Please critique my PSU design

[ddroukas]

This is for a 300B SET roughly based on Thorsten's Legacy Universal DHT amp. The parts list for his tube-regulated supply are getting expensive on paper, so I'm wanting to make something more conventional using a CLCLC design. I've come up with this in PSUDII:

This uses a Hammond 278X (400-0-400 @ 230mA) and two Hammond 193H (5H @ 200mA). I intend to run a common B+ for the 300Bs followed by duplicated RC filters for each EL84 for channel separation. As above, I get 1.1mV ripple for the 300Bs and 0.69mV for the EL84s.

Questions:
1) Have I simulated the power transformer correctly? All my numbers come from Hammond's datasheet on the 278X, outlined below:

2) Are the current taps modeled correctly? I've assumed 90mA at ~420V for the 300Bs and 16mA for EACH EL84 (based on comments in another forum) but have only modeled one in the design. Should this be changed to 32mA to account for both simultaneously?

3) Any other errors or concerns?

Thank you for your help.

 

[famousmockingbird]

Hello, I would think the secondary DCR would be 47? The datasheet says 94 from red-red.

And yes i would change the last current tap to 32mA, I usually have better results with a resistive load in PSUD II.

 

[ddroukas]

Famousmockingbird: There was a similar post on audiokarma regarding what secondary DCR to use and no consensus was made. If I run the simulation with 47.42 (half of the red-to-red resistance) I predictably get greater voltages at my outputs [V(C3) becomes 450.88, V(C4) becomes 443.33]. My ripple is largely unchanged. I know people use PSUDII all the time, surely someone knows how to simulate the PT properly.

Also (and more importantly), I'm still not sure if I should be putting in the unloaded voltage (423.8V) or the advertised voltage (400V).

 

[MelB]

I would go for more C to lower the ESR because it is so cheap. Increase C3 and C4 would be my advise.

 

[ddroukas]

I would go for more C to lower the ESR because it is so cheap. Increase C3 and C4 would be my advise.

I plan on using Solen fast caps, whose datasheet lists the ESR as 0.0008R for the 47uF/630V. The lowest PSUDII will allow me to input is 0.002R. Increasing C3 and C4 to 69uF decreases my 300B tap ripple to 0.957mV (from 1.061mV) but increases my EL84 tap to 0.288mV (from 0.069mV before). Is there any intrinsic benefit I may be missing?

 

[Destroyer OS]

I would go for more C to lower the ESR because it is so cheap. Increase C3 and C4 would be my advise.

That's probably a good idea. Some fatass film caps. I'd be worried you might have a "choked" to death sound. How oversized are the chokes you think? I'd be concerned using a non-circular one with saturation during peaks. You can bet different core shapes that can swing current better.

I'd also consider a Qspeed diode on the return to center tap otherwise your extreme filter may not be so extreme.

 

[ddroukas]

I'd be worried you might have a "choked" to death sound.

I'm unfamiliar with this effect–I regularly see other CLCLC designs but haven't read of this concern. Do you think its better to do a simple CLC design and tolerate higher ripple voltages? What ripples values do you think are acceptable? I'll be using a Lundahl 1623 tapped at 4Ohms and 3k primary.

 

[Destroyer OS]

You've only got 4ma to spare based on what I see. The slightest change could cause saturation is what I'm thinking. And you may have 2x16ma on top of the 180ma? That would exceed the 193H current capabilities.

I don't see a problem with the design, I'm just saying it has no compensation for the mildest of fluctuations. It's a common design but are they typically playing on current limitations so closely?

 

[ddroukas]

You've only got 4ma to spare

Oh, sorry, I failed to catch what you meant. I indeed would have to increase the current ratings.

Anyone have any ideas about simulating with the off-load voltage versus the advertised voltage?

 

[soulmerchant]

The rectifier IFRM (current forward repetitive maximum) of 0.90A has been exceeded with a value of 1.23A......

I'm looking up the specs on this transformer online and it is looking like its 800V CT... have you ever used PSI II before?

 

[soulmerchant]

Once you start to get together a nice supply you can do this: put a stepped load on the first current tap.. make value 2 something like 200mA, after 3 or 4 seconds..

This will give you an idea of what your power supply will do after a sudden significant transient.

 

[famousmockingbird]

Famousmockingbird: There was a similar post on audiokarma regarding what secondary DCR to use and no consensus was made.

To me I would think if they wanted the resistance of the whole winding you would have to put in the whole windings voltage which would be >860 not 432. So maybe I have been doing it wrong this whole time but when Hammond says 94 ohms from red to red I would type in 47. And yes hopefully someone else chimes in on the matter, I tend to agree with Tom Bavis from the other thread you linked.

 

[soulmerchant]

red to red is 865.6V (nominal output!). DCR is 94.84 Ohms. This is straight off the specs sheet.

Um...do you really have 125V mains? I think 115 is more like it these days in the Americas.

And... you have this transformer already or are you just thinking about getting it? It might not be the most suited for your intended job...

 

[316a]

That will be a dead-sounding supply that will kill 5U4G pretty fast , ripple does not need to be vanishingly low . C1 is far too large and C3 too small . L1 too large in value . The step response is ragged and peak currents seen by the rectifier , very high . Better off going for monoblocks and perhaps consider dampers in place of the 5U4G , I would suggest a 200VA unit per channel with the 300B on their own filament supply . IMO and all that

316a

 

[soulmerchant]

Ok, so I did a nice example of choke input for you to see why it is desirable.

Please not "sand is better" comments from the sand PSU gurus who also lurk in these waters...

 

[ddroukas]

The rectifier IFRM (current forward repetitive maximum) of 0.90A has been exceeded with a value of 1.23A......

I don't get this notification when I run the simulation. The IFRM of a 5U4GB is 1.0A, did you put in the 5U4GA?

I'm looking up the specs on this transformer online and it is looking like its 800V CT...

The image in my original post has exactly what Hammond lists on their datasheet. 800VCT is advertised, however the unloaded voltage (which is what PSUDII asks for) is listed as 432.8V. This is the exact question I'm asking: which do most people use to accurately simulate their transformer?

I just checked my mains (NYC at 1905): 124.9-125.1

I've rerun the simulation using Edcor 5H and 2H chokes both rated at 300mA instead of the 200mA Hammonds. I might also change the Hammond PT to Edcor since it offers 800VCT @ 250mA compared to 230mA to be safe. I also didn't realize the maximum input capacitor value for the 5U4GB was 40uF so I reduced C1 to 33uF and increased C3 to 69uF (47+22uF Solen's in parallel). I get these results:

316a: What would you suggest as a better L value? Will running the input cap at 33uF resolve killing the rectifier (5U4GB)? Running dual monoblocks would be desirable but last-choice given space and cost.

 

[soulmerchant]

That will be a dead-sounding supply that will kill 5U4G pretty fast , ripple does not need to be vanishingly low . C1 is far too large and C3 too small . L1 too large in value . The step response is ragged and peak currents seen by the rectifier , very high . Better off going for monoblocks and perhaps consider dampers in place of the 5U4G , I would suggest a 200VA unit per channel with the 300B on their own filament supply . IMO and all that

316a

Fully Agree. My simulation of the supply suggests the 5u4g will not live long.

The biggest problem is C1 which needs to be smaller. I think this is one of those "300b on the cheap" projects.

A word to the wise - consider Tube lab's "power drive" circuit for 300b

 

[famousmockingbird]

Instead of lowering the reservoir cap just add some series resistance to each winding that feed the 5U4GB plates.


EDIT: Since the last section of the supply is really two supplies in parallel I believe you would have to halve the 470 ohm resistor value.

 

[soulmerchant]

I don't get this notification when I run the simulation. The IFRM of a 5U4GB is 1.0A, did you put in the 5U4GA?

yup. I believe your source resistance is incorrect. I get 64.43 ohms for that transformer. I am not a fan of Hammond iron, but at least their specs sheets are easy to read.

Either way you are REALLY close to the max. not good. TV Damper diodes are a far better choice if you can work with them.

The image in my original post has exactly what Hammond lists on their datasheet. 800VCT is advertised, however the unloaded voltage (which is what PSUDII asks for) is listed as 432.8V. This is the exact question I'm asking: which do most people use to accurately simulate their transformer?

when you edit the transformer, click to edit the RMS V first. Enter your nominal output, amps, regulation to get your off-load volts and source impedance.
Click OK.
Then click to edit the Ohms, and click in one of the columns to get the source impedance.
Click Ok.
You should see values just like in my screen shot. That is, unless I made a mistake...

I just checked my mains (NYC at 1905): 124.9-125.1

Very good to know this! In any case, I would look for a different transformer, and go for choke input. If I must have an input cap, I would try to keep it below 5uF.

But maybe I would not use Thorsten's circuit either...

 

[soulmerchant]

Instead of lowering the reservoir cap just add some series resistance to each winding that feed the 5U4GB plates.


EDIT: Since the last section of the supply is really two supplies in parallel I believe you would have to halve the 470 ohm resistor value.

Add serial resistance makes it even slower responding, not-dynamic sounding..