I'm working on a LM3886 amplifier, and I have it working and it sounds mostly great. But when I put my ears up against the speaker, I can hear a little bit of what sounds like faint power supply noise.
I've been messing with the grounding, but I can't seem to nail it.
Right now for testing I have two RCA inputs that are isolated from the chassis (verified with a multimeter). These go to two different PCBs representing the right and left channel. These are connected by two 2-wire twisted pairs. If I don't tie these to ground within the amplifier, it doesn't sound good at all. If I tie them to ground at the power supply caps, it sounds good, but it has the faint hum with my ear up close (this is the current configuration). If I take one channel and ground it instead at the speaker gnd of the appropriate channel the hum goes away and it's totally quiet. If I do the same with the other channel though, the sound gets worse. The RCA cable source is my iPad using a mini audio connector cable that splits into RCA.
Another question is should the INPUT GND be totally separate from the SIGNAL GND/ I read the Taming the LM3886 article, and it seems the consensus is to tie the SIGNAL ground into the speaker negative terminal, but it doesn't really mention what happens to the INPUT GND.
Thanks.
I've been messing with the grounding, but I can't seem to nail it.
Right now for testing I have two RCA inputs that are isolated from the chassis (verified with a multimeter). These go to two different PCBs representing the right and left channel. These are connected by two 2-wire twisted pairs. If I don't tie these to ground within the amplifier, it doesn't sound good at all. If I tie them to ground at the power supply caps, it sounds good, but it has the faint hum with my ear up close (this is the current configuration). If I take one channel and ground it instead at the speaker gnd of the appropriate channel the hum goes away and it's totally quiet. If I do the same with the other channel though, the sound gets worse. The RCA cable source is my iPad using a mini audio connector cable that splits into RCA.
Another question is should the INPUT GND be totally separate from the SIGNAL GND/ I read the Taming the LM3886 article, and it seems the consensus is to tie the SIGNAL ground into the speaker negative terminal, but it doesn't really mention what happens to the INPUT GND.
Thanks.
It could be many things. For start you could disconnect the twisted pairs from the board and put a short link on the inputs. This should answer is it a problem with RCA wiring or somewhere else.
Grounding has the one final rule: If it is silent, you are right. So there are quite some basic rules, but these may need some small "adjustments".
The often heard of "star ground" thing often doesn't work in real life.
If you have an oscilloscope, this is very helpful to watch how hum, better than listening. Of course, have a speaker or load connected while trying.
I connect one amp channel first, which should be dead silent used allone. If one alone hums, something is really wrong.
For the RCA you should isolate them from chassis. Then connect both RCA grounds together and only sent one ground wire back to the amp section. You may use screened wire, but only ground the screens at the amp. So you may end up with two screened wires, connected one side plus a separate ground wire at the RCA inputs.
This way, if you use one ground for both RCA's, the line to the preamp (iPhone) will not form a ground loop inside the amp.
Usually the hum problems start when the second module is connected. First connect the second channel to power and speaker and then search for the best way to route the input ground.
Now, usually you will try to ground the input directly at the amp, which can work or not. Using a 100 Ohm resistor (or two, one to each amp ground) in the ground line can be helpful. Also grounding the input via a 100 Ohm at the power supply cap ground.
Doing a simple sketch of any combination you did can be very helpful. I had amps that had to be connected against the rules to be 100% silent.
While trying, be carefull not to short anything.
Try to move wires just a little, if you can modulate the noise, it may be induced and no ground loop.
You are done if the very faint hiss, you can hear with the ear at the tweeter is louder than any hum.
Do not test hum with an open input. Have it connected to something (preamp) or shorted.
Maybe do some good pictures of your build, sometimes someone can spot problems that way.
The often heard of "star ground" thing often doesn't work in real life.
If you have an oscilloscope, this is very helpful to watch how hum, better than listening. Of course, have a speaker or load connected while trying.
I connect one amp channel first, which should be dead silent used allone. If one alone hums, something is really wrong.
For the RCA you should isolate them from chassis. Then connect both RCA grounds together and only sent one ground wire back to the amp section. You may use screened wire, but only ground the screens at the amp. So you may end up with two screened wires, connected one side plus a separate ground wire at the RCA inputs.
This way, if you use one ground for both RCA's, the line to the preamp (iPhone) will not form a ground loop inside the amp.
Usually the hum problems start when the second module is connected. First connect the second channel to power and speaker and then search for the best way to route the input ground.
Now, usually you will try to ground the input directly at the amp, which can work or not. Using a 100 Ohm resistor (or two, one to each amp ground) in the ground line can be helpful. Also grounding the input via a 100 Ohm at the power supply cap ground.
Doing a simple sketch of any combination you did can be very helpful. I had amps that had to be connected against the rules to be 100% silent.
While trying, be carefull not to short anything.
Try to move wires just a little, if you can modulate the noise, it may be induced and no ground loop.
You are done if the very faint hiss, you can hear with the ear at the tweeter is louder than any hum.
Do not test hum with an open input. Have it connected to something (preamp) or shorted.
Maybe do some good pictures of your build, sometimes someone can spot problems that way.
Last edited:
The IEC mains inlet ground (protective earth) terminal goes to the chassis. The power transformer connects to the power supply. You can then take the ground reference on the output of the power supply to the chassis if you wish. No current should flow in that conductor. Connect V+, GND, V- from the power supply to the amp circuit.
You should have local decoupling by the amp itself. So 470-1000 uF where the power enters the board plus at least 10uF||100nF right next to each LM3886. If you read the Decoupling article of the Taming the LM3886 Series, you'll find that I recommend 22uF||4.7uF. In the simulation I used a 22 uF OSCON, but I find a plain 22 uF electrolytic to work just as well. Similarly, you can walk the 4.7 uF back to 1.0 uF without penalty. That 1.0-4.7 uF must be an X5R, X7R, or X8R ceramic type, though. Also, choose one with a minimum DC working voltage of 50 V. The voltage coefficient of many ceramics is absolutely horrid. While that of the XnR dielectrics isn't pretty either, it's still much better than anything else for that capacitance. Unless you remortgage your house and buy 4.7 uF, 50 V, C0G, I suppose. 🙂
The ground for the speaker output is taken from the local decoupling in the amp circuit. I would make that the ground reference for the Zobel network and LM3886 feedback network as well. All the quiet grounds (so anything on the input of the amp) should connect together and to the speaker (-) terminal. After all, it's the voltage across the speaker terminals we care about so that should be the reference point. It is important that no (or, rather, very little) current flows in that input reference ground.
It's perfectly possible to build an LM3886 amp that's completely free of hum. You may have a very, very slight hiss when you press your ear against the tweeter, but that should be it. That hiss is just the noise of the amp. It usually sits around 30 uV RMS (20 Hz - 20 kHz, A-weighted); 40 uV RMS (20 Hz - 20 kHz, unweighted) in my builds. Those figures include any residual mains hum, by the way.
Tom
You should have local decoupling by the amp itself. So 470-1000 uF where the power enters the board plus at least 10uF||100nF right next to each LM3886. If you read the Decoupling article of the Taming the LM3886 Series, you'll find that I recommend 22uF||4.7uF. In the simulation I used a 22 uF OSCON, but I find a plain 22 uF electrolytic to work just as well. Similarly, you can walk the 4.7 uF back to 1.0 uF without penalty. That 1.0-4.7 uF must be an X5R, X7R, or X8R ceramic type, though. Also, choose one with a minimum DC working voltage of 50 V. The voltage coefficient of many ceramics is absolutely horrid. While that of the XnR dielectrics isn't pretty either, it's still much better than anything else for that capacitance. Unless you remortgage your house and buy 4.7 uF, 50 V, C0G, I suppose. 🙂
The ground for the speaker output is taken from the local decoupling in the amp circuit. I would make that the ground reference for the Zobel network and LM3886 feedback network as well. All the quiet grounds (so anything on the input of the amp) should connect together and to the speaker (-) terminal. After all, it's the voltage across the speaker terminals we care about so that should be the reference point. It is important that no (or, rather, very little) current flows in that input reference ground.
It's perfectly possible to build an LM3886 amp that's completely free of hum. You may have a very, very slight hiss when you press your ear against the tweeter, but that should be it. That hiss is just the noise of the amp. It usually sits around 30 uV RMS (20 Hz - 20 kHz, A-weighted); 40 uV RMS (20 Hz - 20 kHz, unweighted) in my builds. Those figures include any residual mains hum, by the way.
Tom
Instead of the ground wire from each RCA, use a 10R resistor connected to the audio ground.
This will defeat the ground loops.
This will defeat the ground loops.
I think maybe my issue is coming from the fact that I (erroneously) connected signal ground to input ground on the PCB, instead of the load ground on the PCB. It still works fine, but I think I'm hearing some noise in the error signal of the opamp when I put my ear against the speaker.. It's not hum, it sounds more like very mild rectifier noise. I took one channel, from the RCA jack, and grounded it at the speaker terminal of the left channel, and that faint buzz went away. But the buzz on the other channel didn't improve (it got slightly worse), which makes sense I think, since now it's signal ground (right channel) is polluted with the feedback signal on the left channel.
I made a new PCB that's coming that ties signal ground into the load ground like your article says, and leaves input ground separate completely.. I'm hoping that fixes that one issue. As I've modified the grounding many different ways, and it doesn't make a difference (and when I ground that one channel at the speaker terminal, it does go away). So hopefully that's it. It still sounds amazing, but no sense doing this unless it's done right! So i'll try and fix this.
Thanks for the input.
Can you elaborate on this? I'm struggling a bit to understand what's proper here. I have a signal wire from each RCA going to each channel, and it's pair with the ground (outside shield connected) from each RCA jack. How should they be grounded internally from here? Should I run a ground wire from each jack to the input ground? Should I run a small run between RCA jacks, and tie one of them (using only one wire) to the input ground internally? Or should they be left completely ungrounded (assuming that the source is grounded). The LM3886 design has a DC blocking capacitor on each channel, so doesn't that sort of make it ground level agnostic (i.e. it doesn't matter if it's ground is floating relative to the internal one, since the DC gets blocked)?
Thanks.. I used your article when making my schematic, so it's mostly like that. I have a 1000uf||22uf||4.7uf X7R. All caps are 50V or higher in mine.. Right now it's +/-28V as I have both 6 and 8 ohm speakers at home I'm playing with..
I read your articles (thanks! They helped a ton). I struggled a bit with the one on grounding though. I understood how you determined that SIG_GND should tie into LOAD_GND I believe, but I wasn't clear from a practical perspective what that meant in terms of layout. I think I struggled because on one of your LM3886 articles it shows a PCB power supply you made, and I believe it had terminals for INPUT_GND SIG_GND LOAD_GND and PWR_GRND I believe. But in another article your wrote, probably the 'Taming' one, I believe you say SIG_GND should tie into the local channel LOAD_GND. If that's the case, then isn't the SIG_GND on that power supply not needed? I suspect they were written at different times and you modified your approach, but when I read them together I found that a bit confusing.
Last edited:
Never fail to connect the input RCA ground terminal, that would considerably increase noise.
The usual practice is to run a twisted pair from each input RCA to each amplifier input stage.
Then just add a 10R resistor in series with the RCA ground wire, in both channels.
Like this:
Attachments
So, just to rephrase
1) connect the RCA connector ground to the amplifier input ground through a 10R resistor.. This is done twice I take it? Doesn't that create a ground loop though if both RCA connectors share a common ground at the source (since you now have two different paths to ground)?
2) run a twisted pair from each RCA connector to each amplifier (knowing it's grounded to the amplifier input ground at the chassis).
Sound right?
Not like that, just go by the drawing.
The resistor is in series with the twisted pair ground wire (not in parallel).
If the wires are short, the resistor can take the place of the ground wire entirely.
If the wires are longer, you can add each resistor in place of part of each ground wire,
but still in series with the ground wire. Look at the drawing.
So then you're saying don't connect the RCA ground terminals with any internal ground? Just run the RCA shield terminal ground, in series with a 10 ohm resistor, to the amplifier and use that as the input ground?
Yes, that's it.
For example, the ground wire with the resistor, could go to the same point where the feedback loop goes to ground.
But if these are pre-made boards, there should be marked pads for the input ground wires that should be fine to use.
Can you post your schematic, and some photos of the RCA sockets and the amplifier input board to confirm?
Bear in mind that there could be other connections besides the inputs that could be causing hum.
Even how the two large filter capacitors are connected could also cause some hum.
Last edited:
It's my board, so I can change it as I see fit 🙂 Right now it has JST-PH connector for the input, with one wire for the signal and one for the ground. This ground is completely independent from anything, and not connected internally to anything in my amplifier. So my only question was if I need to connect it to my amplifier ground as well, or if the RCA jack ground is enough for the input side of the amplifier.
RCA socket is isolated, and it has a ground lug and the center signal lug.
Your amp has other pins and parts connecting to an audio board ground point.
The input ground wire with resistor should go there. Post your schematic here.
All circuits must have at least one loop. The input ground wire is part of that loop.
The input ground wire with resistor should go there. Post your schematic here.
All circuits must have at least one loop. The input ground wire is part of that loop.
Ok, but that's what I asked above, and I thought you said not to connect it to the amplifier ground. If that's the case, I'll tie the RCA ground also into my low signal ground. Isn't there a loop with the source though? I.e. if I play music using an iPhone via the RCA cables, isn't the ground wire supplied the loop for the audio input?
Yes, that's right.
You don't want an audio loop to be completed by an external ground circuit through the power line.
This is what happens when you don't connect the RCA ground lugs. That will always add noise.
An external source, floating or not, will work as part of the loop, if the interconnection is properly shielded
and not too long.
You don't want an audio loop to be completed by an external ground circuit through the power line.
This is what happens when you don't connect the RCA ground lugs. That will always add noise.
An external source, floating or not, will work as part of the loop, if the interconnection is properly shielded
and not too long.
Ok thanks. Appreciate it, I'll try this when I get the new boards correcting my issue above.
I once had unacceptable hum/noise when the input sockets were grounded on the chassis. Since then, I've always isolated the input sockets and run the input socket ground to the PCB and then from there a ground wire from the PCB quiet earth to the common earth point.
Series resistors in the ground lead would also help.
Series resistors in the ground lead would also help.