Anatech,
spell it out.
For reference here are my recommendations.
8ohm speakers
F-1dB=4Hz to 80kHz
Input filters 2Hz and 160kHz
NFB filter 1.4Hz
PSU filter 1Hz
This results in 20mF of smoothing.
If you can tolerate F-1dB of 8Hz to 80kHz then the smoothing can be reduced to 10mF.
Similarly, tolerate F-1dB of 17Hz and smoothing comes down to 4700uF.
Where do you the rest of our Forum Members stand?
spell it out.
For reference here are my recommendations.
8ohm speakers
F-1dB=4Hz to 80kHz
Input filters 2Hz and 160kHz
NFB filter 1.4Hz
PSU filter 1Hz
This results in 20mF of smoothing.
If you can tolerate F-1dB of 8Hz to 80kHz then the smoothing can be reduced to 10mF.
Similarly, tolerate F-1dB of 17Hz and smoothing comes down to 4700uF.
Where do you the rest of our Forum Members stand?
AndrewT said:
Hi Andrew,
I experimented with a pair of LM3886 configured approximately the same as the Audio Sector chip amps. I used a single 2200uF per rail, in place of the on-board rail decoupling.
Like I said said above, they sounded fine at low volume. I can't remark on high or midrange clarity or detail, improvement wise, because I didn't hear any improvement.
I did hear, and there isn't any doubt in my mind about it, that the bass production at higher volumes was lagging behind the amps with "standard" PS capacitance. Standard, for me, is 3-4700uF caps per rail - total 14100uF.
The speakers they were driving were nominal 8ohm impedance, middle efficiency 2-ways.
Gee, no reason to get short with me on this!
Hi Andrew,
So here goes ....
As the capacitance becomes larger, the conduction angle on your rectifiers becomes smaller. This means that all the energy you need is delivered in a shorter length of time. So, what does that mean? If you have much higher pulses of current, your IR losses in your transformer become greater. The magnetic field is more intense and the core losses become greater. Does this sound reasonable so far? The core heating may become more (it will, but maybe not noticeable), I also wonder if the core can saturate during the high peaks. Hopefully not at idle! Your peak currents through your rectifiers will be much higher, and not by a little bit either. This could cause a failure in your rectifiers when you are using what you thought were ratings high enough. Surprise! Running high power from the amplifier will make all these things worse. I suspect many diode failures have gone unnoticed over the last 20 years or so.
On to noise. The shorter conduction time on the diodes will excite the lead inductances more, and at higher frequencies. The transformer inductance will not help to damp these at all. You will see more ringing on your DC lines. Using fast diodes make this situation worse (not better as was thought). Three terminal regulators have poor HF performance (more HF noise passes through). Add to that the fact that the higher frequencies radiate more easily and you can see your noise problems getting worse still. Your larger capacitors are less effective at higher frequencies and 4,700 uF are large capacitors. You will need to mix in some smaller ones and damp them with low value resistors. This is called a snubber.
From John ...
Finally, I hate quoting myself, but I want to make a point ....
-Chris
Hi Andrew,
We have talked about this before. Also, I indicated that we do not agree, but at no time did I say that you are wrong.
So here goes ....
I was referring to bulk capacitors, not arrays of smaller ones. The array of smaller capacitors is a better way to go as far as I'm concerned. Also, at no point did I suggest using too low a value of capacitance. If you were to read the rest of my posts in this thread, you would see this is what I was talking about.
As the capacitance becomes larger, the conduction angle on your rectifiers becomes smaller. This means that all the energy you need is delivered in a shorter length of time. So, what does that mean? If you have much higher pulses of current, your IR losses in your transformer become greater. The magnetic field is more intense and the core losses become greater. Does this sound reasonable so far? The core heating may become more (it will, but maybe not noticeable), I also wonder if the core can saturate during the high peaks. Hopefully not at idle! Your peak currents through your rectifiers will be much higher, and not by a little bit either. This could cause a failure in your rectifiers when you are using what you thought were ratings high enough. Surprise! Running high power from the amplifier will make all these things worse. I suspect many diode failures have gone unnoticed over the last 20 years or so.
On to noise. The shorter conduction time on the diodes will excite the lead inductances more, and at higher frequencies. The transformer inductance will not help to damp these at all. You will see more ringing on your DC lines. Using fast diodes make this situation worse (not better as was thought). Three terminal regulators have poor HF performance (more HF noise passes through). Add to that the fact that the higher frequencies radiate more easily and you can see your noise problems getting worse still. Your larger capacitors are less effective at higher frequencies and 4,700 uF are large capacitors. You will need to mix in some smaller ones and damp them with low value resistors. This is called a snubber.
From John ...
My reply ...
Finally, I hate quoting myself, but I want to make a point ....
Seems like I recognize that the answer to this question isn't fully understood yet. If anything, I avoided attacking you.
-Chris
Thanks Juergen but the whole point is that Im re-designing the PSU. I could upgrade the caps and diodes but I want to maximise the performance of hte LM3886, fortunately for me I can read a schematic and I can solder. I allready have the whole thing figured out in terms of wiring etc. Its just getting to the bottom of this Capacitance issue thats the problem. There are PCB's available for capacitor banks and Peter Daniel sells a rectifier PCB. This shouldnt be a problem as once the stock caps and diodes are out of the way there is space to mount these PCB's on short standoffs. To convert the amp to dual mono will involve cutting a couple of traces on the PCB and soldering in the output of the PSU accordingly. Its not that difficult. Its possible I may need to remove part of the PCB to accomodate the larger trafo but having looked at the traces I will be able to jumper these without a problem as they are wide and run straight paralel to the side of the heatsinks. I allready have some 1mm gold plated silver wire for this. (these traces go from the speaker A-B switch to the speaker terminals by the way).
Hi Mike,
I would suggest leaving this a stereo amplifier for the time being. Once you have it running and debugged, then play with the grounding issue. That way you will more easily be able to get things working if you have problems. Never do everything at one time in an "upgrade". If you break something, it will be easier to find out what you did.
-Chris
Why do you want to do that? You would probably be further ahead leaving this a stereo amplifier. Grounding may become an issue if you mess with it.
I would suggest leaving this a stereo amplifier for the time being. Once you have it running and debugged, then play with the grounding issue. That way you will more easily be able to get things working if you have problems. Never do everything at one time in an "upgrade". If you break something, it will be easier to find out what you did.
-Chris
Hi mike,
Certainly (or hopefully?) the LM3886 does its job, but the overall layout doesn't look best.
See placement of output inductors, signal routing, supply routing and what else...
So this amplfier has limits and compromises and even if one tries a redesign, cutting traces, rewiring and so on, there would be limits and a uncertain outcome.
I really would suggest to change certain parts or values only.
For example, the 3886 decoupling elyts look very small.
A higher value maybe 220u or larger should fit in the pcb.
Regards
If I look at the pictures, this doesn't seems possible.
Certainly (or hopefully?) the LM3886 does its job, but the overall layout doesn't look best.
See placement of output inductors, signal routing, supply routing and what else...
So this amplfier has limits and compromises and even if one tries a redesign, cutting traces, rewiring and so on, there would be limits and a uncertain outcome.
I really would suggest to change certain parts or values only.
For example, the 3886 decoupling elyts look very small.
A higher value maybe 220u or larger should fit in the pcb.
Regards
Here is the existing PSU:
Local caps on the chip are 100uF 50v.
Those who now have the service manual should be able to see what I mean regards to converting to dual mono.:
The new 300VA trafo is 110 dia and 60mm high. Its a tight squeeze but it might even fit without any messing with the PCB. Next time I take the lid off I will check.
If I do need to make a little room then its only the 2 traces nearest the eadge of the board that would need jumper wires. No ground traces will be affected.
For the supply to the chips follow the trace that starts at LK 09,10. You will notice a split off as the traces run down the left side of the sinks to the left hand channel. The trace would be cut after the split off and one channel is connected where LK09,10 is and the other is connected at the cut in the PCB trace.
The new PSU modules would take their input directly from the trafo. The Mains to the trafo will go via the 2 main fuses of course.
The PSU grounds (at output) to one of the vacant Supply cap traces or any other suitable point. The trafo leads would be shielded with copper braid (covered with heatshrink) and the shield connected to chassis earth.
The new trafo for the pre sections would be mounted vertically on hte side or rear panel and connect into the PCB where the original would, J7, 9, 10.
This is only worked out in my head for now and I need to give it more thought regarding the grounding. Im pretty sure Im on the right lines though. Right?
Local caps on the chip are 100uF 50v.
Those who now have the service manual should be able to see what I mean regards to converting to dual mono.:
The new 300VA trafo is 110 dia and 60mm high. Its a tight squeeze but it might even fit without any messing with the PCB. Next time I take the lid off I will check.
If I do need to make a little room then its only the 2 traces nearest the eadge of the board that would need jumper wires. No ground traces will be affected.
For the supply to the chips follow the trace that starts at LK 09,10. You will notice a split off as the traces run down the left side of the sinks to the left hand channel. The trace would be cut after the split off and one channel is connected where LK09,10 is and the other is connected at the cut in the PCB trace.
The new PSU modules would take their input directly from the trafo. The Mains to the trafo will go via the 2 main fuses of course.
The PSU grounds (at output) to one of the vacant Supply cap traces or any other suitable point. The trafo leads would be shielded with copper braid (covered with heatshrink) and the shield connected to chassis earth.
The new trafo for the pre sections would be mounted vertically on hte side or rear panel and connect into the PCB where the original would, J7, 9, 10.
This is only worked out in my head for now and I need to give it more thought regarding the grounding. Im pretty sure Im on the right lines though. Right?
Hi Mike,
I still strongly suggest you do the upgrades first with the original power supply. Once that is running, have a listen. After that you can try the power supply change.
In this way, you stage your work to make things easier to troubleshoot. You also can hear first hand what changes bring what sound differences.
Thank you for the manual. I'll have a read.
-Chris
I still strongly suggest you do the upgrades first with the original power supply. Once that is running, have a listen. After that you can try the power supply change.
In this way, you stage your work to make things easier to troubleshoot. You also can hear first hand what changes bring what sound differences.
Thank you for the manual. I'll have a read.
-Chris
Re: Gee, no reason to get short with me on this!
Yes 4700uF are not small capacitors.
Extra caps? What value?
I do not take our difference of opinion as an attack. I am NOT attacking you.
But, for the education of our readers, what values do you suggest for a 60W into 8ohm amplifier that has F-1dB ~4hz and sounds good?
We need you to spell it out so that we can see how far apart we are.
Is there any difference in the smoothing requirement if the spec is changed to 60W into 4ohms?
And again 200W into 4ohms?
Yes your comments are very reasonable.anatech said:
Yes 4700uF are not small capacitors.
Extra caps? What value?
I do not take our difference of opinion as an attack. I am NOT attacking you.
But, for the education of our readers, what values do you suggest for a 60W into 8ohm amplifier that has F-1dB ~4hz and sounds good?
We need you to spell it out so that we can see how far apart we are.
Is there any difference in the smoothing requirement if the spec is changed to 60W into 4ohms?
And again 200W into 4ohms?
anatech said:
Good advice.
I was planning to do it this way. I need to de-bug my preamp mods firstly. After that I need to check for DC offset and determine if I really need 4 caps per channel in the signal path. With the lm4562 in place I suspect I will be ok with just one cap per channel infront of the power stages. Then I'll try the Peter Daniel config on my poweramps with a pair of BG n's for local decoupling and no NFB cap. I'll either use the same Ricken resistors as he likes of try some Kiwames. If all this proves sucessfull then I'll do the PSU mod with new trafos etc.
Mike
I'd say the amount of capacitance you have in there is proportional to the power output of the amp. I really dont see the point of having say 22,000uF of supply capacitance when your output is only 40W into 8 ohms.
I think that for an LM3886, 4700uF x2 per rail, or at the most 6800uF x2 is the best approach. Using two such capacitors minimises the ESR effect.
http://www.tnt-audio.com/clinica/ssps1_e.html is a good article on this, although theres a certain amount of "audiophileness" in there that I don't agee with. Overall though it's worth a read.
I think that for an LM3886, 4700uF x2 per rail, or at the most 6800uF x2 is the best approach. Using two such capacitors minimises the ESR effect.
http://www.tnt-audio.com/clinica/ssps1_e.html is a good article on this, although theres a certain amount of "audiophileness" in there that I don't agee with. Overall though it's worth a read.
Hi.
So your advising roughly 9'400uF to 13'600uF per rail? Based on that If I wanted to go for a bank of caps I could use say:
2 x 4700uF
2 x 470uF
1 x 47uF
Total = 10387uF per rail
This wouldnt take up too much space and could be mounted on breadboard. I could even add a couple of Polyprop bypass caps around 0.47uF. All in the total capacitance is close to the current set up. To my mind this is good as it's close to what the engineers settled on, but better for mid as its an array of smaller caps.
Theres the added factor of the mods planned at the chips though. If I follow Peter Daniels approach of no cap in the NFB etc, will this have an effect on the PSU requirements?
Mike.
So your advising roughly 9'400uF to 13'600uF per rail? Based on that If I wanted to go for a bank of caps I could use say:
2 x 4700uF
2 x 470uF
1 x 47uF
Total = 10387uF per rail
This wouldnt take up too much space and could be mounted on breadboard. I could even add a couple of Polyprop bypass caps around 0.47uF. All in the total capacitance is close to the current set up. To my mind this is good as it's close to what the engineers settled on, but better for mid as its an array of smaller caps.
Theres the added factor of the mods planned at the chips though. If I follow Peter Daniels approach of no cap in the NFB etc, will this have an effect on the PSU requirements?
Mike.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.