This is what I basically use in almost any tube/valve power and pre-amplifier.
It's very simple and effective, works even great for those who like the behavior of a tube rectifier.
In a push-pull I mostly just use it for the pre-stages + screen voltage. (since the ripple mostly cancels out in the PP stage)
@ljgmdad L1 and L2 are basically not even needed. (same for L3).
It will only be a waste of space and money in this case.
Btw, obviously we can't use 1N4148's in that position, and sometimes LTSPICE also give issues when simulating those under the wrong conditions.
The drops also depend on impedance of your household wiring and characteristic of appliances.
The problem with increasing C1 or even C2 will be the surge current through L1 at startup. I'm not sure the MOSFET is the final answer, though. I need to look into the OB2 or other options.
The tube rectifier I had intended to use has a max plate current of approx 850mA. I wanted to stay within that spec even at power-on.
Now, I'm considering the simplicity of solid state rectification as potentially desirable... Heck, the DHT filament supplies already have solid state. Why not the B+. I don't know yet.
Now, I'm considering the simplicity of solid state rectification as potentially desirable... Heck, the DHT filament supplies already have solid state. Why not the B+. I don't know yet.
PSUD2 is excellent for simulation purposes but what really matters is implementation, only then You can measure or hear how much ripple or other type of hum You really have.
I have built 4 tube phono stages, with & without regulators and the PSUD always gave excellent results for ripple but implementing it without hum or buzz is a different matter.
It does take a lot of practice ( & learning ) to build tube amp of any kind without hum ( on high efficiency horns of 96dB everythnig not right in psu, layout or wiring can be heard, for example ).
Personally I admire the guys who can build phono, line + se amp without noises on high eff. speakers, maybe only with natural tube hiss on levels we never listen at.
Best wishes at building, Krca
I have built 4 tube phono stages, with & without regulators and the PSUD always gave excellent results for ripple but implementing it without hum or buzz is a different matter.
It does take a lot of practice ( & learning ) to build tube amp of any kind without hum ( on high efficiency horns of 96dB everythnig not right in psu, layout or wiring can be heard, for example ).
Personally I admire the guys who can build phono, line + se amp without noises on high eff. speakers, maybe only with natural tube hiss on levels we never listen at.
Best wishes at building, Krca
Great input! It seems that I probably could eliminate the 8H choke and C2. But things don't look as good without any of the chokes. See below.
First off, the baseline shows no sign of 120Hz ripple. (I changed the bridge diodes to something usable).
Now, if I eliminate L2, ripple appears but is under the 1 mV threshold...
Now, eliminating both L2 & L3 is not so pretty...
Now, bringing L3 back but eliminating C2 looks reasonable again:
So, one Hammond 8H choke and one 20uF cap could be eliminated. We're talking about $50 in savings to go from zero observable ripple in simulation to, what, about 600-800 uV of ripple. I'm pretty tempted to just spend the $50 and the extra real estate.
There are no floating voltages in SPICE - your AC source has one terminal connected to ground, whether you want or not (in this ccircuit you don't!). One way to handle this is to remove the other ground connections, make differential measurements (V2-V1). (you could insert a transformer, pretty complex model, though). Or you could add a second voltage source at 180 degrees, two dodes instad of the bridge - output will be higher by one diode drop. This may be simplest.
You're making this way too hard on yourself.
Three chokes? Especially when dealing with such low currents, that is way over the top. Good luck managing all of the inductance/ringing/complicated interactions between all the power supply components.
Transformer specs I entered: 175 V AC @ 150mA. Choke I used is a toroidal from Antek that I've used before.
I bet this would sound great. PSUD estimates .38 mV.
Three chokes? Especially when dealing with such low currents, that is way over the top. Good luck managing all of the inductance/ringing/complicated interactions between all the power supply components.
Transformer specs I entered: 175 V AC @ 150mA. Choke I used is a toroidal from Antek that I've used before.
I bet this would sound great. PSUD estimates .38 mV.
Maybe have time to screw around with it tomorrow. Important to the execution is knowing true transformer and choke DCR ratings, and I plugged in some arbitrary values just to sketch out a CLCRC approach. You could probably get away with less capacitance and DCR and tighten things up a bit. You could also split from C2 and decouple with additional RC for each channel.
The software is free and fairly easy to pick up, so if you've got a Windows machine, try it out! I've got two young boys and not much spare time at the moment.
Honestly, a simple CLC followed by Tom's Maida regulator is probably as good as you can get: insanely low ripple, insanely low output impedance. My friend did that for a build of Salas's 6V6 preamp and it sounds amazing. You could also consider Salas's HV Shunt Regulator. Haven't tried one myself but people say it is top notch.
But those are both approaches that utilize PCBs and require substantial heatsinking -- maybe that's not the vibe you are looking for? I think a vacuum tube-rectified choke-input supply is as worthy an approach as those others. They can all sound good and I'm sure they also all sound different.
Try not to agonize over "the best way" to do it. Pick the one you want, do it well, and consider changing it later as a fun experiment.
The software is free and fairly easy to pick up, so if you've got a Windows machine, try it out! I've got two young boys and not much spare time at the moment.
Honestly, a simple CLC followed by Tom's Maida regulator is probably as good as you can get: insanely low ripple, insanely low output impedance. My friend did that for a build of Salas's 6V6 preamp and it sounds amazing. You could also consider Salas's HV Shunt Regulator. Haven't tried one myself but people say it is top notch.
But those are both approaches that utilize PCBs and require substantial heatsinking -- maybe that's not the vibe you are looking for? I think a vacuum tube-rectified choke-input supply is as worthy an approach as those others. They can all sound good and I'm sure they also all sound different.
Try not to agonize over "the best way" to do it. Pick the one you want, do it well, and consider changing it later as a fun experiment.
@b_force I started the thread asking for advice on an acceptable level of B+ ripple for a single ended headphone amplifier. Presumably lower is better when highly sensitive transducers are strapped to your ears. At the very least, I expect to take measures that may not be normally taken in speaker amps. So, I'm not so much seeking an insanely low ripple as trying to determine what the design point ought to be before buying components and starting to breadboard.
Last edited:
Caveat: I've never designed or implemented a choke-input supply. You will need to carefully select transformer and choke, determine critical inductance, assess where and what value resistor to strap across it, consult 6AX4 datasheet for the series resistance it likes, etc etc. That said, the 6AX4 model below looks pretty good to me! That 40 mV will of course be lowered by the output transformer too. You could always use an additional RC filter for each channel.
Choke input sketch
6AX4
Transformer specified 275V @ .15A
Ripple at 40mV
SS Bridge
Ripple at 15 mV
Choke input sketch
6AX4
Transformer specified 275V @ .15A
Ripple at 40mV
SS Bridge
Ripple at 15 mV
I will add that you could damp the PS further with some added resistance. But the values are pretty arbitrary in the absence of chosen transformer and choke. This is illustrating an approach that's trying to balance damping with the fastest possible charge/discharge of C2 / the capacitor(s) feeding each stage. This preference is based on recently converting 5U4GB rectified PS (for 2A3 output) to use SS diodes.
There's no way to say how low is low enough without knowing your HP's, without knowing your hearing acuity, the OPT loss, and the turns ratio... What output signal level is too much for your equipmment and ears? Then you have to calculate that backwards through the OPT to the primary.
Thus "an insanely low" ripple might just be the sensible thing to shoot for here. I'm liking the idea of the source follower more and more.
@Thekak thanks for taking the time to try out a choke input version. I think we want more than one LC stage for sharper cutoff and higher attenuation of the ripple frequency.
Going back the simple CLCRC filter, I reproduced your sim with some of my available components to achieve similar Vripple output. The design does require some large capacitors which does preclude the use of oil caps. And one interesting observation in spice is that I'm 10 ohms of series resistance away from the output response looking like this:
Versus looking like this:
Going back the simple CLCRC filter, I reproduced your sim with some of my available components to achieve similar Vripple output. The design does require some large capacitors which does preclude the use of oil caps. And one interesting observation in spice is that I'm 10 ohms of series resistance away from the output response looking like this:
Versus looking like this:
Why resistors before and after the diodes? Why before at all? Why that much series resistance before the first unnecessarily-small value capacitor? You are using SS and can do better than 20 uF. 20 -> 300uF seems unnecessarily dramatic too. Will generate a lot of large charging current pulses. What is DCR of choke & transformer?
I don't have LT Spice. Would like to learn when I've got some time to dedicate to it. That response you are showing is pretty surprising.
I am totally fine with all electrolytic capacitors in a PSU. Been there and done all that capacitor obsessing. You wouldn't want to be committing to film caps until you've determined final values anyway using cheaper ones (IMO).
I don't have LT Spice. Would like to learn when I've got some time to dedicate to it. That response you are showing is pretty surprising.
I am totally fine with all electrolytic capacitors in a PSU. Been there and done all that capacitor obsessing. You wouldn't want to be committing to film caps until you've determined final values anyway using cheaper ones (IMO).
Last edited:
Oh I was adding the 120R as series resistance to emulate the 6AX4, which I don't have a model for at the moment. I will not use SS diodes in the final design. As I mentioned, that 100R series resistance smoothed things out as shown. 20u is about right for the first cap in a cap input design with these tubes. Maybe I could get away with about 40u. The DCR of that Hammond 8H choke is 259R.