Stick to the topic gentlemen. Off topic posts have been removed.
MOSFET capacitance are not much of an influence on a cap-muliplier.
After all, the MOSFET gate has a µF-level cap to ground, just to make it a cap multiplier; and the output to ground has a similar value of cap - so 1nF or so of Coss or Ciss can't do much harm, since they are small compared to the external C values.
A well-chosen FET with a few tens of mA of standing current, and 20-40V of voltage overhead will give a small-signal output impedance of the order of 1Ω - flat from 10Hz to 50kHz, and the (bode) phase almost untouched.
Although a closed-loop regulator can get lower impedance, adding it in series with moderate impedance load (the output stage of an SE amp) means that the transfer function and phase shifts end up in the output. IME, this is easily audible (try for yourself, and compare even to a passive solution).
With SE valve amps, a series-regulator makes the system equivalent to a valve amp in series with a transistor amp. Almost always, this means a transistor amp which has not been characterised for distortion in the output current (since current is the dynamic part of the output).
NAIM NAP250 power amps used a series regulator based on the amplifier stage itself, using the same power transistors - and for good reason.
But translation of this idea to a 300B-SE is not going to work elegantly.
No such problem with a FET multiplier - since it acts as a source-follower, which most folks already know about from the improvement in grid driving of power valves. No loss of timbre or stereo space with these at all - and many improvements to to the greatly reduced blocking distortion and low drive impedance.
With a cap-multiplier, I hear no degradation compared to the virtues of a LCLC supply, and the bass of the cap multiplier is naturally much better.
Rod Coleman words are words of wisdom.
It is so cool to have you in this forum Rod. You truly rule. Have the greatest technical respect for you.
Art
It is so cool to have you in this forum Rod. You truly rule. Have the greatest technical respect for you.
Art
Were there ever any good high voltage BJTs with a decent linear area? I looked at the few made today and they're all for switching.
With MOSFETs you have to be careful that you don't blow the gate oxide. So you need to clamp Vgs with a zener or similar. Unfortunately, this also means that if the input voltage drops below the output voltage, the capacitor in the cap multiplier will discharge through the zener. The cap multiplier won't have any ripple rejection while the input voltage recovers. This means that you get hum every time the input voltage dips, for example when a heavy load turns on. I found that out the hard way. It's pretty annoying to have your amp go "hummmmm" every time the fridge/furnace/whatever turns on.
One solution to this is to drop significant voltage across the cap multiplier, but that defeats the purpose of using it in my opinion.
A proper regulator is a better option. I'm obviously partial to my 21st Century Maida Regulator but there are other options as well. One could geek out and use a tube for the pass device. The 6080 would be an option.
Tom
Tom,
the STF3LN80K5 that Ale recommends is equipped with protection diodes. I am not sure what kind of service you have that would cause such prodigous voltage drop when the AC turns on but I think I should be ok with that. What happens at turn off? Does it 'burp' -off?
What is 'significant' for you? How much are you dropping in your 21st Maida? I think I would be dropping enough not to cause issues in the preamp. Hopefully. 🙂
the STF3LN80K5 that Ale recommends is equipped with protection diodes. I am not sure what kind of service you have that would cause such prodigous voltage drop when the AC turns on but I think I should be ok with that. What happens at turn off? Does it 'burp' -off?
What is 'significant' for you? How much are you dropping in your 21st Maida? I think I would be dropping enough not to cause issues in the preamp. Hopefully. 🙂
Tom; any "proper" regulation would have the same behavior, it can't output more voltage than gets at it's input. And to cover voltage dips caused by fridge/furnace/whatever anyway it needs to drop enough voltage.
I've found that for tube amps one source follower is more than "proper", i.e. adequate. If you feed the gate by a filtered voltage, you may call it "a capacitance multiplier". If you feed it by a regulated reference voltage, you may call it "a voltage regulator".
The devil is in details.
For proper filtering you need a good time constant for feeding the gate of the source follower. But it does not mean that you need a big cap there that discharges through Zener. Some 1 uF film cap and few hundred kiloohm resistor would be more than adequate. However, you need some voltage divider, either shunting the cap by few megaohm resistor, or adding it before the RC, as a bleeder. For example, you are dropping 15V on the FET at nominal voltage. Let voltage between gate and source be 5V when the source follower works (roughly), that means you need a voltage divider that supplies 10V lower than the rectified voltage. So, if the rectified voltage is say 310V, take 10k and 300K in series. From their connection goes RC, say 1 MOhm and 1Uf. From 1 Uf 1K gate stopper to the gate. And one Zener between the gate and the source, to protect the FET. If it does not have one inside, add one diode like 4007 reverse-parallel between source and drain. But some FETs have already a reverse diode and a Zener inside.
Thank you all for the interventions.
Thank you Pano for setting everything.
It seems that both @Rod Coleman and @Wavebourn suggest the use of a simple capacitance multiplier.
I like the simplicity of that circuit and the capability to have a soft-start with it, even if the ripple will be more.
In case of tubes like GU50 that need a much lower voltage on g2 (around 200-250 vs 400-450 V on anode in SE "pentode"), would you prefer to have two cascoded psu (EG 220+220 Vdc), each with its own capacitance multiplier, or a single PSU with two capacitance multiplier in series, the first for the anode and the second for G2?
What I see in the first option is more safety for the screens, that even if the mosfet fails closing the circuit (that seems less probable than opening the circuit), g2 will be safe. Downside is more ripple. The second option has opposite pro and cons.
Thank you Pano for setting everything.
It seems that both @Rod Coleman and @Wavebourn suggest the use of a simple capacitance multiplier.
I like the simplicity of that circuit and the capability to have a soft-start with it, even if the ripple will be more.
In case of tubes like GU50 that need a much lower voltage on g2 (around 200-250 vs 400-450 V on anode in SE "pentode"), would you prefer to have two cascoded psu (EG 220+220 Vdc), each with its own capacitance multiplier, or a single PSU with two capacitance multiplier in series, the first for the anode and the second for G2?
What I see in the first option is more safety for the screens, that even if the mosfet fails closing the circuit (that seems less probable than opening the circuit), g2 will be safe. Downside is more ripple. The second option has opposite pro and cons.
No need to worry about the reverse diode (output to input) - all silicon vertical power MOSFETs have an antiparallel body-to-drain diode, rated for the full drain current level, ready made for this task.
As Anatoliy says (good to see you here again) a divider is mandatory for a multiplier. One extra reason: the FET must be kept away from the triode region of the curves - the low-VDS zone where VGS-vs-ID is influenced by VDS. Ignoring this point gets you extra unexpected output ripple.
There's no need for concern over dropping some volts for this purpose - we already need a good heatsink to cover the startup thermal pulse, when VDS starts low.
As Anatoliy says (good to see you here again) a divider is mandatory for a multiplier. One extra reason: the FET must be kept away from the triode region of the curves - the low-VDS zone where VGS-vs-ID is influenced by VDS. Ignoring this point gets you extra unexpected output ripple.
There's no need for concern over dropping some volts for this purpose - we already need a good heatsink to cover the startup thermal pulse, when VDS starts low.
The ripple can be very low; fractions of a mV are not difficult, even at 100mA loads. And since there is no longer any need to consider a vacuum rectifier, large capacitors can be accommodated in the Raw DC, reducing it further.
Large pH-neutral electrolytics like the Kemet ALS are higher quality, and retain their performance for much longer than cheap parts.
Yes, MOSFETs almost always fail short circuit.
I rebuilt an old amp for a friend. It originally applied 600V to the anodes AND screens of EL34s.
I added a choke, to make a choke-input Raw DC for the screens ca. 400V, followed by a cap muliplier to drop it to 370V
The anode got the same 600V. No need for separate transformers: both supplies on the same winding.
It was certainly much quieter, and the EL34s were safer.
My cap-muliplier board is getting an overcurrent trip - latches OFF if tripped, have to cycle the power to restart it. Programmable trip current.
I think it's maybe handy to explain a bit more context?
Because all we know at this point is that you want some kind of regulator for a SE amplifier?
Are you specifically looking for ultra low noise and PSRR?
Or does it just has to suppress the ripple adequately?
Same for the actual regulation itself.
The regulation of a capacitance multiplier isn't amazing.
When just resistors are being used it also won't have a fixed voltage. Meaning it will follow its input voltage.
I think this is also where previous discussions got confused by people.
Yep - if the voltage MUST be tightly controlled (±0.1%) a capacitance multiplier alone won't deliver and some kind of regulator is in order.
However, if you are OK with your high voltage power supply output wandering up and down a few volts in sympathy with the voltage from your wall socket, then a capacitance multiplier might be just right. It can reduce ripple to very low levels, deliver a quiet power supply, and a MOSFET capacitance multiplier has a nice, low output impedance - not the absolute lowest, but still pretty handy. (Plus a capacitance multiplier placed before a regulator can help get the best performance from the regulator.)
Well with a properly regulated power supply, we can only go as low as the lowest mains dips.
Unless we are gonna use an active SEPIC/buck-boost power supply, or put an active PFC stage in front, but let's not go that way.
So another way of looking at it, is that you will get just a little more headroom if you don't regulate so tightly.
The only downside of that approach is that some bias current might shift a little, although technically that can also be compensated for.
It's important to know before coming up with ideas how deep somebody wants to go into that rabbit hole?
Which is my question to @zintolo
Sure we can get a super well designed voltage regulator with the lowest noise, best PSRR and uber tight regulation.
But if somebody is not really looking for that, that's a lot of work and trouble for very little gain.
Besides the discussion how significant of an (subjective) improvement that might be.
But as I mentioned before, the only way of answering that question technically, we have to know exactly what kinda of amplifier and performance we are talking about.
Otherwise there is no technical objective way of saying anything about its significance.
This is why I suggested to just simply start with a capacitance multiplier first.
See how that performs, feels and meet the expectations.
You can always expand from there! 🙂
Unless we are gonna use an active SEPIC/buck-boost power supply, or put an active PFC stage in front, but let's not go that way.
So another way of looking at it, is that you will get just a little more headroom if you don't regulate so tightly.
The only downside of that approach is that some bias current might shift a little, although technically that can also be compensated for.
It's important to know before coming up with ideas how deep somebody wants to go into that rabbit hole?
Which is my question to @zintolo
Sure we can get a super well designed voltage regulator with the lowest noise, best PSRR and uber tight regulation.
But if somebody is not really looking for that, that's a lot of work and trouble for very little gain.
Besides the discussion how significant of an (subjective) improvement that might be.
But as I mentioned before, the only way of answering that question technically, we have to know exactly what kinda of amplifier and performance we are talking about.
Otherwise there is no technical objective way of saying anything about its significance.
This is why I suggested to just simply start with a capacitance multiplier first.
See how that performs, feels and meet the expectations.
You can always expand from there! 🙂
With a MOSFET, the load regulation is more-or-less dependent on the Rds, so that can be pretty tight.
(very much simplified).
However, as mention before, they are not as hardy as regulator transistors.
Also the power dissipation will be more.
I understand that this is gonna be a single ended design, so the load regulation is pretty constant to begin with.
So in that sense it doesn't really matter what we use.
Plus I think with using a Sziklai Pair instead of a Darlington we can get a decent enough load regulation as well as a lower dropout voltage.
You do have to simulate this though, because the total response will have a little dip.
Ideally you want to have that dip around 100-120Hz
Any Class-A SE stage runs the DC standing current AND the whole cycle of signal current through the positive supply. This means that the whole of the signal is handled (and potentially mangled) by any regulator/stabiliser or passive supply.
For this reason it really does matter what we use, since however tranquil the outut voltage may look on a scope, you need to know what is happening to the current, including the phase.
Folks here have been building SE power amps for years using power supplies with impedances of tens of ohms, or even higher; and many of these sound really good. Changing from these all-passive supplies to a cap-multiplier (of output impedance around 1Ω) improves the bass and makes low noise very easy. Most importantly, all the good things about the passive supply remain. This is because a properly implemented multiplier presents a signal-impedance very close to a plain resistance right across the audio band, with very little in the way of phase shifts, reactance, or unwanted EQ effects.
With closed-loop regulation, using an error amplifier, keeping to a plain resistive output is difficult (the more so at high voltage); and phase has to be manipulated to preserve the phase margin, or prevent it crashing through crossover at >20dB/decade. This is the reason that industrial chips like the LM317, LT108x (etc) present an inductive looking output phase that collides with the output capacitors to produce a lump in the impedance at some audio frequency, if we don't take care.
If you believe you need less output resistance than 1Ω ( a bit lower than this, if you can trade off some extra standing current), I think you should ask yourself why.
Compare an all-passive solution for sound before committing to any proposed low impedance solution.
Last edited:
That's correct, although the power factor of a Class-AB and Class-D amplifiers is even worse, since there is a much bigger difference between (resistive) DC load and AC load.
This was exactly my point before it went into a whole discussion 🙂
Couldn't say it any better 🙂
Either approach is fine, but it matters a whole lot to make this clear and obvious BEFORE you start working on solution and schematic!
Thank you Jan, this is is what I've got from the simulation of the circuit on the first post: