Feedback on first tube amp power supply design...

Hi!

I'm looking to build a Baby Ongaku (see attached PS schematic) but the power transformer has a higher voltage secondary than what is specified (360V vs 420V/480V). Since this is my first build directly from a schematic I thought this would be a good opportunity to play with the power supply design to accommodate this transformer - both because I have it and because it's a chance to learn something.

Since the voltage on the PT secondary is higher than specified, I thought this would lend itself to a choke input - as opposed to the sorta choke input in the schematic - to drop the voltage for a B+ of 350V. I have the current at 146mA as I intend to make a stereo version of the amplifier. It doesn't seem like much will chance downstream of the power supply.

I've modeled the PS in PSUD (see attached) to the best of my ability and am curious to get some feedback. Again, this is my first PS design and there's a good chance I've missed something. To date, I've built a couple of point-to-point kits and a couple of amps from a schematic with wiring diagrams and great documentation from MJ magazine (Japan).

The things I'm the most curious about are the values for the filters, the current draw, and how the ripple looks.

Thanks in advance!


Screen Shot 2023-07-13 at 1.03.20 PM.png

Screen Shot 2023-07-13 at 12.50.21 PM.png
 
Moderator
Joined 2011
In this case, you should actually measure that voltage, it's likely to be higher.
And also make sure the input cap can easily handle it.

A 10% rise without loading (from full load) is more likely than a 5% rise.
 
Last edited:
In this case, you should actually measure that voltage. And also make sure the input cap can easily handle it.
It's waiting for me in Japan! It's a Hashimoto transformer and ordered it wired for 120V. Here are the specs...

480V-420V-240V-0-240V-420V-480V-170mA
AC 0-2.5-5V-5.5A(×2)
6.3V-3A
5V-3A
480V: 60.5 Ohm Winding Resistance
420V: 53.1 Ohm Winding Resistance
1. Choke input, Full-wave rectifier: 250mA DC
2. Capacitor input, Full-wave rectifier: 170mA DC

I would like to use a pure choke input so I'd exclude the .68uF before the choke in the original schematic.
 
So I've modeled this in PSUD2. Personally I would reconsider the design.

Without modeling anything, I would reconsider the use of 4 5AR4 rectifier tubes. There's no reason to do that. You could use a single 6AX4, for example, saving yourself money at the same time.

Or you could be even more reasonable and use solid state rectifiers.

As for the design itself, as shown, the time constants for the C sections are poor and, in my experience, will not produce sufficient transient response. Bass response could be poor as well. I always aim for 8-16 mS response for the first capacitor, with C2, C3, CN separated by a few mS each. These curves show very slow charge and discharge. It actually looks awful IMO!

1689872060852.png


There's a few ways you could handle this. You could create an LCRC PSU that then leads to independently coupled RC stages, one for each driver and output tube (so 4 altogether). That would result in "the best" performance and you can tailor ripple requirements for each stage as you see fit. It would also be more complicated and require more time and effort to construct and test.

I'll share an LCRCRC design that may work pretty well. Personally I prefer AT LEAST one RC section for the driver/input stage and one RC section for the output stage. Obviously this is not that. But what this sketch does show is the improved time constants for C1, C2, and C3:

1689872080934.png


Ripple is about ~20 mV on this sketch. That's probably sufficient for the output stage (and will be reduced further by the output transformers) but may not be sufficient for the driver stage.

I like DHTRob's guide to using PSUD. In my experiences so far, following his guidelines has improved the sound of my SET amps.

https://www.dhtrob.com/overige/pdf/dhtrob_psu.pdf

Also an excellent resource:
https://a-direct-heating-triode.blogspot.com/2018/06/lcr-phono-preamp-part-3.html
 
You're going to use a quad of 5AR4?!?!?! Is that what 5AR4 X4 means in your screenshot?
Is the PSU also going to double as a hot plate?
I just realized that! The labeling in PSUD can be a little quirky and just skimmed the parameters. PIV and IFSM are fine but yes... Changed to 5AR4 and things were considerably different in my tinkering. Thank you for pointing that out.

I always aim for 8-16 mS response for the first capacitor, with C2, C3, CN separated by a few mS each. These curves show very slow charge and discharge. It actually looks awful IMO!
This is why I'm here and I truly appreciate your feedback... How are you visualizing the mS response at each cap? Again new to the program and designing power supplies from scratch. Regarding the curves, I've remodeled them (see below) one with a 5AR4 and one with a 5U4GB for more rectifier voltage drop to hit my target B+ of 350V (the 5AR4 has an additional RC for more filtering and for more voltage drop). Curious to get some feedback on the designs as they stand.

Thanks again for everyone's time and consideration!

1. 5AR4
Screen Shot 2023-07-21 at 9.45.30 AM.png
Screen Shot 2023-07-21 at 9.45.48 AM.png
Screen Shot 2023-07-21 at 9.46.12 AM.png



2. 5U4GB
Screen Shot 2023-07-21 at 9.41.31 AM.png

Screen Shot 2023-07-21 at 9.41.48 AM.png
Screen Shot 2023-07-21 at 9.42.08 AM.png
 
Use the time scale on the bottom and see when you've reached to ~80-90% voltage. I am more or less parroting what the maintainer of the Directly Heated Triode blog suggests.

I built a 2A3 amp with a 5U4GB initially. Kept adding capacitance and noticed it was sounding quieter (less ripple) but also sounded less defined. That's when I delved into PSUD. I've made several changes to that PSU, the latest being to switch to SS rectification. To me it is hands-down better with transients as well as bass response.

In my experience, vacuum tube rectifiers can be great performers, but it is more challenging to have great performance AND the lowest ripple. It's easier to get both with SS.

Everyone has different preferences and, maybe even more importantly, different systems. If this is your first PSU, consider trying one setup, then tweaking it so you can determine what you like best. It could be as simple as changing capacitor and resistor values. It could be adding on RC cells. Etc.
 
Use the time scale on the bottom and see when you've reached to ~80-90% voltage. I am more or less parroting what the maintainer of the Directly Heated Triode blog suggests.
This is great. So I'm measuring from 0mS to the point at which the voltage has risen to 80-90% of the idle voltage at the first filter cap in my supply? So in the 5AR4 plot below, my frequency response for the first cap would be approximately 15mS (the drawn pink line). 360V is 85% of the idle voltage at the first cap. I would repeat this for each cap and in a perfect world they would only vary by 2 or 3mS. Sound right?

Screen Shot 2023-07-21 at 12.05.24 PM.png


If this is your first PSU, consider trying one setup, then tweaking it so you can determine what you like best. It could be as simple as changing capacitor and resistor values. It could be adding on RC cells. Etc.
I think this is the play!
 
That's more precise than I've ever been! Others may jump in and correct me. The theory behind the writer of the DHT blog's suggestions has to do with the transient timings for specific music instruments. I think he lays it out in that post I linked but also might be somewhere else.

I think you're on the right track. I forgot to mention that you can also swap in one tube rectifier for another (depending on pinout and heater requirements of course) and get different performance. But I am personally against that as a final or long-term approach since you should design the PS specifically for one rectifier to get the best performance out of it. But perfectly valid for experimentation.

Try lowering capacitance or switching to SS rectification, for example, and you will see that the time to reach 80-90% charge can be reduced dramatically.

EDIT:
I should probably not be saying 80-90%. Am not an expert on the subject matter nor technically trained!
https://unacademy.com/content/jee/s...s the time taken by,connected to it in series.
 
Ripple is about ~20 mV on this sketch. That's probably sufficient for the output stage (and will be reduced further by the output transformers) but may not be sufficient for the driver stage.
Quickly and probably obvious... when visualizing ripple in PSUD, is it easiest to just zoom in on the voltage at a specific cap and read the variation from peak to trough? So in this instance at the first cap (C1) it would be about 1.5V (403.2V - 401.7V).

Screen Shot 2023-07-22 at 11.55.54 AM.png
 
Yes.

The two basic simulations I always run are the first 500ms and then 50ms after 10 or 15 seconds. The first gives you PSU ramp-up behavior and the latter gives you ripple.

You can also set up simulations to help detect ringing or power supply instability. I think you can Google and find tips on that, or look for PSUD threads on this forum.

By the way, are the capacitor, choke, and transformer resistance specs correct on your setup? As in, are they the real values? You must always get as close as possible to the real values.
 
By the way, are the capacitor, choke, and transformer resistance specs correct on your setup? As in, are they the real values? You must always get as close as possible to the real values.
The PT and choke specs are from Hashimoto. The tube data is all from PSUD and the cap calculations should be about right but struggled with the resistance calculations.

Was a little stuck and feel that they’re not far off… found some formula online and did my best with the math and data sheets.
 
  • Like
Reactions: 1 user