New project and my first circuit...

[mr mojo]

Howdy folks!

Starting a new project, this time from the ground up by designing the circuit as well. What I've got here is a power supply for a stereo 6CK4 SET amp, signal side coming soon. I was hoping some members would be kind enough to share any comments, criticisms etc to make sure I'm on the right track.

I would certainly appreciate it!

Best,
mr mojo

[Eli Duttman]

I assume 120 H. is a "typo" and you plan on using 2X 20 H. parts. Since you're going to spend on "iron", consider a choke I/P filter.

The 5U4 is directly heated and does not have a cathode sleeve. Taking the B+ off pin 8 allows for both directly and indirectly heated rectifiers. If the use of the 5U4 is "cast in concrete", taking the B+ from a CT on the 5 V. winding has a reputation for sounding "best".

BTW, steer clear of Russian and Chinese 5U4Gs. They are GUANO. A NOS 5U4GB by TungSol, RCA, or GE will serve you much better.

[kevinkr]

I assume this is going to be a single pair of 6CK4 in SE, in which case you really don't need more than 100mA for a pair of them, at say 310Vdc you would be running them at around 30mA each.. The extra supply capacity really doesn't buy you much in a class A se amplifier, provided the source impedance is comparable.

Also I would use a little more capacitance after the choke, and imho a single 5 - 10H choke with a couple of 47uF/630V solen film caps will provide all the filtration you need. This will provide a low source impedance relative to the combined output tube RP and opt series impedance which will result in lower cross-talk and lower output impedance in the lowest octaves, unless you plan to use the ultrapath topology in which case it might be less of an issue.

The 5U4 has a pretty fast warm up characteristic as well, a 5AR4 would be much kinder to the cathodes of the 6CK4's as they warm up.

[mr mojo]

Eli,

Thanks for the feedback-I really appreciate it!

You're right, couple of "typos" on that last pic. Here's an updated image:



Kevin,

My thanks to you as well. Instead of the 20 H chokes-which are a bit harder to source, would a pair of 10 or 15 H chokes work as well? Will they need to be rated for @ 150-200ma?

On Angela's site:

Hammond 10H, 125ma, 155 ohms, 500VDC

Hammond 10H, 200ma, 82 ohms, 800VDC

Hammond 10H, 150ma, 102 ohms, 800VDC

How do I know how many ohms would be best? I like the 800VDC ratings since I'll be filtering 400v.

Also, this combo is supposed to yield about .01% ripple. Is that overkill for a tube audio amp?

Best,
mr mojo

[SHiFTY]

IMHO you should use more capacitance, and just a single Pi filter. The KISS principle and all. You could use a second choke to supply the driver circuits but it's not really necessary.

You should use the max input capacitor rated with the 5U4G, which is about 33-40uF depending on voltage, at 350V you should be fine with 40uF. Use a good NOS 5U4G and you will not have any problems.

Then, a single high quality, lowest possible DC resistance choke of about 10H followed by 100uF of film cap (the solens are pretty good apparently) will give you much better filtering and lots more energy storage and better bass.

Try downloading the duncanamps.com power supply designer software, you can model the ripple of various power supplies at various loads.

[SY]

Shifty, I strongly second the idea of modeling this supply in PSUD. I've been amazed at the accuracy of its predictions.

But I have to quibble with the idea of simplification you suggested- very often, you get better ripple performance by using several smaller sections rather than a single bigger section. It does add to parts count, but...

[SHiFTY]

By having a second section you massively add to power supply impedance though. With small caps voltage would sag a lot with a large transient. (yes I know it's class A but it still dips a bit)

If you were going to use two chokes I would suggest using a pi-filter for each channel - i.e. a single input cap from the rectifier then a choke and cap for the left, and a separate choke and cap for the right. This would be the best IMHO...

[SY]

To put a fine point on it, I ran a couple of simulations. The first is for a two section as drawn, 8u caps all the way, the first choke being 12H, the second being 20H.

With a 200mA load and some reasonable guesses about choke and transformer DCRs, the sim shows 420V out with 24mV of ripple. If the load is stepped to 250mA, the output voltage drops to 399V, with a slight increase in ripple to 29mV.

Now, if we reconfigure to a simple CLC, but retaining the same total inductance (22H), add the DCRs and knock off 25%, and combine the last two caps so that the second cap in the CLC is 16u, we get a similar output voltage, 424V, but with a ripple of 570mV, nearly 20 times worse than the CLCLC. Have we picked up any extra regulation? Well, with the load stepped to 250mA, the voltage drops to 403.

So in both cases, the extra 50mA step caused the same change in output voltage- supply impedance hasn't seemed to improve. But splitting up the Ls and Cs had a remarkable effect on the ripple voltages.

I note that both sims showed an overcurrent warning for the 5U4G at the 250mA level. This suggests that at 200mA, you're running the tube close to its limits...