Yes, that was referred to towards the end of the original thread. It shows certain signs of being inspired by Carlos's approach but without really understanding it. Also the PCB grounding isn't something I'd suggest anybody copies.
I'm going to take back what I said about star grounding in power supplies. I've just simulated with LTSpice two different layout styles, and really it makes no difference to the output noise, so please ignore me.😀
What I found does make a significant difference is the ESR of the two 10,000uF caps. So there is definitely mileage in paralleling smaller values here rather than using just two big ones. Oh, the splitting of the 1R into two 0.5R makes a few dB of improvement too, depending on the ESR of these caps.
HI abraxalito,🙂 thanks for the tip on star ground. Will post it after correction you said.😉
thanks
Andrew
What do you think about putting small value (0.1 ohm?) power resistors on the secondary side of the transformer before the bridge and/or between the bridge and first reservoir caps to reduce charging currents and provide additional RC filtering? Would you have a preference for before or after the bridge or both?
What do you think about putting small value (0.1 ohm?) power resistors on the secondary side of the transformer before the bridge and/or between the bridge and first reservoir caps to reduce charging currents and provide additional RC filtering? Would you have a preference for before or after the bridge or both?
Last edited:
Does it matter if it goes before or after the bridge though?
In my simplistic worldview of electronics I would have thought if it were before the bridge it would be somehow 'isolated' from the first smoothing caps by the diodes and not work as a filter, which it would definitely do were it positioned AFTER the bridge.
That said, I've seen a few schematics where it goes before the bridge. In those cases is it just serving to limit inrush current into the bridge and caps at switch-on?
In my simplistic worldview of electronics I would have thought if it were before the bridge it would be somehow 'isolated' from the first smoothing caps by the diodes and not work as a filter, which it would definitely do were it positioned AFTER the bridge.
That said, I've seen a few schematics where it goes before the bridge. In those cases is it just serving to limit inrush current into the bridge and caps at switch-on?
Last edited:
If you are looking for such a solution, just take a smaller transformer. That will give you more R, better filtering, less peak currents. You save the resistors and save on the xformer and save space. Sometimes you can have your cake and eat it too 😉
jd
Yes,
the resistance of the transformer is often used as the sole contributor of the first r in the rC filter.
You can add on the cable resistances to correct to the real value of r and you can add on a discrete resistance for an enhanced r value.
But, for certain the first r is NEVER zero ohms.
A similar logic follows for LCLC filters.
The cabling before the Cs never has zero inductance, resulting in lClC (these l beings lower case Ls and not i)
In reality 2pairs of capacitors forming a dual polarity PSU is actually lrClrC and we generally ignore the lr components.
the resistance of the transformer is often used as the sole contributor of the first r in the rC filter.
You can add on the cable resistances to correct to the real value of r and you can add on a discrete resistance for an enhanced r value.
But, for certain the first r is NEVER zero ohms.
A similar logic follows for LCLC filters.
The cabling before the Cs never has zero inductance, resulting in lClC (these l beings lower case Ls and not i)
In reality 2pairs of capacitors forming a dual polarity PSU is actually lrClrC and we generally ignore the lr components.
Last edited:
If the R is in series with the diodes, the order doesn't matter. IOW, it doesn't matter which comes first and which 2nd. The current goes through the two elements in series and doesn't change if the order is reversed.
The voltage on each the elements stays the same (because the current is the same), and since voltages in series add, they add the same no matter the order in which the current flows through them.
jd
Just out of curiosity, is there any data which shows the efficacy of this approach compared to a more conventional design?
Sy
Ben Duncan did a series of articles in HiFi News in the late 80's I believe which included a power supply module with R-C-R-C filtering using multiple smaller value reservoir caps in parallel to replace the one or two larger values found in a conventional supply. The idea behind paralleling the caps was to reduce their equivalent series resistance (ESR) and equivalent series inductance (ESL) for reasons which will soon be apparent. He had some graphs showing a dramatic reduction in ripple with the R-C-R-C approach, not surprisingly. You might try Googling it.
As Andrew correctly says, all conventional power supplies are really R-L-C supplies, with the R and L being formed by the resistance and inductance of the transformer secondaries and the wiring before and between the reservoir caps. Adding series Rs or Ls in the form of discrete resistors and/or inductors just increases the inherent values of the supply.
The series resistance (Rseries) forms a voltage divider with the impedance of the reservoir caps (Zcap). For a single RC stage, the attenuation will be 20 * log10 [Zcap/(Zcap + Rseries)] dB. Zcap falls with frequency, reaching a minimum value equal to the cap's ESR at its resonant frequency before rising again due to the ESL. The cap's ESR will therefore ultimately limit the attenuation available, so the lower the ESR, the more attenuation for a given Rseries, hence the paralleling of caps in Duncan's supply.
You could always increase attenuation by increasing Rseries but that's only really an option for preamp power supplies because of bigger voltage losses at the currents typical for power amps.
I think I've got that right!!
Ben Duncan did a series of articles in HiFi News in the late 80's I believe which included a power supply module with R-C-R-C filtering using multiple smaller value reservoir caps in parallel to replace the one or two larger values found in a conventional supply. The idea behind paralleling the caps was to reduce their equivalent series resistance (ESR) and equivalent series inductance (ESL) for reasons which will soon be apparent. He had some graphs showing a dramatic reduction in ripple with the R-C-R-C approach, not surprisingly. You might try Googling it.
As Andrew correctly says, all conventional power supplies are really R-L-C supplies, with the R and L being formed by the resistance and inductance of the transformer secondaries and the wiring before and between the reservoir caps. Adding series Rs or Ls in the form of discrete resistors and/or inductors just increases the inherent values of the supply.
The series resistance (Rseries) forms a voltage divider with the impedance of the reservoir caps (Zcap). For a single RC stage, the attenuation will be 20 * log10 [Zcap/(Zcap + Rseries)] dB. Zcap falls with frequency, reaching a minimum value equal to the cap's ESR at its resonant frequency before rising again due to the ESL. The cap's ESR will therefore ultimately limit the attenuation available, so the lower the ESR, the more attenuation for a given Rseries, hence the paralleling of caps in Duncan's supply.
You could always increase attenuation by increasing Rseries but that's only really an option for preamp power supplies because of bigger voltage losses at the currents typical for power amps.
I think I've got that right!!
Last edited:
Yeah, I've read the Duncan articles which show the expected results (see Scroggie's articles on the same subject back in the 1950s). But there are performance claims here that go beyond reducing 100/120Hz ripple and it would be interesting to see any backup data beyond, "I plugged this in, and it sounded better," with no controls or verification.
Sorry, didn't intend to teach granny how to suck eggs! I guess you were talking about the 'snubber' here were you? No, haven't seen anything convincing on that score.
I think the value of R-C-R-C filtering is about more than just 100/120Hz ripple though. Articles I've seen show harmonics of these frequencies coming out of the bridge at fairly high amplitudes up to the high 100s of kHz to low MHz range. Those harmonics will go straight through linear regs like shite though a goose and end up in the amp circuitry or get radiated around by sloppy wiring with the same end result - RF intermodulation distortion or even ocillation.
I think the value of R-C-R-C filtering is about more than just 100/120Hz ripple though. Articles I've seen show harmonics of these frequencies coming out of the bridge at fairly high amplitudes up to the high 100s of kHz to low MHz range. Those harmonics will go straight through linear regs like shite though a goose and end up in the amp circuitry or get radiated around by sloppy wiring with the same end result - RF intermodulation distortion or even ocillation.
Last edited:
Just so happened that I played with LTSpice quite a lot today wondering where the best place was to fit the series resistors. I lost quite a lot of output voltage putting the resistor before the first cap compared to after it. So I wouldn't recommend having the resistor there from an efficiency point of view. Similar to what Jan has said, if someone wants a resistor there it makes more sense to increase the secondary resistance by winding it with thinner wire, saves a component. The efficiency of putting a resistor there is poor because the current is most impulsive (smallest conduction angle) - better I think to have a relatively smaller first cap if the aim is to limit inrush currents, rather than a series R here.
Fully agree.
But there's another angle to this R-C-R-C. It may well be that it is a great way to reduce ripple, but what about that amp downstream? The rectifier and caps are disconnected from the mains some 80-90% of the time (when the diodes are reversed), and during those intervals, the caps have to provide the amp with power. Guess what: the amp also sees those resistors between the caps. So, when the amp tries to suck energy from those caps, that is also resisted (ahh! there's where that name comes from 😉 ) by the resistor between the two caps. So the first cap has only limited utility to provide energy to the amp.
So like always in engineering, you need to balance conflicting requirements: the amp would love an unlimited cap with zero ripple, the xformer and rectifiers will blow up trying to squirt current into an unlimited cap and even if they survive it will have gobs of ripple. This time, you CAN'T have you cake and eat it too...
jd
Yep, all good points which I hadn't considered in that way before. But I did approach that from a roundabout way in my sim runs today.
I noticed that the harmonics from feeding just a resistive load were falling off faster than 6dB per octave. Just as an example, 100Hz at around 2A gave -20dBV but 4kHz was about -100dBV. So then I thought "shouldn't I try to match the rate of fall off of the ripple harmonics to the PSRR of the chipamp?". It was only then that I realised I'd have to include simulating an amp putting its own ripple back on the supply to see what was really going on.
At that point, with both a 50Hz power source and a 40/400/4kHz audio source the simulation ran very much slower so I didn't make much more progress. 😛 The LM3886 also has very assymmetric PSRR, the +ve side being very much better than the -ve - so would it make any kind of engineering sense to use a smaller cap on the +ve side for this amp?
- Status
- Not open for further replies.