Would you run a cap under these ripple current conditions?

I need to replace big 22mF (22 000uF) capacitors in my vintage amp that have no match in size so I would place multiple parallel capacitors on a board that will go in place of the big beasts. I have found suitable replacements that have a rated ripple current of 2800 mA r.m.s./85℃, 120Hz. I have computed their ESR from Tan(δ) (0.18) to 0.05 ohm and did some simulations (see result below) with this value set on the simulated caps.

Capture d’écran 2023-11-15 182622.png


The computed RMS value is well below the 2.8A rms allowable ripple current but that 5A peak has me worried. Note that the voltage source simulating the power transformer has no parasitic series resistance specified, if I set a 0.5 ohm parasitic resistance for the transformer the peak drops to 1.9A (872 mA RMS). Also note that the current draw set for the simulation (2.7A) would be for the amp running at full power in class A, but the amp is AB and of course it is never run at full output.

As I write this I feel everything is well and good but I guess I'd like some reassurance 😱... Just to avoid fireworks!

Thanks in advance!
Joris
 
Last edited:
Buy an accurate capacitor ESR meter that will test 4.7 millifarad electrolytic caps, and measure it yourself. Be sure to set the meter to its 120 Hz measurement setting. There are lots of them on the market; the one I like is called DER EE DE-5000. Amazon sells it (link 1) but usually you can find it for 30% less on eBay if you look hard. Just make sure you purchase a complete measurement package containing DE-5000 and ALL the accessories and probe options.

Then include ESR in your simulations, and also use SPICE models of real diodes (I suggest VS-E5PH3006); your screen capture image suggests you simulated using ideal diodes. Real diodes have nonzero dynamic impedance.

_
 
Last edited:
Thanks Mark for your reply.

The diode specs are not displayed on the simulation's schematic but I used the model for the RFN10NS3S 350V/10A diode. The amp uses the discontinued KBU802G which is 100V/8A. Switching to a closer spec'd RF1001NS2D (200V/10A) yields similar results.

Thanks for the recommendation and link, I'll save a search on ebay for a deal. However if at all possible I'd like to have a bit of assurance before commiting 100$ in PCBs and caps that are not certain to fit the bill... Isn't the Tan(δ) specification from the datasheet at all reliable? I have adjusted that value to account for the large capacitance as directed by the manufacturer as well (When capacitance is over 1000μF, tanδ shall be added 0.02 to the listed value with increase of every 1000μF).
 
An idea of what the capacitor pcb will look like - 4 of them to replace the large caps labelled C4-x in the picture. The actual caps will be taller than the 3D rendering. This amp is very crammed up and was know to have over-heating problems due to a sealed design - I drilled ventilation holes after purchase 😎

Rega Elicit output caps board outline v.0.1.jpg
IMG_20211209_160021_0691.JPG
 
The computed RMS value is well below the 2.8A rms allowable ripple current but that 5A peak has me worried.
No problem. The failure mode is overheating, so peaks aren't relevant, just the rms being within spec. Electrolytics can handle huge over current spikes before failing, think of a xenon flashtube circuit where the cap is discharged fully in under 1ms. Still its best not to deliberately short them deliberately, that probably will be overstressing them - as well as spot-welding the terminals...

The ratio of rms to average current does increase the more spiky the waveform though, so its good to aim for a large conduction angle with this type of circuit - it also reduces stress on other elements. In your sim you should ensure the transformer resistance is modelled as this will affect the current waveform.
 
  • Thank You
Reactions: SomeJoe
Make sure to retain as much copper on your pcb's as possible, i.e. use thickmost traces
The add-on board is designed to be a drop-in replacement, so the stock pcb will remain as is - Dupont pins will interface with the contacts on the amp board. (the plastic bits will be put topside). I used star-wired 5mm traces which should be capable of carrying more than 6A on regular 1oz copper. I streamlined the pcb so that there is less "sandwiched" area.

Rega Elicit output caps board outline v.0.1 back.jpg
Rega Elicit output caps board outline v.0.1.jpg


Digging further down Digikey's catalogue I found a nice cap series (Rubycon ZLJ) rated 3.5A ripple RMS@120Hz, 10 000 hrs@105°C. 33% more expensive for 20% less capacitance, but I'd rather have a thougher part and peace of mind. The largest 50V cap is 3.9mF so a bit less total rail capacitance at 19.5mF instead of 22mF, but that's within the error margin and I bet those modern caps, paralleled, will blow away the old ones even with that 10% less charge.

Thanks to all, I feel confident now and I learned a lot, this forum is great!
Joris
 
Last edited:
I need to replace big 22mF (22 000uF) capacitors in my vintage amp that have no match in size (snip)
What did you find in the right size? If you found some, you should just use today's denser capacitors and have quieter rails.
Is an array of 5 the best that fits there? Can we see a better picture? It appears you're not strictly bound by the original cap's diameter...
 
The original caps's size is 50 mm diam., height 44 mm, so "chubby" caps having larger diameter than height and the amp has a low profile case design that leaves not much free space over the caps. I have searched both Mouser and Digikey for a replacement without success; This 5-cap arrangement was the best I could find that could get close to the original capacity while fitting in 44 mm height, including 1.6mm of pcb thickness and soldered leads. I do not want to drill holes in the case and have taller caps sticking out.

IMG_20211209_155815_0688.JPG
IMG_20211209_165821_0693.JPG
 
If you could stretch to 45mm seated height,
I have measured 46mm free space over the amp pcb, so technically yes these would fit height-wise, but I must mention the case gets somewhat hot, and 1mm clearance or less is too close for comfort. Only one 15mF might do, but I'd rather be closer to the 22mF if I can; two per position can't fit (see below right). That's sad, this series has the 22mF but at 50mm seated height - so close....

Capture d’écran 2023-11-16 192855.png


As for drilling holes in the stock pcb, check out the transparent shot below - It would be a challenge. I still believe my small add-on PCB is easier to implement.

IMG_20211120_193548_0645.JPG
 
The original caps were 50mm Dia while the silkscreen shows 30mm Dia right?
Looks like you might win a few mm if you could hang the new caps over the edge of the pcb too. But it would take some jiggering of the pcb traces underneath...
 
Drill holes in the PCB for plastic tie-downs.
Yeah... That could work, there is a possibility to fit the 22mF 30mm diam. x 50mm mentionned above on its side, with a pair of tie-wraps. Involves pcb drilling plus some extra wires fro the new caps to the PCB, which will have to be large I think to avoid oscillations?

Anyways next week end I will confirm the caps are faulty and if so will remove one, try to find a similar 30mm diam. x 50mm height in the parts bin and try fit. Thanks
 
I used star-wired 5mm traces which should be capable of carrying more than 6A on regular 1oz copper.

Yes they are, but they will add the other one or two milliohm to your ESR. Why not minimize that added resistance by utilizing the whole copper plane? Since you're obviously going dual layer, you'd end up with two full ground planes, so to speak. You might even carve those boards out of copper clad by hand then, no need for a board house at all 😉
 
  • Like
Reactions: rayma
Since you're not re-doing the whole layout and your amp is not a chip amp, I am not sure if the reasoning applies here. But then again, I'm not sure I fully understood what was said in that thread, as some of the suggestions seem rather strange to me. And while we're at it: please understand my posts here as my suggestions, and not as the only way to make it right ™. It's just how I would do it, based on what I think to know.