Power supply ringing

[Tom Bavis]

Pentode stages look more like a constant current load than a resistor really, so using the "current" load is not unrealistic. Triode output stages have a plate resistance LESS than V/I, will provide additional damping compared to the "resistive" load model..

Using two equal value chokes has always given ringing when I have done simulations. Larger values of C lower the ring frequency, but maybe more importantly, lowers the circuit Q. Making output C larger means less of the ringing gets to the load.

 

[BinaryMike]

kward, some caution is probably needed when using a stepped current tap and looking at the shape of any resonant response, especially when there is no resistive loading on that node with the current tap.

I think it's fair to parallel the stepped CCS with a resistive load to get a bit more insight, but we're bumping up against PSUD's limitations here. If the load is a triode, as Tom said, then actual load impedance is lower than a simple load E/I calculation suggests. Some folks would switch to SPICE simulation at this point, but it's difficult to argue against the breadboard approach if you have access to all of the necessary components.

 

[primosounds]

For Primo's PSU, changing to the values requested, gives this response.
This produces 40mV ripple on C2 and 1.1mV on C3 when run at steady state with 120 mA current draw.

Don't know if it will sound better though.

Hi KW, in the sim you ran for me here why did you insert the 120ma/240ma current load between L1 and L2.

 

[Retrovert]

I actually just acquired a CheapoModo that needs assembling, so once I get around to putting it together I can find the optimal snubber for the power transformer (lifted from a Heathkit AA-100).

Nice cornerstone. That AA-100 was reported to have good iron, both power and output, so that should stand you in good stead for power. Too bad the rest of unit are reported to be so unsuitable, both in construction and design. That power transformer, and the output transformers as well, were probably made by Midwest Coil & Transformer (the EIA code will be definitive) and while probably custom made, you might be able to obtain the specifications for a similar unit. Could be it was just a rebranded stock unit.

I have the power supply half of the Heathkit W3-M Williamson amplifier (amplifier and power supply are separate units) with a monster 10 H choke. (No, I did not slip a decimal on that. "Naah. That's not a choke. This is a choke.") It's going to power a trio of WA-P2 preamplifiers, with plenty of juice to spare. Should be rock stable in that application. Once I get my Cheapomodo built I'll see where the ringing shows up on that.

What I found interesting about modeling the W3-M design in PSUD2, is that merely doubling the filter capacitor (from 400 µF to 800 µF) critically damped the system with no overshoot. Adding a second inductor, of any size, did nothing in and of itself, it was all in the additional capacitance. PSUD2 does not, as far as I can tell, model swinging chokes only smoothing ones, so I don't know what effect that would have.

Please do report back on your snubber results. It should be interesting to see what happens with geriatric iron.

As kward noted, the differences between theory and practice can be substantial.

 

[kward]

Hi KW, in the sim you ran for me here why did you insert the 120ma/240ma current load between L1 and L2.

It's supposed to simulate the output stage current draw at idle and at max output power. But it's probably too much current for your actual amp, then again I wasn't sure so I took a guess. In any case the exact guess I made is not as important as how the PS reacts to the change. Note using the 240 mA value causes the simulator to complain about exceeding max current draw on the 5AR4 (but again it was a guess).

 

[thorpej]

It's supposed to simulate the output stage current draw at idle and at max output power. But it's probably too much current for your actual amp, then again I wasn't sure so I took a guess. In any case the exact guess I made is not as important as how the PS reacts to the change. Note using the 240 mA value causes the simulator to complain about exceeding max current draw on the 5AR4 (but again it was a guess).

PSUD reports a similar problem with my 40W amp simulation -- In my case, it's complaining about exceeding the max current at start-up (760mA vs the max of 750mA), but a real 5AR4 wouldn't have this problem because it soft-starts due to the indirectly-heated cathode.

 

[Retrovert]

PSUD reports a similar problem with my 40W amp simulation -- In my case, it's complaining about exceeding the max current at start-up (760mA vs the max of 750mA), but a real 5AR4 wouldn't have this problem because it soft-starts due to the indirectly-heated cathode.

It's telling you about a real problem that kills tubes.

The power supply pulling current through the rectifier (even if the tubes are not, the capacitors certainly are) before the rectifier's cathode is fully hot and the islands of charge have coalesced into a unitary emission body. As soon as emissions start at any level the tube passes a current, even though it shouldn't be doing this as this current originates with isolated point-source emission on the cathode. This causes thermal runaway at those spots. (The hotter portions emit better, so they emit more which makes them hotter, which goes round in a circle.)

RCA, particularly Schade, wrote about rectifier lifespan and it appears to comes down to early conditions which seriously stress the cathode. That work was first done 75 years ago and nothing has changed since then, it was only built upon and expanded.

In general, the cathode's characteristics depend upon the temperature, current flow, and the degree of activation. That activation is both a function of manufacturing technology and the life characteristics of the cathode as it ages and deteriorates. The cathode is robust, true, and particularly so in rectifiers because of the abuse they endure, but this routine overload condition is why people blow up rectifiers with alarming regularity. Particularly the expensive ones.

Adding a delay, via a MOSFET to keep the cathode from passing any current whatsoever into the rest of the power supply provides greater lifespan. (I don't want to hear any nonsense about sand in the power supply affecting the "sound". Power supplies only have a "sound" if they are poorly designed or have inadequate output for the load.) There's no reason to avoid silicon rectifiers, provided a delay is used to ensure B+ cannot be applied prior to all tube cathodes fully reaching the specified operating temperature.

 

[primosounds]

For Primo's PSU, changing to the values requested, gives this response.

View attachment 862134

This looks quite a bit better, but I think it can be better still with some more tweaking.
If you run the above at steady state at 120 mA current draw, the ripple at C2 is 158mV (peak to peak), and ripple on C3 is 0.5mV.

You might want to try some more tweaks to see if you can keep the ripple about the same, but get rid of that hump half way down the step. So just as an experiment, here's one tweak I tried:
View attachment 862136

This produces 40mV ripple on C2 and 1.1mV on C3 when run at steady state with 120 mA current draw.

Don't know if it will sound better though.

Hey, Kev, i tried to run the exact same values on psud to get the graph you showed. I inputted the same values but i could not get the graph to show the same axes as you. My graph starts a 0 at x,y. What did you do to get the graph to the points shown? Also i noticed your graph starts at 1.8 sec and response begins at 2.sec. Why is that? Also why did you put the stepped load in between the 2 inductors. It seems that the load would be at the end of the filter components. cheers, primo

 

[thorpej]

I'm keenly interested in the ripple current topic, as well. I've read a tech note two on it from cap manufacturers, but it always seemed a little opaque to me.

So, I went back to my 60W UL amp PSUD simulation and adjusted the simulation window to look at the current at the first (reservoir) cap (500uF 500V JJ TC-series). I set the simulation to run for 2sec @ 2sec so I could see the steady state at idle. I(C1) = 483.47mA RMS -- this is the AC ripple current through that cap? (The rest are all down in the 100s-of-uA range). At full power (simulation set for 2sec @ 7sec, with the load stepping at 5sec), I(C1) jumps to 784.36mA RMS.

The data sheet for this cap has a "Iac" spec of 1000mA, which I assume is the max ripple current rating for the cap.

Am I looking at the right stuff?

 

[kward]

Primo, It's possible I made a bad assumption (not having seen your actual schematic) on where you've got the tap for the B+ line feeding the output transformers. I assumed it was after the first filter choke. Correct me if it's somewhere else.

Regarding the scale of the graph, you can "zoom" in and out on PSUD. Take your mouse and position it over the graph a little to the upper left of the position where the step starts. Hold down left mouse button (on PC anyway), and drag to lower right a bit. Let go after dragging to appropriate spot. This will zoom in. You can reset the zoom by clicking the magnifing glass icon top right of the graph pane with the red X superimposed. Sometimes you need to zoom out first before selecting the area to zoom in on, if the area you want to zoom in on is right at the top of the graph pane, and you do that by clicking on the magnifying class icon with the up arrow.

I'm talking about these icons:


Going left to right, first magnifying glass is zoom in, second is zoom out, third is cancel whatever zoom you had previously applied. So, you can zoom in in two different ways, by using the magnifiying glass icon, or by selecting and dragging with the mouse.

 

[primosounds]

"Primo, It's possible I made a bad assumption (not having seen your actual schematic) on where you've got the tap for the B+ line feeding the output transformers. I assumed it was after the first filter choke. Correct me if it's somewhere else."
Hi Kev, the OPT B+ is actually taken off the last LC and the driver tube's B+ is taken off another RC stage which follows and consists of 3.3K x 40uf. Sorry, for not giving you more accurate information. I really appreciate your help and interest in my power supply. I have been mostly copying other designers ps designs and just upgrading the current requirements as needed. But, working with psud does give me a base from which to judge what i hear with the design of the power supply.
Regarding this ps, i am very happy with the dynamic quality but it is still a hair or hair and half's width too much energy in f,s, and t sounds. It is probably due to the Altec speakers which are horn loaded and tends to beam the mids and highs right at you. It would be better to try the amp with more typical boxed speakers

 

[primosounds]

Primo, It's possible I made a bad assumption (not having seen your actual schematic) on where you've got the tap for the B+ line feeding the output transformers. I assumed it was after the first filter choke. Correct me if it's somewhere else.

Regarding the scale of the graph, you can "zoom" in and out on PSUD. Take your mouse and position it over the graph a little to the upper left of the position where the step starts. Hold down left mouse button (on PC anyway), and drag to lower right a bit. Let go after dragging to appropriate spot. This will zoom in. You can reset the zoom by clicking the magnifing glass icon top right of the graph pane with the red X superimposed. Sometimes you need to zoom out first before selecting the area to zoom in on, if the area you want to zoom in on is right at the top of the graph pane, and you do that by clicking on the magnifying class icon with the up arrow.

I'm talking about these icons:
View attachment 862998

Going left to right, first magnifying glass is zoom in, second is zoom out, third is cancel whatever zoom you had previously applied. So, you can zoom in in two different ways, by using the magnifiying glass icon, or by selecting and dragging with the mouse.

Hi KW, i have been playing with psud alot and have some questions, still. First, is how do you save the graphs? My computer wants to save them in Adobe, but then i can not open them when i go back to retrieve it from my computer files. 2. is there some write up which explains the graphs and how they relate to what is heard, at least in general terms. For example you were wondering about ringing in the ps, yet what does that look like graphically? Is any of that explained in the notes or by Duncan amps? cheers.
BTW, i made my ps , the one you were helping me with, 11uf x 6H x 33uf x 6H x 200uf. The rectifier is a Graetz + 5ar4 into a 220ma load.

 

[trobbins]

I use the MS snipping tool to grab some or all of the PSUD screen - alternatively PrintScreen and then an image editor. Unless you have the full Acrobat software then free Acrobat is sometimes a real pain.

PSUD2 is a very general tool, used for many different reasons and many different applications - it's not really aimed at giving an opinion on any particular aspect of power supply design. Some of us use the tool to gauge ringing performance, but it's up to us to then interpret that information.

 

[Retrovert]

PSUD2 prints schematics and plots.

This allows one to install a PDF printer driver (free or commercial) to generate a file at higher quality than screen resolution.

 

[primosounds]

So, I went back to my 60W UL amp PSUD simulation and adjusted the simulation window to look at the current at the first (reservoir) ........
Am I looking at the right stuff?

No one seems to know the answer to your question. Did you find more info or are you still waititng?
Well, with this PSUD there must be an ideal graphical representation that one is trying to achieve? From using it i think that it would be a straight line starting at close to 0, going straight up and then 90 degrees right angle and straight across. Is this correct in an ideal graph? Real world sets off at an angle, the steeper the better?, and gently tops off (the sharper the turn the better?) and goes straight across the graph. If i am wrong about this please set me straight.
Well, playing with psud is fun but i wonder how what is graphically represented relates to real world listening experience. I realise that is very dependent on speakers, and amp design but still i would have thought that there was someone who tried to correlate findings with listening experiences given the amount of use that psud findings are referred to. If any such information exists please let me know. cheers.

 

[trobbins]

Thorpej, perhaps best to post or link to an image of your psud2 circuit and plot - then others can more easily appreciate your situation.

 

[kward]

Hi KW, i have been playing with psud alot and have some questions, still. First, is how do you save the graphs? My computer wants to save them in Adobe, but then i can not open them when i go back to retrieve it from my computer files. 2. is there some write up which explains the graphs and how they relate to what is heard, at least in general terms. For example you were wondering about ringing in the ps, yet what does that look like graphically? Is any of that explained in the notes or by Duncan amps? cheers.
BTW, i made my ps , the one you were helping me with, 11uf x 6H x 33uf x 6H x 200uf. The rectifier is a Graetz + 5ar4 into a 220ma load.

I use the screen capture function of Windows 10. (ALT+Print Screen). Then I paste that into Paint and save it as a .jpg file. I've actually never read the documentation on PSUD. I just learned by fiddling around with various example circuits. This thread has touched on two sources of ringing: from the power transformer directly (via use of a rectifier which starts and stops current flow abruptly), and ringing from LC filter stages after the rectification stage, which is easy to see in simulation with PSUD.

If I had time and nothing else to do, I would like to study the effects on sound quality of a ringing power supply from these two sources. Seems like you would be able to measure this at the speaker terminals with either an oscilliscope, a distortion analyzer, or a spectrum analyzer, and if you can measure it, you might also be able to hear it. Anyway all of this is still speculation on my part. One day maybe I will build a purposefully poorly designed power supply next to a well designed power supply and test by driving the same audio amp, with same speakers, and same source material, and see what I can measure and discern in listening tests.

 

[primosounds]

Well, it sure seems like there is a need to have the psud results correltated to real world results.

 

[Brice]

I am doing this now. Re-configuring a PS and adjusting the voltages based on the simulation.
I'll post my results when I get to this in a couple of weeks.

 

[primosounds]

That would be very interesting. I am not sure if the psud "ideal" graph is the best "sounding" power supply but at the least it gives a good idea of your voltages and current use which is important of itself.
Since i have been making my own amps lately and they are mostly single ended amps. without the push pull advantage of noise cancellation the power supply design is critical to a quiet amplifier, along with layout and component placement.