Required voltage rating for PS smoothing/filtering caps?

[Deke609]

Mouser sells some Kemet AC filter metalized polypro caps rated 700 VDC and either 330 or 440 VAC rms that i think I might be able to make fit. Do I need 700V AC/DC or just DC?  Once rectified, I thought the current would be considered DC for cap purposes, since it never goes negative. But I'm not certain.

cheers and thanks, Derek

[Paul Birkeland]

700DC is fine.  This is just to replace the ones that come after the doubler?

[Deke609]

Many thanks PB. Yes, I'm not touching the doubler.

[Paul Birkeland]

Even a 600V rated film cap is probably fine in those positions.  They don't really take much of a beating in terms of AC current, though the doubler caps sure do!

[Deke609]

Awesome! Thanks PB. That may make some smaller diameter caps an option.

cheers, Derek

[Deke609]

And here they are. 100uF 700VDC each. So a little shy of the stock capacitance. Mounted but not yet connected. I need to cut and router some simple risers to raise the (still unfinished!) base by about 2 inches. 


One of the new caps is likely only visiting. I am going to experiment with replacing each of the 120Rs with chokes. A 4H 65 DCR 225mA rated in place of the first R, and a 6H 150R 200mA rated in place of the second R. But PSUD shows that the combination of the chokes and the 2 new caps (or even the stock 110uF effective value caps) results in initial B+ overshoot and then dip -- and from what I've read that looks like a recipe for ringing. So I played around with different configs and found that a 200uF cap in last position completely eliminates it. In PSUD ripple is reduced to about 0.02mA pk-pk! For reference, PSUD shows about 20mV pk-pk ripple with the stock filter config, and approx. 0.8mV pk-pk with the last RC replaced by the 6H choke and a 100uF cap. We'll see if any of this actually works out in reality. I'll scope the amp this weekend to get some baseline measurements before making any changes.

[edited to add PSUD results for the stock filter config and for when the last RC stage is replaced by an LC]

cheers, Derek

[Deke609]

I played around with PSUD some more this morning and found what looks to be an even better looking LCLC config: 4H-10uf-6H-200uF.


I attach screenshots of the various configs: my rough estimate of stock (2X 120R 110uF); the overshoot/dip with two chokes (4H-100uF-6H-100uF); my first attempt at correction (4H-100uF-6H-200uF); and finally what looks to be a pretty good solution (4H-10uF-6H-200uF). That saves me from having to buy a 200uF cap as i can just connect the two existing 100uF caps in parallel.


Does anyone see/foresee any issues with the 4H-10uF-6H-200uF config? Anything in particular that I should be watching out for?Any big problemswith my PSUD modelling?


cheers and thanks, Derek

[Paul Birkeland]

FYI in the model, I would use a 160mA constant current load rather than the 10K resistive load you have, though this shouldn't really change your results.

[Deke609]

Thanks PB. I ran the models with 160mA constant current as you suggested and the only change I noticed was about a 30V drop in B+. But since the drop is the same for both stock and modified, I figure that's ok.


cheers, Derek

[EricS]

These are such interesting results - thanks for sharing the sims.  In the Class-A solid state world, the smallest ripple tends to result from equal size caps in the pi filter.  I would not have thought about such a "lopsided" approach to cap size for a tube build.  I have not yet measured PSU ripple in my 300B prototype.  That will happen once I am satisfied that it makes music.

[Deke609]

In the Class-A solid state world, the smallest ripple tends to result from equal size caps in the pi filter.  I would not have thought about such a "lopsided" approach to cap size for a tube build.

Eric - I suspect there's no difference between SS and Tube worlds for ripple. The lowest ripple I got from my models was 0.02mV with the combo of 100uF and 200uF. But the shape of the curve was still a bit wonky and I fear oscillations/ringing, so I was looking for a smooth voltage rise. My latest config with 10uF and 200uF results in about 0.5mV ripple - so 25 times more. Based on my playing around with PSUD it looks like ripple is largely function of total capacitance. For S&G's, I modeled the chokes with 2 X 500uF and got negligible ripple well under 0.01mV.  But the voltage rise time was not so good - taking a lot longer to get to B+. I figure (but do not in fact know) that rise time may be relevant to transient response - so I figure the faster the better (provided it is smooth). 


The 10/200 uF combo appears to give me faster rise time than stock with about 40 times less ripple - and this to me looks like a good compromise between fast response and low ripple.


But I'm just guessing here. I'm still new to all of this

[edited to correct my lousy arithmetic: 20mV ripple (stock) / 0.5mV ripple (modified) = 40 times, not the 80 I initially stated - sheesh!]

cheers, Derek

[Raymond P.]

Looks like one of your goals is to prevent ringing on B+ when power is applied? A resistor or current source load to model the rest of the amp probably (obviously?) isn't going to show what's happening on B+ since the rest of the amp has a "wake up" phase (tubes warming up, etc.) Is it possible to load your PSU model with a piecewise current source that has a ramp up period? I guess the difficulty with that is modeling the ramp up behavior. I suppose you could scope your amp's B+ to get some approximation.

[Deke609]

Thanks Raymond. Yes, I am trying to avoid ringing. I have two mostly uneducated thoughts about this: (1) B+ ramp up on turn on is representative of what may happen when there is sudden spike in demand from a loud transient; and (2) ringing in one part of the circuit could cause unwanted effects in other parts of the circuit. One or both of these thoughts may be wrong.


PSUD does allow for a two-step current load. One can set the initial current load and specify a second current load and when it is to take effect. I haven't tried it with the recent models. Do you have any suggested values I should try? Edit: Oh, I see that you;ve already suggested that I scope the amp to estimate this. Thanks - I will do this tomorrow when i am playing around with the scope.


many thanks, Derek

[Deke609]

And here's my four-footed Kaiju. The feet were intended only to be provisional and I had low expectations for how they'd look. But I actually like the look of the thing, even in its unfinished state - which is too bad b/c it is extremely hazardous: the large gap makes for a easy path for finger and power supply to meet. The new caps are still not hooked up , so no danger at present, but I will have to figure out how to make it safe before making my intended changes. Happily, the feet are just press-fit with one retaining screw each from the backside - so they can be removed and reused in a different configuration.

[EricS]

That's clever and very nice looking!  I've been thinking about feet under my chassis and found some on ebay that look decent, but I was wondering if they would hold up under the weight.