|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
173.185.144.238
I am getting what looks like a weird ripple for the power supply the way it is loaded with the 813's. Wondering if I am doing something wrong. Data is
full wave tube 5ar4 rectifier
power transformer has a half no load output of 432.8 vac per Hammond (278x)
first cap is 2.2uf
first choke is 10 H/65 ohms (Hammond 193M)
second cap is 94 uf
second choke is 5H/55 ohms
third cap is 94 uf
load is 160 mA based on measuring an average of 40 mA plate current across the cathode resistor (355vdc/750 ohm cath resistor)
When I run this I get a low value for the ripple of about 2 or 3 mV, but a very high frequency that make the band quite thick looking. Never have seen this before. Am I doing something incorrectly?
Follow Ups:
My personal view here, which has given me good results in several units.
I was pointed in this direction by Jeff and it's based on John Swenson's work. I suggest you look up his posts about PSUD.
Note: I have not gone as far as John and Jeff in terms of LSES but I may try that sometime. What I have used is their methods of step load analysis, and for me it really works well.
The method is to edit the current load and specify a step load, with the step about 10% or so more. I make it come in after 15.1 seconds. Then observe the change of output voltage after the step (specify 'after a reporting delay' of 15 seconds).
If there is oscillation, it's a *poor* PSU which is likely to sound harsh.
If the drop is steady without oscillation but slow, the sound may be somewhat slow and verging towards boring, probably with poor leading edges.
The ideal (I am told, and believe) is a fast clean drop without oscillation, with a very slight overshoot. Ideally it stops dropping after about 0.06 seconds.
This cannot be achieved with large capacitors, so to get reasonably low ripple requires a 2 stage PSU.
My best SET amp uses a cLCLC supply tuned to give this response and it sounds *very* good, with fast clean leading edges and rhythm and dynamics enough to satisfy a diehard Naim fan friend.
Now to your PSU.
It is accepted practice to increase the C and/or L values as you go along a multi stage PSU. Your use of 10H and 5H in that order contravenes this and is most likely the reason for the oscillation on startup.
I modelled it with a step load 160mA to 180mA after 15.1 sec and this clearly shows 5 oscillations after the step.
Swapping the chokes over (5H then 10H) gives an improvement, 3 oscillations, but taking 0.3 seconds to get close to final value.
Changing C2 to 30uF ... now it gets interesting!
Smooth drop, no oscillation, no overshoot, getting there in just under 0.2 seconds; not too bad a PSU. About 19mV of ripple (analyse for say 900mS after delay of 14 seconds to get the ripple).
If you can tolerate more ripple, you can go further.
Change C3 to 50uF and now there is slight overshoot, no oscillation, and a response of 0.1 seconds; reasonable.
Change C3 to 40uF and it's ugly! Definite oscillation. Now you have to play with various values to try to get what you want.
With just a bit of playing I found good results with:
4uF; 5H 55 ohm; 20uF; 10H 65 ohm; 40uF
Smooth drop, no oscillation, just detectable overshoot, and response
under 0.1 sec (looks like about 0.75 sec). A very good PSU IMHO though I expect a few comments on this ...
I suggest you try these models and see if you can do better; also check the ripple and see if it's low enough for your needs.
I hope this helps; I repeat, it has really worked for me.
Regards
Mike
(in the freezing UK)
and do it the
WOW !!! Imagine !! A graduate EE adding a 200 Ohm resistor in series as the filter to the Finals, and then, using an additional 108 Ohms of DC resistance to the two chokes !! 308 Ohms of UGLY series resistances.And look, a 'C" input filter !! Oh well.
I liked the graphs and such, but the UNconcious EE author's suggested design will work and sound like doggie - doo.
Jeff Medwin ...... a Business Major...... Penn State.... 1966
Edits: 01/12/17
You go right ahead on it sport, that's not for me. psdu2 has always worked fine for me until I switched to windows 10. It is not yet set up for that apparently. Good thing I did not sell my windows 98 laptop.
Edits: 01/04/17
PSUD has significant limitations when it comes to designing a supply for audiophile use. It's nice for quickly predicting ripple content and loaded/unloaded output voltage, but not much else. Many "old-school" designs perform well in those two areas, but will color the sound. If you want to design power supplies that really perform, you'll need to learn SPICE.
--------------------------
Buy Chinese. Bury freedom.
I get a headache just looking at that link! I can live with psud2, at least for now.
Edits: 01/04/17
I wasn't referring to the link, but to SPICE, which is a means of modelling circuitry using a computer in order to make theoretical measurements and performance predictions. SPICE was very expensive software not generally available to hobbyists until Linear Technology began to offer a version for free, named LTspice.
PSUD is almost certainly using some version of SPICE internally, but with a simpler and more limited user interface. Just as one example of its limitations, PSUD can't tell you the output impedance of a power supply over frequency. I consider this a critical characteristic, and I won't build a supply now without analyzing it in SPICE.
--------------------------
Buy Chinese. Bury freedom.
I'll try to take a look at it, but I have a feeling most of it is going to be more complicated than I like to get.
Start with something simple. Stick a voltage source on the page and ground the bottom. Then wire two series resistors across it. Tell the software to RUN. When it's done, look at the voltage and current around the circuit. It's a small step from there to viewing RC time constants, LC filters, power supplies, tube amplifiers etc. Not for everyone, I know, but it's an extremely useful tool.
--------------------------
Buy Chinese. Bury freedom.
I might be able to handle that.... : 0 ). Thanks for the info.
Almost everything you need in one place! That plus a copy of LTspice will get it done.
--------------------------
Buy Chinese. Bury freedom.
Without any issues?
Yes, I recently downloaded the latest version and it works with no issue on Windows 10.
Maybe for you, but not for everyone. I had issues using it on my brand new laptop with windows 10. Gave different results than other folks were getting with older OS. When I switched to my old laptop with NT, it was ok again. If you look at the website, windows 10 is not among the OS's listed as working with it.
z
You may not have hit the right combo yet. This is the first time this has happened to me.
it seems like sometimes a well-worn model (where I have made multiple changes) has a "memory", and changes don't do what I expect.
If I close such a model down and re-enter the same data a fresh new model, I get different, expected results.
I just figure that it is par for the course with hobby freeware.
Just went to the website and windows 10 is not one of the operating systems listed. Ah #$%&*, now what do I do? May have to dust of my old laptop and load it on there.
Edits: 01/03/17
Maybe not. See my reply to TK below. This is a question I really need an answer to.
Is this it?
--------------------------
Buy Chinese. Bury freedom.
Edits: 01/03/17 01/03/17
Opps. I forgot to say the resistance of the power trans is 95 ohms. Everything else is the same though. I got a much different chart, with a rounded corner at 411.8 vac. I tried changing my values to yours and I still don't get the same peak at the start that you do. How is that possible?
Are you using one of those? This is an old program Vinnie, and I guess it's not worth the author's time to recompile it for newer systems. I use PSUD2 (2.0.2) with Win XP and consider myself lucky that it runs.
--------------------------
Buy Chinese. Bury freedom.
TK, Thank You for that bit of info. I still have my older computer with Windows XP, so this should work on that with my other audio softwares.
give
Loaded it up on an old thinkpad I got with windows nt to run SE CAD and it seems to be doing ok there. That's two programs I have to run on ancient software now or not at all. Oh well, at least I can use it. It is a very handy program for quickly doing what ifs, and it also has been fairly accurate for me in the past. I would hate to lose it.
Edits: 01/04/17
I have an old lap top with windows 98, guess I will have to try that. Blah!
Don't use the CC use just a resistor for the load try that.
Lawrence
How do I determine the value of a resistor that is equivalent to the four 813 tube currents?
.
nt
Looks good to me. Tried to save image as a picture, but did not save.
did you get a really heavy saw tooth ripple?
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: