And Whay Goldmund and other Hi-end companies uses big to small film caps in their psu-s?Like 2.2uf-0.47uf-0.047uf.
It would be much better to put those low values of caps extremely close to the power and ground rail connections to the active output device, which is the ONLY place that matters at all, for the PSU performance.
They would not be able to do any good, at the output device, if they had the parasitic inductance of the power and ground rails between them and the output device.
Also, they MIGHT create an unwanted LC resonance, if they are placed close to any large electrolytic.
So I see no advantage to placing them near the PSU reservoir caps, and a possible disadvantage.
There is a large thread, here, "Paralleling Electrolytic and Film Caps", where it was proven that it is unwise to parallel film caps (or C0G or NP0 ceramic) with larger electrolytic caps (which tend to have lower ESR, i.e. parasitic equivalent series resistance). The small caps' very low ESR enables their C to form an almost-undamped LC resonance with the parasitic inductance of the large electrolytics. You don't want that.
Also note that some people who measure only with their ears might hear the effects of the high-frequency ringing caused by the LC resonance and think that it "sounds better", when in fact it's just a form of distortion.
Last edited:
and some even hail the attribute of "added detail" instead of sorting the problem.
I've built headphone amplifiers in the past using a star ground, separating what needs separating and joining the grounds at a middle, star point. The last one I built, according to some RMAA measurements, has a noise floor as low as my xonar DX's -113dB, and I believe it could go even lower if I had a better sound card to measure with... Apart from a DMM, that's all the measuring gear I have. A scope would be nice, some day.
Anyways, I built a LM3886 amplifier last month, again, using a similar grounding scheme and there's a little hum (yeah, I have the transformer near, twisted all the AC and DC wires) and noise if you have the source disconnected. When it's plugged in there's just a little noise when you get your ears up next to the speakers, that becomes far more noticeable if you get near the tweeters... I don't know if that's as low as the 3886 can go in terms of noise, but after reading this thread I believe I can improve it further by using some better grounding. It already sounds really good to my ears, but I'll improve whatever I can
Let me see if I understood right... the following is how I'd do it after reading this thread.
Let's say I have my power supply already built. A 24v+24v 240VA center tap transformer, fuses, a sturdy 35A bridge rectifier, a good deal of capacitance (10k uF per rail give or take). I make the other leg of the v+ and v- rails as the 0v point, there I connect the center tap.
Moving on to the amplifiers, let's say I have power ground as follows:
connected as tight and as near each other as possible. Having done that, I proceed to take both power grounds from each amp and join them at a new midway point which is, if I understood correctly, main audio ground? Which should have both speaker returns connected to it, right?
Low level signals as the RCA input gnd should be connected to the main audio ground, right?
This main audio ground should then, be referenced to the 0v point in the power supply at some point so a sturdy, thick wire runs from the 0v point to the main audio ground, completing the star. Right? Here's a drawing of the above.
----------------------------------------------------
As a side question, I've seen some more advanced chipamp designs here which separate "noisy" grounds from "quiet" grounds with a 1 ohm resistor or something alike. Is there a need for this in a simple, straightforward LM3886 implementation?
Thanks.
Anyways, I built a LM3886 amplifier last month, again, using a similar grounding scheme and there's a little hum (yeah, I have the transformer near, twisted all the AC and DC wires) and noise if you have the source disconnected. When it's plugged in there's just a little noise when you get your ears up next to the speakers, that becomes far more noticeable if you get near the tweeters... I don't know if that's as low as the 3886 can go in terms of noise, but after reading this thread I believe I can improve it further by using some better grounding. It already sounds really good to my ears, but I'll improve whatever I can
Let me see if I understood right... the following is how I'd do it after reading this thread.
Let's say I have my power supply already built. A 24v+24v 240VA center tap transformer, fuses, a sturdy 35A bridge rectifier, a good deal of capacitance (10k uF per rail give or take). I make the other leg of the v+ and v- rails as the 0v point, there I connect the center tap.
Moving on to the amplifiers, let's say I have power ground as follows:
- LM3886's pin 7
- HF decoupling
- MF decoupling
- output zobel
connected as tight and as near each other as possible. Having done that, I proceed to take both power grounds from each amp and join them at a new midway point which is, if I understood correctly, main audio ground? Which should have both speaker returns connected to it, right?
Low level signals as the RCA input gnd should be connected to the main audio ground, right?
This main audio ground should then, be referenced to the 0v point in the power supply at some point so a sturdy, thick wire runs from the 0v point to the main audio ground, completing the star. Right? Here's a drawing of the above.
----------------------------------------------------
As a side question, I've seen some more advanced chipamp designs here which separate "noisy" grounds from "quiet" grounds with a 1 ohm resistor or something alike. Is there a need for this in a simple, straightforward LM3886 implementation?
Thanks.
Sadly I don't. My house is quite old and has no earth grounding... Over the years we've changed most of the two prong sockets (where both leads are circular and have no defined orientation/live/neutral) for the V shaped ones with the third lead for earth ground, but it's not connected to anything.
Apart from that I'm using a wooden case for my amplifier so there's no chassis to ground... and there's no way for one to touch the insides, only if the case gets open. Nevermind the wood, there's a 80mm computer fan running at 9v off a separate transformer + LM317 regulator module so there's plenty of airflow and nothing gets hot even after hours of loud listening. I understand I ought to add the earth connection anyways, to have the added security if I ever plug this amplifier into a more modern electrical installation, but for the time being I'd like to get the amplifier grounding itself 100% right before tackling the safety earth.
Apart from that I'm using a wooden case for my amplifier so there's no chassis to ground... and there's no way for one to touch the insides, only if the case gets open. Nevermind the wood, there's a 80mm computer fan running at 9v off a separate transformer + LM317 regulator module so there's plenty of airflow and nothing gets hot even after hours of loud listening. I understand I ought to add the earth connection anyways, to have the added security if I ever plug this amplifier into a more modern electrical installation, but for the time being I'd like to get the amplifier grounding itself 100% right before tackling the safety earth.
Last edited:
I've built headphone amplifiers in the past using a star ground, separating what needs separating and joining the grounds at a middle, star point. The last one I built, according to some RMAA measurements, has a noise floor as low as my xonar DX's -113dB, and I believe it could go even lower if I had a better sound card to measure with... Apart from a DMM, that's all the measuring gear I have. A scope would be nice, some day.
Anyways, I built a LM3886 amplifier last month, again, using a similar grounding scheme and there's a little hum (yeah, I have the transformer near, twisted all the AC and DC wires) and noise if you have the source disconnected. When it's plugged in there's just a little noise when you get your ears up next to the speakers, that becomes far more noticeable if you get near the tweeters... I don't know if that's as low as the 3886 can go in terms of noise, but after reading this thread I believe I can improve it further by using some better grounding. It already sounds really good to my ears, but I'll improve whatever I can
Let me see if I understood right... the following is how I'd do it after reading this thread.
Let's say I have my power supply already built. A 24v+24v 240VA center tap transformer, fuses, a sturdy 35A bridge rectifier, a good deal of capacitance (10k uF per rail give or take). I make the other leg of the v+ and v- rails as the 0v point, there I connect the center tap.
Moving on to the amplifiers, let's say I have power ground as follows:
connected as tight and as near each other as possible. Having done that, I proceed to take both power grounds from each amp and join them at a new midway point which is, if I understood correctly, main audio ground? Which should have both speaker returns connected to it, right?
- LM3886's pin 7
- HF decoupling
- MF decoupling
- output zobel
Low level signals as the RCA input gnd should be connected to the main audio ground, right?
This main audio ground should then, be referenced to the 0v point in the power supply at some point so a sturdy, thick wire runs from the 0v point to the main audio ground, completing the star. Right? Here's a drawing of the above.
----------------------------------------------------
As a side question, I've seen some more advanced chipamp designs here which separate "noisy" grounds from "quiet" grounds with a 1 ohm resistor or something alike. Is there a need for this in a simple, straightforward LM3886 implementation?
Thanks.
I would want the main audio ground to be very close to what you named the "0-Volts" point (i.e. very near the output ground of the power supply), to minimize the length of conductor that the input signal ground shares with the power and output ground returns, since their currents all have to go back to the power supply caps' grounds, whereas the input signal ground only uses it as a reference voltage, since its actual ground-return currents flow back to the source through the input jack cable. Also make sure that the MAG is downstream from the last reservoir cap, by at least a small distance, so no charging-pulse-induced voltages will share any length of conductor with any of the ground returns.
Otherwise, i.e. if there is a significant length of conductor from the MAG to the reservoir caps' ground (resulting in a not-insignificant parasitic impedance), then the voltages that are induced across the parasitic impedance of the conductor from MAG to PSU caps ground, by the ground-return currents from the output and power grounds, will appear, at least in part, at the input signal ground reference point, at the amplifier.
All the amplifier "knows" about the input signal is the difference between the voltages at the two ends of a resistor from its input to "ground". If the "ground" end of the resistor has a time-varying voltage (caused by the shared length of conductor from MAG to PSU caps), then that voltage is effectively SUMMED with the input signal voltage. That would be "a bad thing".
Also, assuming that you are using multiple parallel reservoir caps per rail, in the PSU (which you probably should be), you could try having one cap (in each rail) be closest to the rectifier, and have it's own separate ground connection to the center tap. The caps after that one could all share another separate connection all the way to the center tap. I'm not sure how much that might help and have not yet tried it but it "sounds good on paper".
Note, too, that the reservoir caps' ground buss, if there is one, should not have "legs" that reach out to each cap's ground pin. The buss should run directly to each cap ground pin.
Cheers,
Tom
Last edited:
With what load?
With no load, there should be no ripple to speak of.
With a boring load, such as a constant DC current, you can get the textbook ripple. But what would it mean?
Real ripple is caused by music. And it's not shaped like the textbook kind. You can see its derivative (time rate of change) by clicking on the link in my signature block. It's the music itself. It's the capacitor current.
The music is going to pull AMPS of current from the caps. That's where the music comes from. So you are going to have voltage ripple, no matter what, because pulling huge currents out of caps will always make their voltage fall (and pushing huge currents into them will always make their voltage rise). Then along comes a charging pulse. So there will also be some 100 Hz or 120 Hz component.
Most amps don't care too much about rail-voltage fluctuations. The feedback adjusts the commanded resistance of the output transistors, to keep the whole thing linear. Otherwise, the small-signal part of the amp would just blindly set the transistors' resistance to be proportional to the input signal, and since the rail voltage would be changing, the same resistance wouldn't give the same current to the speaker, all the time. That's one definition of non-linearity.
So the only really-big reason to worry very much about the ripple is that it sets the rated maximum output power of the amp (subject to the transformer voltage and the load impedance). (Luckily, we can make the ripple as small as we want, by increasing the reservoir capacitance.) If the PSU's reservoir caps are too small, and that causes the ripple voltage (i.e. the rail voltage) to dip down too far, while the output signal voltage reaches upward at the same time, then there might not be enough room left between them for the voltage across the output stage itself (between power rail and speaker output). That would cause the onset of clipping, (exactly like when a regulator's dropout voltage spec is violated), and cause ripple-shaped chunks to be gouged out of the output signal.
Anyway, sorry to have blathered-on about all of that for so long. I keep trying to get more people to realize that the VOLTAGE of a power supply is not the interesting or important part. The current is where the action is, and is what the power supply needs to be very good at providing!
-----
To answer my question, above, about what the textbook ripple would mean, with a boring constant DC current load: If you made the DC load current be equal to the maximum peak output current, for the peak of a sine wave that would draw the maximum rated RMS output power, THEN the ripple would have some meaning: It would need to dip down no farther than the same sine's peak voltage plus the voltage across the output stage, i.e. between the power rail and the output. Otherwise, you would need to lower your previous estimate of the rated max output power, or, add some reservoir capacitance.
[Yes, many people would use the RMS value of the max-rated-power sine wave, for the constant DC load current, for such an experiment. But that would not guarantee that clipping could not occur, for some other signal shape. The way I see it, the amplifier is capable of producing the peak level, so the reservoir caps should be large-enough to enable it to do so for at least an entire charging-pulse period, or, actually, several in a row. Even a low bass frequency, way below 50 Hz or 60 Hz, even as just a single sine wave, would have relatively long flat peaks, compared to the 100 hz or 120 Hz charging period gaps. So they could look almost like constant DC at the peak level. So use the peak level, not the RMS level, to calculate your reservoir caps, and your caps will never run out of steam for low bass (unless your transformer can't keep up, or the caps' ESR is too high).]
Last edited:
For your hum and noise problem, I would be interested in seeing some actual photos of the build, especially since it sounds like you did most things right. Can you characterize the sound of the hum/noise a little more accurately? Is it mostly 100 Hz? Or 50 Hz? And is there anything else in there with that? And does it instantly stop, when you unplug the power cord or does it instead drop to nothing over some possibly-very-short finite amount of time (as the caps discharge)?
I assume that your input signal and input signal ground pairs are twisted well, all the way from the input jacks to the amplifier or board input. Correct?
I assume that your input signal and input signal ground pairs are twisted well, all the way from the input jacks to the amplifier or board input. Correct?
Looking at your sketch it seems you have separated the PAIRs of wire that form CIRCUITs.
The RCA input ground is actually the RETURN wire of the signal CIRCUIT. The Flow and Return of the signal circuit MUST be close coupled all the way from the input socket to the group of components form the input of your amplifier.
The same applies to your speaker circuit. The Flow and Return MUST be close coupled.
The same applies to the Power Feed to the amplifier. The two Flows and the Return MUST be close coupled This is the Power triplet.
And applies to ALL other Circuits where current flows around a loop.
The RCA input ground is actually the RETURN wire of the signal CIRCUIT. The Flow and Return of the signal circuit MUST be close coupled all the way from the input socket to the group of components form the input of your amplifier.
The same applies to your speaker circuit. The Flow and Return MUST be close coupled.
The same applies to the Power Feed to the amplifier. The two Flows and the Return MUST be close coupled This is the Power triplet.
And applies to ALL other Circuits where current flows around a loop.
I don't agree.
The Points that get referenced to the Main Audio Ground (MAG) are
Signal circuit
Speaker circuit
Power Circuit
PCB Power Ground
PCB Zobel Ground
PCB Decoupling Ground
PSU Zero Volts
I ignore the PSU Zero Volts.
I try to locate the MAG @ the physical centroid of all the other Points.
This usually places the MAG somewhere between the Speaker terminals and the amp PCB output. If possible close to the amp PCB input. There is even an argument that favours the MAG to be located ON the Speaker Return Terminal.
The circulation of currents around all the connected circuits does not require the PSU Zero Volts wire connection to be short.
The transformer+rectifier+smoothing group can be a long distance from the MAG and still all the Audio Circuits perform very well.
Thank you very much you two. Your replies are very enlightening.
Well, my amplifier has this little issue: If there's no source connected (input left floating) I can hear increased noise floor (It could be called background hiss?) and some 50hz hum (220v mains here). I used to have a lot of that at the beginning of the build since the transformer is physically near the PCB (It's all in the same wooden box, which measures 30x20x10cm - 11.81x7.87x3.93 inches).
After twisting the transformer primary leads, secondary leads tightly, twisting the input signal pairs and the output wires per channel tightly too, I'd say 85% of that disappeared. If I increase the volume then I can hear a bit more noise and the 50hz noise increases too, but it's greatly reduced.
If I pull the plug, it all disappears in its entirety and there is just black background, nothing. Music playing during this no power instant sounds a bit clearer than when plugged in. I suppose that's how the circuit would perform if the transformer was far, and maybe the layout was done a bit better? Or I got too spoiled by the noise floor on my headamp, I don't know. Does this suggest a problem?
The zobel doesn't get hot, the LR network (as per the datasheet) after the zobel doesn't either, the 3886s themselves can be left for hours on and they won't get hot, they can play music for a while passively cooled and you can't hear SPiKE or any other protection kicking in. I'm using the T version, mounted to cpu heatsinks directly (wooden box, no problems with grounding) with MX2 high performance thermal paste, and a 80mm fan running at 9v. Ice cold! -12 mV of DC offset on both channels.
All that is basically gone after plugging in a source, for example a DVD player o a phone. In that case the noise floor (hiss?) gets even lower (can be heard if you put your ears right next to the woofer, and at about 5-10cm from the tweeter) and the 50hz hum is, let's say, 95% gone. You can even put the amplifier at 100% volume and you've got to get really near the speakers to hear it. Mind you, when playing music this is easily forgettable and mostly nitpicking, but I believe it can be improved.
I'll get you some pictures of my build today if I can, maybe tomorrow.
Well, Andrew, when I drew that I only took into account the ground wiring. But yes, you are completely right and that's what I'm doing with my amplifier, each signal/output/AC/DC wiring has its return counterpart twisted together, tightly.
In my case I have a preamplifier (Rod Elliott's P97, I used his schematic to make a PCB for my particular amplifier) in the same big PCB which is 20x10cm and the input wires from the RCA connectors are twisted all the way to the preamp input. The speaker wires are twisted, too. Everything that comes, goes and has its pair closely coupled and twisted. AC, DC wires, you name it.
So, in my 20x10cm PCB I currently have:
Right now I'm redesigning the layout of the thing according to your lessons, and it's coming along nicely. I'm adding a 5x8cm PCB, in order to fit what's already there, and some other stuff better. I'll get a 15v+15v 0.5A transformer for the preamplifier so it has its own power supply, to leave the big 240va 24v+24v transformer only for the 3886s. But yes, I've identified some issues that will be corrected in the next revision.
So, I'm gonna experiment with the grounding layout in my existing amplifier and see if that background hiss gets better, but this is going to require a new PCB in order to fix it completely.
I can't thank you enough for your explanations, they're gold!
Well, my amplifier has this little issue: If there's no source connected (input left floating) I can hear increased noise floor (It could be called background hiss?) and some 50hz hum (220v mains here). I used to have a lot of that at the beginning of the build since the transformer is physically near the PCB (It's all in the same wooden box, which measures 30x20x10cm - 11.81x7.87x3.93 inches).
After twisting the transformer primary leads, secondary leads tightly, twisting the input signal pairs and the output wires per channel tightly too, I'd say 85% of that disappeared. If I increase the volume then I can hear a bit more noise and the 50hz noise increases too, but it's greatly reduced.
If I pull the plug, it all disappears in its entirety and there is just black background, nothing. Music playing during this no power instant sounds a bit clearer than when plugged in. I suppose that's how the circuit would perform if the transformer was far, and maybe the layout was done a bit better? Or I got too spoiled by the noise floor on my headamp, I don't know. Does this suggest a problem?
The zobel doesn't get hot, the LR network (as per the datasheet) after the zobel doesn't either, the 3886s themselves can be left for hours on and they won't get hot, they can play music for a while passively cooled and you can't hear SPiKE or any other protection kicking in. I'm using the T version, mounted to cpu heatsinks directly (wooden box, no problems with grounding) with MX2 high performance thermal paste, and a 80mm fan running at 9v. Ice cold! -12 mV of DC offset on both channels.
All that is basically gone after plugging in a source, for example a DVD player o a phone. In that case the noise floor (hiss?) gets even lower (can be heard if you put your ears right next to the woofer, and at about 5-10cm from the tweeter) and the 50hz hum is, let's say, 95% gone. You can even put the amplifier at 100% volume and you've got to get really near the speakers to hear it. Mind you, when playing music this is easily forgettable and mostly nitpicking, but I believe it can be improved.
I'll get you some pictures of my build today if I can, maybe tomorrow.
Well, Andrew, when I drew that I only took into account the ground wiring. But yes, you are completely right and that's what I'm doing with my amplifier, each signal/output/AC/DC wiring has its return counterpart twisted together, tightly.
In my case I have a preamplifier (Rod Elliott's P97, I used his schematic to make a PCB for my particular amplifier) in the same big PCB which is 20x10cm and the input wires from the RCA connectors are twisted all the way to the preamp input. The speaker wires are twisted, too. Everything that comes, goes and has its pair closely coupled and twisted. AC, DC wires, you name it.
So, in my 20x10cm PCB I currently have:
- Power supply
- Regulated power supply for the preamp (LM317+LM337) running off the big power supply
- preamp
- 2x apex's LM3886 boards
Right now I'm redesigning the layout of the thing according to your lessons, and it's coming along nicely. I'm adding a 5x8cm PCB, in order to fit what's already there, and some other stuff better. I'll get a 15v+15v 0.5A transformer for the preamplifier so it has its own power supply, to leave the big 240va 24v+24v transformer only for the 3886s. But yes, I've identified some issues that will be corrected in the next revision.
So, I'm gonna experiment with the grounding layout in my existing amplifier and see if that background hiss gets better, but this is going to require a new PCB in order to fix it completely.
I can't thank you enough for your explanations, they're gold!
Last edited:
Huh, seems I can't edit my message anymore. There is something wrong there and I want to correct it.
When I unplug the amplifier the background noise is still there, up until the capacitors' juice runs out. My memory doesn't work that well without the daily dose of coffe, it seems. Sorry.
I've done a few changes to the existing grounding layout and it seems to be better. I'm doing the listening tests with something sensitive like headphones, later I'll hook it up to the speakers and see how much better it actually is.
When I unplug the amplifier the background noise is still there, up until the capacitors' juice runs out. My memory doesn't work that well without the daily dose of coffe, it seems. Sorry.
I've done a few changes to the existing grounding layout and it seems to be better. I'm doing the listening tests with something sensitive like headphones, later I'll hook it up to the speakers and see how much better it actually is.
Last edited:
Oh, silly me. I tried disconnecting the preamp from one of the amplifiers, the nearest to the transformer. Surprisingly there's zero hum and background noise is almost silent, listening to it with my headphones. Plugging in a source, completely silent, black background and music is clearer... Plugging in a speaker, being much less sensitive, dead silent. And I thought it was alright...
My approach to having everything in the same PCB has bit me, it seems. The preamp isn't getting along nicely with the rest of the parts, or it doesn't like sharing the same power supply with the amplifier. When I get its transformer and make its dedicated power supply I suppose it'll get better, but I'll have to review the circuit, make some improvements there, too.
Well, it's been quite the learning experience, apart from the little rocks in the road. My grounding scheme for the next and final PCB is better, thanks to you and this wonderful thread. I'd buy you guys a beer or two if I could!
My approach to having everything in the same PCB has bit me, it seems. The preamp isn't getting along nicely with the rest of the parts, or it doesn't like sharing the same power supply with the amplifier. When I get its transformer and make its dedicated power supply I suppose it'll get better, but I'll have to review the circuit, make some improvements there, too.
Well, it's been quite the learning experience, apart from the little rocks in the road. My grounding scheme for the next and final PCB is better, thanks to you and this wonderful thread. I'd buy you guys a beer or two if I could!
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.