the rail too.

Any experience or general observations are appreciated. As always.

[uneumann]

I'm thinking of some refinements to one of my designs. I've often thought a faster power supply would be a good thing as under heavy OD you can literally hear the voltage pump back up for a half a second after long runs and heavy chords. This design uses a resistor in line with the HV rail for a controlled amount of sag. I don't want to change the sag, just the recovery time .....

Any experience or general observations are appreciated. As always.

This is a bit hard to follow, and you said "any observation" ...
If I get your gist, you want controlled OT stage sag, and fast recovery, and low ripple.
My take on this is a raw supply that is heavily filtered (to get rid of hum) and provide as stable a DC source as possible. That is followed by the OT/screen load fed through your variable sag resistor with just enough filtering to control your sag attack/decay. If you want the preamp to feel the same sag, you'll have to add another filter stage after that - again with enough capacitance to decouple and control sag decay. Alternatively, run your preamp off a parallel branch of the raw supply - but it will experience less sag that way.

As for heater hum... not sure how the PS is going to impact that much... so I'll leave that aside.

[Chuck H]

As for heater hum... not sure how the PS is going to impact that much... so I'll leave that aside.

That was probably my main concern. I haven't had that problem yet, but I've never used a relatively small value cap at he plate supply either. This design also uses a couple of split plate loads in the preamp, so anything on the rail is a bad thing and I just wanted to ask.

The amount of "sag" and where is occurs is fine. Calculated actually. So no need for branching the rail. Using the Pi filter and the lower screen grid uf's It looks like I can cut the recovery time in half for the screens. Not as much on the plates, but still better. I like the design as it was and have built four. One for me and three for customers. This is for my personal amp and tastes.

Thanks

[g1]

Traynor YCV40 uses what I would consider quite small value filter nodes (although raw B+ is regulated), so you may want to have a peek at that. It also uses quite small value screen resistors which I'm not sure if it is at all related (even for EL34 version YCV50). I asked about the screen resistors once and it was implied that with such low capacitance at the node, any significant amount of screen current would result in a major drop in screen voltage, thus limiting current. Or I may have completely misunderstood the explanation for the small screen R's.

[nickb]

I think you will run into a fundamental problem. The time constant of the power supply is dependent on the effective power supply resistance and the filter capacitance viz.

t = 2.2RC

R is pretty much the same as the power supply source impedance since the load i.e, the amplifier is much higher. So like it or not the sag and recovery times are very close to the same. No amount of tweaking R or C is going to help. A PI network changes the formula but not the fundamental problem.

Now just for fun, it would be possible come up with a (fairly simple) circuit that senses the drive level and switched in different series 'sag' and 'recovery' resistors feeding a new filter capacitor:



This would give you control over sag and recovery times independently. To make the recovery time faster than you have now, you have to find a way of decreasing the supply impedance before these switches. The solution would be a regulated PSU. So, all in all, it is possible but perhaps a little more complicated than you might wish.

It would be possible to conceive of all kinds of refinements like settable sag voltage, linear control rather than switches, a different method of operating the switches, the switching and time constants could the rolled into the regulated PSU etc.

[Chuck H]

I get what you're saying (though I don't know all the formulas). This, I assume, is why reducing capacitance DOES decrease the time. Of course, the side effect is that there will be less reserve current available for dynamics. Less time to charge=less time to drain. There's probably a balance that will sound best to me and I intend to find it. I don't actually want a "fast" amplifier. That's why I put the sag resistor in the design in the first place. I just want to experiment with making it less slow. But this is a good exploration of the aspects just the same

[trobbins]

Are you driving your amp with a step change from overdrive to low signal drive? Most natural sources of sound that would push an amp in to overdrive, such as a heavily strummed guitar string, would have a decay time that would probably dominate any power supply rebound - unless of course you were in to a kill switch.

[Chuck H]

Are you driving your amp with a step change from overdrive to low signal drive?

Not sure what that means.?.

Most natural sources of sound that would push an amp in to overdrive, such as a heavily strummed guitar string, would have a decay time that would probably dominate any power supply rebound - unless of course you were in to a kill switch.

Well if I let notes and chords decay naturally the transition would be smooth. But we don't always do that when playing. If you simulate common and popular power supplies on PSUDII you'll see that there's a recharge of up to nearly a second for some designes to recover from heavy current. This is normal. I'm just trying to refine the old standard to be more friendly with modern use. Short of adding voltage regulation though. I'd like it to resemble an improvement on the classic designs. Not a technical recreation of the ideal, if you get my meaning.

[trobbins]

A step change would be say two synthesizer notes played sequentially without any delay between the notes, and without any decay on the trailing edge of the first note - where the magnitude of the first note pushed the amp in to overdrive, and the amplitude of the second note was soft but discernible. When the first note finished (abruptly with no decay), then the second note would initially have an extra character from the rebounding power supply voltage.

[loudthud]

There are many factors that can come into play. If the filter cap at the OT node is too small, bass response suffers. You start to see a half wave rectified version of low notes on the B+. Recovery time at that node will be slow if there is too much resistance. A rectifier tube has a nonlinear resistance. The resistance drops as the current increases. But it may take several tubes in parallel to get really fast recovery.

For a given amount of ripple suppression, an LC filter will have about half the recovery time of an RC. But don't get stuck on the chokes typically used in guitar amps. Something below 1 Henry can do wonders in the right place. The math gets complicated real fast. Use a modeling program. But the hard part is tuning the B+ to the circuits in the amp. A Trainwreck Express at 450V B+ just doesn't have the magic of one with B+ lowered to 395V. You gotta find the sweet spot.

It seems to me that really fast players don't like saggy amps. They might as well go solid state.

[Chuck H]

A step change would be say two synthesizer notes played sequentially without any delay between the notes, and without any decay on the trailing edge of the first note - where the magnitude of the first note pushed the amp in to overdrive, and the amplitude of the second note was soft but discernible. When the first note finished (abruptly with no decay), then the second note would initially have an extra character from the rebounding power supply voltage.

Well then... That was exactly what it sounded like and "I" made the mistake of thinking it may have been a reference to an electronic circuit. Sorry to put you through what is an obviously pedantic exercise by asking you to describing it.

[Chuck H]

Thanks LT. I don't plan on changing to a choke though. It would mean too many other changes to an amp I already like well enough. But, again, good info on the aspects of the issue.

I wasn't aware that recto tubes have lower resistance at higher current! An interesting twist to the issue of simulating the tube effect with common resistors.

To all. And just in case I haven't been clear... I like the voltage drop at current as it is. It allows me to get more control from the guitar with technique and the guitars volume control. It makes the lead tones compress in a good way while I can still get some dynamics on cleaner tones. I'd just like to hear it with a little faster recovery AS WELL AS the voltage drop at current I have now. I've been able to simulate some design changes that should cut the time constant for the screen in half (the biggest offender in this issue) and at least a bit shorter on the plates. I'll have to see if LT's caution about the LF applies with listening tests.

If I seem resistant to changes, I'm not. I really appreciate the ongoing commentary and there are bits to consider throughout. So thanks again for all the responses so far and I'm still open to any more commentary. I just hope I didn't overstate my goals in the OP as seems like some suggestions amount to driving this small nail in with a sledge hammer But technical discussions can be that way, so it's all great.

[Enzo]

Just thinking aloud (or whatever this process is). Your whole B+ chain is sagging as you have it, I assume. Have you considered maybe separating the output stage B+ from the preamp string? Branching paths. I guess where I am going is have you monitored B+ drop and recovery stage by stage (well, node by node really) to see what the B+ downstream is doing? Possibly some of the recovery time involves the whole B+ string recovering rather than just the output parts. Might this be a place for a diode after teh screens to prevent the peamp B+ from being dragged down as much? Just a thought.

Or, I assume your sag resistor is in series with the entire supply? Would it work to have the resistor in series with the OT CT lead? That would leave the rest of the B+ unaffected, or at least less so.

On the other hand I am tired and might not be thinking this clearly. I need to go home, cook a pack of ramen and have a beer.

[Chuck H]

Thanks Enzo. Branching the HV rail was proposed earlier and I dismissed it. Maybe I didn't give it enough consideration since this is the second time it's been brought up. So thanks also to uneumann for introducing the idea.

I think I resisted because this amp derives it's (considerable) clipping starting with the power tubes. I just figured that would be the place to target, but indeed it there is some dramatic affect happening in the preamp that should be examined as well. Obviously that is where the recharge time will be the longest. I'll check out how much sag and how long the recovery is for the preamp as part of this process then.

[loudthud]

There are a couple of different mechanisms that cause preamp stages to compress. You can control some of them with subtleties in the design. One cause is the simple second order distortion that these stages produce before clipping. If you plot RMS output Voltage as a function of RMS input Voltage, you don't get a straight line. Perhaps a half dB fall in gain from small to large signal before clipping. There is also a small drop off in gain caused by B+ reduction. Again, it's only a fraction of a dB, but these things add up. An unbypassed cathode resistor reduces both effects. The so called cold clipper stage hardly compresses at all, aside from what the clipped wave it produces does to the next stage. The small signal gain is held fairly constant.

Bias shift also has an effect on gain. Think about the plate curves of a triode with a load line. As the operating point moves to the right caused by sag coupling of the previous stage, the gain goes down because the curves are closer together. Split loads might be problematic. The split load reduces the signal but not the bias shift form the saggy B+. A Voltage divider after the coupling cap reduces both and a smallish coupling cap attenuates the sag portion of the signal.

But preamp Voltages move pretty slow. Full sag takes a couple of seconds and recovery just as long. In general the lower the B+ to the preamp, the more it will compress. So be mindfull of this when you select those dropper resistors. You may need a smaller resistor and a larger cap.

A couple of cases to point out. A Vintage 5F6a has a smallish filter cap for the preamp, 8 or 10uF IIRC. The reissue had one or two 22uF caps. Some players who have played both say the didn't get it right on the reissue. The TW Express has lots of filtering on the preamp. It's preamp doesn't compress much at all in my experiments even though the preamp B+ sags quite a bit. Most of the compression is at the screen node. It's not so much gain compression as it is clipping compression. Most of the compression occurs after hard clipping begins. The main B+ is fairly stiff with lots of filtering and low ripple. Seemingly wildly divergent designs yet both highly regarded. Could there be more than one secret formula?