Ripple rejection???

[Chuck H]

One of my designs that I build is a 2xel84 amp. I have a customer who wanted the same amp in a 4xel84 model. Sounds amazing, but... These amps are built to pound the piss out of the power tubes and this 4xel84 build is ripple'ing more than I would like. The owner has no complaints. It's not as bad as an AC30 or similar but I am going into this amp to make some upgrades and I would like to squash the extra ripple while I'm in there. So...

What is the better arrangement for ripple rejection? Parallel caps or series caps with balancing resistors? I'm guessing it's a wash but the added ESR of series caps may be a factor. All experiences welcomed. TIA

[R.G.]

You don't say, but I'm taking it that the question is about power supply filter caps, especially the first filter cap, so that's where the answer is pointed.

To a first order, ripple is suppressed by the total raw capacitance in the first filter cap. You want as many farads there as you can get. There is a point of diminishing returns, but more uF is always smoother. In that sense, parallel caps or series caps is a moot point, you just want the most uF you can get, series or parallel.

You can make the amp eat up rectifier tubes that way. Rectifier tubes have a maximum capacitance specification, because the more capacitance in the first filter cap, the higher the current peaks every charging cycle. You exceed the peak cathode current at some point, and the tubes start dying. If you're going to just add more capacitance, it has to be with semiconductor rectifiers, not rectifier tubes, or you have to do things more cleverly than just adding more caps.

ESR is a second order effect. It can be a problem in high current supplies, but no tube amp supply is really "high current" by today's standards. Putting caps in series also puts the ESRs in series, making the result bigger than for a single cap, while paralleling them parallels the ESRs and makes the result smaller. However, you would only need to put higher uF caps with lower voltage ratings in series. These caps have smaller ESRs in general anyway. It may not matter. Check the specs on the caps you use, compute the approximate ESR. ESR adds a spike on top of the ripple peak, mostly. It's likely that ESR is not a big deal with new, modern filter caps.

You can also get clever. Splitting the filter caps into a first filter cap and a second filter cap with an inductor between them will provide much more ripple suppression than just paralleling the same two caps. Or you could use an active-capacitor filter, with a high power source follower and a cap on the gate. This sacrifices some of the peak voltage, but gives you the electronic equivalent of a huge second filter cap.

[EFK]

Chuck forgive me for interjecting a more amateur question here, but I'm very interested in the topic. When you say, "...this 4xel84 build is ripple'ing more than I would like" are you indicating an audible effect or something that is manifesting only on a scope? And what exactly is the audible effect if so - background hum, ghosting or phasing sound at higher volume, scratchy nasty noises (I've found that can happen with too little initial filter capacitance), or ??? I'm very interested, especially when it comes to the concept of trying to preserve the same tone and feel of a 2 tube amp while moving to a 4 tube amp, or vice versa, because won't changing the capacitance change the feel of the amp?

[Chuck H]

Thanks R.G. I am talking about the main filter. If it's just a matter of capacitance that's not a problem since I'm using a diode rectifier. This is probably the route I'll take then as EFK makes a good point about preserving the tone. Changing to an C/LC filter or a MOSFET arrangement will alter other things more than I want.

EFK, It's ghosting. Classic ghosting like you hear in nearly all 4xel84 amps. I also wonder if, say, you are using an 80uf main filter for a two tube amp and need 160uf for a four tube amp to achieve the same ripple rejection, what else changes about the tone or feel of the amp. It may be using the same value filter for 2 or 4 tubes IS what's incorrect and the higher value filter for the higher current makes the two circuits more similar than using the same value.

[trobbins]

You may well find that ESR is lower with a series connection, due to the use of lower voltage rating caps. I typically don't use electrolytics above the common 400-450V max of commercial offerings with good spec sheets - and of the few 500-600V new caps I've come across, they have had pitiful specs with no data on ESR or more importantly ripple current ratings.

[Chuck H]

Thanks. It all gets real easy now. I'm just going to parallel another filter identical cap on the main filter.

Using duncan PSU I mocked up the PS for this amp and found that indeed ripple is almost directly proportional to current with respect to uf's. ie: if 70ma current needs 40uf for adequate ripple control then 140ma needs 80uf for the same result (actually a tad better). I also did find that ESR plays a roll too. More than three ohms of ESR starts to show poor performance. Three ohms or worse isn't uncommon for older caps or even some new caps. Especially those rated over 450V.

[EFK]

So, could paralleling something like .1uf plastic caps onto 500V caps help with the poorer ESR performance, or am I way off track here?

[Chuck H]

Well... I suppose you could use Duncan PSU to check. Only thing is ESR is not linear and the program doesn't account for that. I'll try it anyway.

EDIT: Nope... The program can't simulate that. I know a lot of buzz is around about this practice but considering the impedances and the frequency of ripple I really wonder what a .1u or even a 1u cap could do. At an impedance of even 10 ohms isn't a 1u cap only working well above the frequency range of ripple current?

[Gingertube]

Tube Amps generally have high levels of intermodulation distortion.
One thing which can be done is to clean up the power supply so that there is one less signal (power supply ripple) for your signal to intermodulate with.
Modern electrolytics have come a long way from the units of even 10 or so years ago - thanks largely to switch mode power supplies which demand low ESR (and ESL). Modern caps will all have sub 1 ohm ESR.
I use and recommend Panasonic ED Series for guitar amps. I use distributed filtering making sure of-course that the first cap which feeds the power tubes via the output tranny has enough capacitance.
Since I don't build a lot of amps I generally buy a bulk quantity of 22uF 450V and then run 3 or 4 in parallel for that first cap.

If you scope the output with the amp running into clipping you should NOT be able to observe any 120Hz ripple on the top or bottom of the waveform envelope. If you can then you need more capacitance.

OR

Therer was a "rule of thumb" which said that ripple voltage (peak to peak) should be less than 5% of the DC voltage. I would set that as the upper limit of acceptable ripple. For HiFi amps I try for less than 2%.

Cheers,
Ian

[Chuck H]

Thanks Ian. Sort of a blanket observation rather than a contribution to the subject though. What have you noted about the relationship between main filter capacitance and an amps current demands and/or what's your opinion on the common practice of using a .1uf or 1uf film cap in parallel with an electrolytic cap to improve it's performance?

[Gingertube]

Chuck H,
I do use a parallel film cap but not on every electrolytic capacitor. I generally use one at the Output Tranny Centre Tap main power feed and at the other end of the line at the input preamp tube.
Generally the 1/100th rule works well - That is use a 220nF across a 22uF or similar.

Electrolytics (particularly the larger values) turn inductive above their self resonant frequency. That is what these caps are added to address. At the main (1st) filter site the parallel film cap helps suppress diode switching noise. At the other end (1st gain stage) it helps keep high frequency response up and hence the harmonic content. That is generally knocked down as you progress through the preamp and become less important.

Cheers,
Ian