Although i share your reluctance towards regulators a good discrete regulator design can sound way better with the 627/634 than any of the 3-terminal ****. An unregulated supply will very likely sound more open and dynamic but will suffer in other areas such as bass definition and soundstaging.
In your case the extra voltage is a bonus rather than a problem. Once you know the current you can calculate resistor values which will drop the excess voltage in a CRCRC filter. You may find PSUD very useful in this.
For an extra-creative and possibly rewarding approach you may want to experiment with LCRC filtering. Chokes for such small currents are cheap and easy to find. PSUD is again essential.
As for the snubber values, you'll have to experiment and see what sounds good to you. Of course resistor and capacitor types will have to be consistent in the comparison.
See ARC PH3 power supply for use of snubbers.
me thinks you're wrong. polyprops are of course non-polarised so you may call them bipolar but not unipolar.
Supply noise and snubbers
I'm joining late on this topic but I'll throw in my 2 cents anyway.
First issue with big caps is NOT the inductance. The problem is an indirect result of the intermitrtent current consumption of chip amps (or any class AB power amp).
The caps are calculated for a certain supply ripple at max current, but what happens with low current consumption, which is most of the time with a typical amp?
The DC voltage rises up. This, in turn, LIMITS THE CONDUCTION TIME OF THE RECTIFIER DIODES PER AC CYCLE. Short current spikes from the line generate high-frequency noise which becomes a source of many headaches.
Pull out the scope and look ar the switching point of the recifiers. Then "zoom in" (which might be tough to trigger correctly) until you see the actual swicthing point on a microsecond scale. A bad supply will produce significant ringing. You really want to tame this ringing with better diodes and/or a snubber.
In practice, running a supply with a constant load is desirable, and that can be done with low power chip amps. But the logic of doing so is questionable, given that you could build a class A amp with that wasted power!
Someone smart can come up with an adaptable supply, but the logic of that is also questionable, as the supply will become much more complicated than the circuit it's driving...
A good solution for chip amps is to pick a chip that has a very good PSRR and design a supply for 10% ripple at max power. This means that we do not just drop in a bigger capacitor bank for the heck of it, but optimize the value per load!
I'm joining late on this topic but I'll throw in my 2 cents anyway.
First issue with big caps is NOT the inductance. The problem is an indirect result of the intermitrtent current consumption of chip amps (or any class AB power amp).
The caps are calculated for a certain supply ripple at max current, but what happens with low current consumption, which is most of the time with a typical amp?
The DC voltage rises up. This, in turn, LIMITS THE CONDUCTION TIME OF THE RECTIFIER DIODES PER AC CYCLE. Short current spikes from the line generate high-frequency noise which becomes a source of many headaches.
Pull out the scope and look ar the switching point of the recifiers. Then "zoom in" (which might be tough to trigger correctly) until you see the actual swicthing point on a microsecond scale. A bad supply will produce significant ringing. You really want to tame this ringing with better diodes and/or a snubber.
In practice, running a supply with a constant load is desirable, and that can be done with low power chip amps. But the logic of doing so is questionable, given that you could build a class A amp with that wasted power!
Someone smart can come up with an adaptable supply, but the logic of that is also questionable, as the supply will become much more complicated than the circuit it's driving...
A good solution for chip amps is to pick a chip that has a very good PSRR and design a supply for 10% ripple at max power. This means that we do not just drop in a bigger capacitor bank for the heck of it, but optimize the value per load!
motherone said:
Carlos's solution and fig 8/C - Fig 9 in the bypassing PDF are the same. A couple of things though.
1. It is not clear the improvement in listening came from PSU damping - in which case the PDF applies, or it is addressing some other form of chip instability related to PSU HF impedance.
2. Even if PSU damping were the case, then total series inductance as seen before the "snubber" location determine the appropiate range of values for the damping resistance, as much as the downstream decoupling capacitance.
This boils down to the fact that what is an optimum combination for a certain instance of component values an layout may turn out no to be as good in another.
I for myself have not tackled the issue as to scope and listen a definite "before" and "after" situation, so am still in the dark at what is possibly going on. If someone else has done, please let it be known.
Rodolfo
Thanks for the advice demogorgon & analogsa. CRCRC filtering was definitely in my plans. I think the best approach would be for me to try out an unregulated vs a discrete regulated supply. As for values of resistors, I am quite willing to fiddle and experiment, but I would be grateful if someone could suggest a ballpark starting point. The pre is based on Pavel Macura's design, though I will be fiddling a bit with the caps. According to his specs, Rin = 100k and Iout = 200ma max. I assume that the resistors need to be 3w or better. I have obtained a 2.2k 25w resistor. Can I check on the voltage drop of different R values by simply connecting this resistor from a rail to ground and then measuring the voltage across it?
The 2.2k resistor can be of the 1/4 watt variety, but should probably be 1/2 watt just to be on the safe side. I'm using a 3.8k 1/4 watt resistor and it's working wonderfully.
CarlosFM, thanks for this great tip, it did make my 3875 BGT GC have better bass for sure and i couldnt notice decrease in sound quality in the higher register like there was when i tried high cap before without snubbers.
of course now you were a help im gonna ask for more
dont you just hate that 🙂 i get it in "bf1942 mapping and modding" all the time!
do you think this would be the right setup to use this method on a bridge/paralell (4x 3886) channel
1 set of rectifiers/ big caps and snubber per channel with each chip having its own 100uf and 100nf caps near the pins. or would it be better for each pair in the bridge to have a shared cap of say 200uf/100pf. that way the component count would be less, and each pair are very close together so the 1 cap would be as cloase to both chips in the pair as the seperate ones would be anyways.
and do you think i will have to mess around with your snubber values much considering it is going to be 4 chips instead of 1.
thanks and have great holidays!
of course now you were a help im gonna ask for more
dont you just hate that 🙂 i get it in "bf1942 mapping and modding" all the time!do you think this would be the right setup to use this method on a bridge/paralell (4x 3886) channel
1 set of rectifiers/ big caps and snubber per channel with each chip having its own 100uf and 100nf caps near the pins. or would it be better for each pair in the bridge to have a shared cap of say 200uf/100pf. that way the component count would be less, and each pair are very close together so the 1 cap would be as cloase to both chips in the pair as the seperate ones would be anyways.
and do you think i will have to mess around with your snubber values much considering it is going to be 4 chips instead of 1.
thanks and have great holidays!
Re: Supply noise and snubbers
Hmm.. This theory seems to make more sense to me than the inductance of the caps. So what's a good suggestion for how to optimize? Anything you can point us to?
Ori said:
Hmm.. This theory seems to make more sense to me than the inductance of the caps. So what's a good suggestion for how to optimize? Anything you can point us to?
analog_sa said:
Well, i _was_ right, altough i didn't know the right name for a bipolar cap.
We Norwegians have english as a third language you know.
My early experiences with Carlosfm snubber
Hi all,
Tried the snubber (on the diodes: 4700uF x 2, plus 4700uF, 100nF bypass; on chip power leads: 1R in series with 220nF, 47uF x 2 Panasonic FC, 100nF bypass)
Been in for one day, not playing music, just powered.
Impressions: Immediately obvious was that highs seem to have more resolution and delicacy and are aslo more seperated (in a good way). Highs with the snubber are definately better than without
Mids: Mids appear to be cleaner and also more seperated from the mix.
Bass: Seems as deep as before, but there is a touch of mid/upper-bass warmth which was not there before.
If the bass "problem" goes away with break-in, I will definately be keeping the snubber in the power supply. 🙂 😱 😀 😉
Carlosfm-thanks for the idea, you've done great work! Let me know if your initial impressions were similar. Does the bass improve with time/burn-in?
Ryan
Hi all,
Tried the snubber (on the diodes: 4700uF x 2, plus 4700uF, 100nF bypass; on chip power leads: 1R in series with 220nF, 47uF x 2 Panasonic FC, 100nF bypass)
Been in for one day, not playing music, just powered.
Impressions: Immediately obvious was that highs seem to have more resolution and delicacy and are aslo more seperated (in a good way). Highs with the snubber are definately better than without
Mids: Mids appear to be cleaner and also more seperated from the mix.
Bass: Seems as deep as before, but there is a touch of mid/upper-bass warmth which was not there before.
If the bass "problem" goes away with break-in, I will definately be keeping the snubber in the power supply. 🙂 😱 😀 😉
Carlosfm-thanks for the idea, you've done great work! Let me know if your initial impressions were similar. Does the bass improve with time/burn-in?
Ryan
Re: My early experiences with Carlosfm snubber
Ryan, change the 220nf cap for 100nf.
Dr.H said:
Ryan, change the 220nf cap for 100nf.
Carlosfm,
What are your impressions and what chnage chould come about from reducing the size of the snubber cap from 220nF to 100nF?
What are your impressions and what chnage chould come about from reducing the size of the snubber cap from 220nF to 100nF?
Dr.H said:
Try it.
Do it as on the schematic, use 100nf.
It's just about right.
Detail, speed, everything is so right and so much better than higher cap values.
I also try the snubber (100nF) and it seems to work: more bass. But I don't find the mid and treeble as good as without big cap + snubber. In my case the sound seems better with 220nF. For a matter of room, I choosed snap-in caps (nichicon), maybe a C038 or C039 woul sound better. Don't you think Carlos, that the value of the snubber cap and its effect depends on the properties of the 10 000µF cap?
What is the brand and model of the filter cap you use?
What is the brand and model of the filter cap you use?
jram said:
Did you also bypassed the big caps with 100nf film?
Anyway, you are all free to experiment with the snubber cap value.
jram said:
Mostly BC (ex-Philips) 4,700uf in parallel, but also Panasonic FA (discontinued, replaced with FC) works very well on my small LM1875 bench amp.
I've also tried really cheap 10,000uf/63v caps (the first test was made with those and a bunch of wires and aligators) with decent results.
EDIT: the bypass and snubber caps are BC MKT 100nf/100v.
Attachments
Yes I did.carlosfm said:
Did you also bypassed the big caps with 100nf film?
I will give a try with the parts you've mentionned.
Thanks.
- Status
- Not open for further replies.