Here’s a nice simple ripple eater
https://hifisonix.com/projects/hifisonix-micro-ripple-eater/
I’ve called it the ‘Micro Ripple Eater’ (MRE) because it’s compact but it can deliver up to 10A per rail with a decent heatsink.
Any questions, I’ll be happy to answer them
🙂
https://hifisonix.com/projects/hifisonix-micro-ripple-eater/
I’ve called it the ‘Micro Ripple Eater’ (MRE) because it’s compact but it can deliver up to 10A per rail with a decent heatsink.
Any questions, I’ll be happy to answer them
🙂
Last edited:
Thanks for the inexpensive and nice design! Very cool!
Did you consider using CFP for lower dropout? Maybe there is some advantage to EF but I am wondering if the fear of CFP oscillating in this context is well founded.
Did you consider using CFP for lower dropout? Maybe there is some advantage to EF but I am wondering if the fear of CFP oscillating in this context is well founded.
Thank you 🙂
Yes - if you go or CFP, it becomes quite restrictive on what you can use for the pass transistors. With this approach, the transistors are not critical - you can basically use what ever meets your output current and worst case Vce situation is.
Yes - if you go or CFP, it becomes quite restrictive on what you can use for the pass transistors. With this approach, the transistors are not critical - you can basically use what ever meets your output current and worst case Vce situation is.
Thanks, that makes a lot of sense. I have a ton of 21193/4 I wasnt sure what to do with. So the cost of using them is zero.
Last edited:
Member
Joined 2009
Paid Member
The original “Ripple Eater” was an active ac-shunt, a JLH design if I remember, but this one looks like a cap multiplier, or I am missing something!?
(Sorry - binned my first reply to @Bigun before Bonsai's post - because I've just put in 18+hrs on the quotidian and the resulting text was nigh-unintelligible in any wider sense🙂
Yes -it's a cap-multiplier, but look at the turnover frequency: 3K9 x1000uF means about 0.05hz.
If your local mains (UK here) is 50Hz, that's 3 orders of magnitude below, potentially a -60dB filter on ripple; so Bonsai's claim of -40dB is very, very conservative!
But such a low turnover also means - a very slow start-up; and that's not universally acceptable. Some PA designs will squawk a bit on that...
So you could, say:
- add a zener about 1-2v above your target output voltage in parallel with C1 and C2, reduce them to 100uf, start-up faster, and lose no significant voltage regulation whatsoever; in fact, potentially improve it under certain circs (esp if your raw psu is poorly regulated/ high/wild in voltage. Consider total V drop x I avg across the board and heatsink requirements, though)
- All the usu Cap Multiplier tricks to suit requirements can apply, in other words...:
- hey, for one you could also split the 3k9 in two parts, add another 100uf to 0v (say 1-1K8/100uf/1-1k8/100uF) to gain 40dB/octave rolloff, but from a higher frequency, for about 8-10x faster start-up - nb the limit on HF attenuation will be the layout, and the parasitics in the devices & parts used etc.)
This is all stuff to play with in a neat, compact, v high-current capable pcb, - as I see it.
Grateful to @Bonsai for supporting such experimentation : )
Yes -it's a cap-multiplier, but look at the turnover frequency: 3K9 x1000uF means about 0.05hz.
If your local mains (UK here) is 50Hz, that's 3 orders of magnitude below, potentially a -60dB filter on ripple; so Bonsai's claim of -40dB is very, very conservative!
But such a low turnover also means - a very slow start-up; and that's not universally acceptable. Some PA designs will squawk a bit on that...
So you could, say:
- add a zener about 1-2v above your target output voltage in parallel with C1 and C2, reduce them to 100uf, start-up faster, and lose no significant voltage regulation whatsoever; in fact, potentially improve it under certain circs (esp if your raw psu is poorly regulated/ high/wild in voltage. Consider total V drop x I avg across the board and heatsink requirements, though)
- All the usu Cap Multiplier tricks to suit requirements can apply, in other words...:
- hey, for one you could also split the 3k9 in two parts, add another 100uf to 0v (say 1-1K8/100uf/1-1k8/100uF) to gain 40dB/octave rolloff, but from a higher frequency, for about 8-10x faster start-up - nb the limit on HF attenuation will be the layout, and the parasitics in the devices & parts used etc.)
This is all stuff to play with in a neat, compact, v high-current capable pcb, - as I see it.
Grateful to @Bonsai for supporting such experimentation : )
Last edited:
No it's a special cap multiplier that eats the ripple and sweats it out through the heatsink, new bio-electronic technology 😀
To understand how the ripple eater works in practice, you have to think about the cut off frequency (c. 0.05Hz as Martin pointed out). The pass transistors remain saturated and block any ripple voltages above this.
If you have a step load change, say because the amp suddenly draws a lot of power, the output voltage will go down to the new voltage (so a step change) but the ripple will still be removed. Think of it as a low pass filter. Likewise, when the current draw goes down, the reverse happens. So what is happening in practice is you are giving up a bit of voltage (1-2V across the series pass transistor) to shift the ripple down in frequency to below 1 Hz.
Removing ripple like this effectively raises the amplifiers PSRR by >30 dB in practice at 50 Hz and more at HF.
Note that the MRE will not remove SMPS HF noise - this requires a proper HF filter. However, for unregulated SMPS that present some LF ripple under load at the output, the MRE will remove it.
If you have a step load change, say because the amp suddenly draws a lot of power, the output voltage will go down to the new voltage (so a step change) but the ripple will still be removed. Think of it as a low pass filter. Likewise, when the current draw goes down, the reverse happens. So what is happening in practice is you are giving up a bit of voltage (1-2V across the series pass transistor) to shift the ripple down in frequency to below 1 Hz.
Removing ripple like this effectively raises the amplifiers PSRR by >30 dB in practice at 50 Hz and more at HF.
Note that the MRE will not remove SMPS HF noise - this requires a proper HF filter. However, for unregulated SMPS that present some LF ripple under load at the output, the MRE will remove it.
Hello Jan,
Yes for class AB heat sinking requirement is minimal. For class A you’ll probably need a bit more
Yes for class AB heat sinking requirement is minimal. For class A you’ll probably need a bit more
Indeed Very nice.
In the BOM you specify NJW3281 and NJW1302, those are not not easy to source and in my opinion not critical for the operation of the circuit.
Maybe you should include som alternatives in the BOM?
Stein
In the BOM you specify NJW3281 and NJW1302, those are not not easy to source and in my opinion not critical for the operation of the circuit.
Maybe you should include som alternatives in the BOM?
Stein
Last edited:
Thanks Stein.
Indeed, it’s very tolerant of the transistors used.
For the TO3P/TO-247: any device that with a Vce > worst case + or - supply voltage will work and that meets the current requirements of your application. To cater for amp faults and very low speaker R testing, I suggest 10 A or greater collector current.
For the TO-126 device: same comment re voltage as above and at least 500 mA collector current.
Indeed, it’s very tolerant of the transistors used.
For the TO3P/TO-247: any device that with a Vce > worst case + or - supply voltage will work and that meets the current requirements of your application. To cater for amp faults and very low speaker R testing, I suggest 10 A or greater collector current.
For the TO-126 device: same comment re voltage as above and at least 500 mA collector current.
Hello Bonsai
greetings if 1000uf 63 volts and 100uf 63 volts caps are used can
55dc+/- supply e used.
warm regards
Andrew
greetings if 1000uf 63 volts and 100uf 63 volts caps are used can
55dc+/- supply e used.
warm regards
Andrew
This micro has few components more than the original sx/kx amp ripple eater?? (which i have in the works now)
Should i try to add R1, R2, R9, R10, D5, D6 and C3, C4 to older board, if even possible?
Should i try to add R1, R2, R9, R10, D5, D6 and C3, C4 to older board, if even possible?
OK, this is pretty nifty, but in the final result, what does this do for an amp above just having a decent cap bank? Sonically. I put about 27,000 on each rail of my 60W amp, and can feel nothing on the woofer. Whatever ripple is there is below the noise floor of the Focusrite I use on my bench for IO. The IPS and VAS are isolated by a pi filter on the rails, good stiff CCS and CM. I was going to put a multiplier rather than the pi filter, but don't see the need.
The slow start is very nice BTW.
The slow start is very nice BTW.