I have a question about properly grounding my point-to-point LM3886 amp, which I initially sent to AndrewT on PM, but he asked me to copy it here together with his answer, and I agree, so here it is:
...here's what I applied already (from one of your (AndrewT's) earlier posts):
So the questions that remain are these:
1. Where do I connect the IN-? Do I connect it directly to the Main Audio Ground, or connect it together with the bottom of Ci and R2, and then to Main Audio Ground? (there was a post on this and ground lift, but I can't find it)
2. Where to connect the chip's ground pin (pin 7)? Do I connect it to the Power Ground?
Here's a schematic of what I think is right:
***********************************************************
And here is the answer:
Complete the INPUT circuit.
Signal In and Signal Return MUST form a complete circuit.
This circuit MUST have very low loop area.
The amplifier measures the voltage applied to the INPUT SOCKET, i.e. between Signal Hot and Signal Return.
The amplifier then processes that difference in voltage and sends a scaled copy to the Output.
To enable that processing the NFB MUST also be taken to the Signal Return.
Again this loop area must be minimised in this extra circuit that touches the INPUT circuit.
Finally reference the Input circuit to the Main Audio Ground. You can do this at the PCB input terminals or at the Chassis mounted input terminals or for a stereo amp at the common link joining all the input terminals together.
REMEMBER
every circuit allows current to flow around a loop. That loop ALWAYS comes back to where you started your analysis of the current from.
Every loop must be completed. Often these individual loops need to be referenced to each other. This should be at individual points. You select the points of contact between loops to minimise the interference caused by the link.
...here's what I applied already (from one of your (AndrewT's) earlier posts):
Connect the HF decoupling together.
Connect the MF decoupling together.
Connect these two and call them Power Ground.
Connect the output Zobel to Power Ground.
Connect Power Ground to Main Audio Ground.
Connect Speaker Return to Main Audio Ground.
Connect the MF decoupling together.
Connect these two and call them Power Ground.
Connect the output Zobel to Power Ground.
Connect Power Ground to Main Audio Ground.
Connect Speaker Return to Main Audio Ground.
So the questions that remain are these:
1. Where do I connect the IN-? Do I connect it directly to the Main Audio Ground, or connect it together with the bottom of Ci and R2, and then to Main Audio Ground? (there was a post on this and ground lift, but I can't find it)
2. Where to connect the chip's ground pin (pin 7)? Do I connect it to the Power Ground?
Here's a schematic of what I think is right:
An externally hosted image should be here but it was not working when we last tested it.
***********************************************************
And here is the answer:
Complete the INPUT circuit.
Signal In and Signal Return MUST form a complete circuit.
This circuit MUST have very low loop area.
The amplifier measures the voltage applied to the INPUT SOCKET, i.e. between Signal Hot and Signal Return.
The amplifier then processes that difference in voltage and sends a scaled copy to the Output.
To enable that processing the NFB MUST also be taken to the Signal Return.
Again this loop area must be minimised in this extra circuit that touches the INPUT circuit.
Finally reference the Input circuit to the Main Audio Ground. You can do this at the PCB input terminals or at the Chassis mounted input terminals or for a stereo amp at the common link joining all the input terminals together.
REMEMBER
every circuit allows current to flow around a loop. That loop ALWAYS comes back to where you started your analysis of the current from.
Every loop must be completed. Often these individual loops need to be referenced to each other. This should be at individual points. You select the points of contact between loops to minimise the interference caused by the link.
The schematic in the first post shows how I actually connected everything in the end. I did have a problem with LOTS of hum and buzzing, thought I did something wrong, but it was down to separating the MAG from the Chassis Ground, which was polluted by my PC and who knows what. Now there's virtually no hum in the speakers 
Unfortunately my camera died so I can't post pics, but I'm very proud on the whole P2P adventure - my HF decoupling is the length of 2 small Wima 100nF caps, MF decoupling less than 4cm per side, both connected with a AWG20 thick copper wire only a few mm apart. Soldered the Zobel (length of R and C) to that same wire, and pin7 (chip gnd) was soldered to this Power Ground, and in the end I connected the middle of this wire to the MAG.
I didn't go for symmetry, as it was clear to me that the shortest possible traces could be done if V+/V- and output are kept on the left of the chip (pins 1, 3 and 4), and the input on the right (pins 9 and 10).
Speaker returns are taken directly to MAG, and I soldered the signal returns with a piece of wire and from the middle of it a wire to the MAG. Tried to keep all the wires going to MAG the same length, so only PSU grounding wires are a bit longer.
So, any better ideas for grounding?
Also - should I lift the signal ground with a small value resistor or should that be considered only when humming is a problem?
Unfortunately my camera died so I can't post pics, but I'm very proud on the whole P2P adventure - my HF decoupling is the length of 2 small Wima 100nF caps, MF decoupling less than 4cm per side, both connected with a AWG20 thick copper wire only a few mm apart. Soldered the Zobel (length of R and C) to that same wire, and pin7 (chip gnd) was soldered to this Power Ground, and in the end I connected the middle of this wire to the MAG.
I didn't go for symmetry, as it was clear to me that the shortest possible traces could be done if V+/V- and output are kept on the left of the chip (pins 1, 3 and 4), and the input on the right (pins 9 and 10).
Speaker returns are taken directly to MAG, and I soldered the signal returns with a piece of wire and from the middle of it a wire to the MAG. Tried to keep all the wires going to MAG the same length, so only PSU grounding wires are a bit longer.
So, any better ideas for grounding?
Also - should I lift the signal ground with a small value resistor or should that be considered only when humming is a problem?
If you connect the input ground to the 22k and the feedback ground, you'll apply noise from common-mode currents emanating from your source (DAC, CD player, preamp). To avoid this, give the input ground its own dedicated wire back to MAG. This is likely to induce a little more hum though from stray mag-field so keep the trafo distant.
@Speedskater - there is no circuit board, I soldered it point-to-point. I actually started with a chipamp.com boards, but after lots of changes I had problems with lifted traces, so I went for P2P.
@abraxalito - I might try that just to see if there is change, but right now I have just some slight hum, heard only when I put my ear to the speaker. But, as AndrewT wrote, shouldn't signal return and NFB loops be as short as possible? This way I would add more than 30cm (2x15cm) wire from signal ground > MAG > nfb.
@abraxalito - I might try that just to see if there is change, but right now I have just some slight hum, heard only when I put my ear to the speaker. But, as AndrewT wrote, shouldn't signal return and NFB loops be as short as possible? This way I would add more than 30cm (2x15cm) wire from signal ground > MAG > nfb.
They should be as short as reasonably possible, nevertheless still constrained by other important aspects including not coupling ground noise into sensitive nodes of the circuit. I suggest having a listen - if you don't hear the effects of moving the input ground to a 'safe' place then leave it as is.
@abraxalito - let me get this right, which of these are you saying I should do:
1. connect signal gnd to bottom of R2 and then wire to MAG, also give nfb gnd a separate wire to MAG
OR
2. connect signal gnd directly to MAG, and connect bottom of R2 to nfb and then wire to MAG?
@aparatusonitus - here's my PSU - notice the two separate ground wires going to MAG and no ground wires going to amp
1. connect signal gnd to bottom of R2 and then wire to MAG, also give nfb gnd a separate wire to MAG
OR
2. connect signal gnd directly to MAG, and connect bottom of R2 to nfb and then wire to MAG?
@aparatusonitus - here's my PSU - notice the two separate ground wires going to MAG and no ground wires going to amp
An externally hosted image should be here but it was not working when we last tested it.
Ok, so you are using single PSU to power both channels? If that's the case do it like in the images below ;-)
Edit: Take output (speaker) ground for left channel from power ground point at left channel pcb, do the same for right channel.
Edit: Take output (speaker) ground for left channel from power ground point at left channel pcb, do the same for right channel.
Attachments
Last edited:
I don't like diagrams of posts 9 & 10.
Post9 shows the Zero Volts carrying charging pulses being used as the MAG. No the MAG is NEVER on the PSU Zero Volts.
Post10 shows the power connections split up into 4 single wires. The Zero Volts should be joined and the V+, V- and Zero volts must be close coupled as a power triplet to pass over the MAG on the way to each amp PCB.
Similarly the Signal Grounds are split creating a hum loop as soon as the Source to Amp Interconnects are plugged in.
Post9 shows the Zero Volts carrying charging pulses being used as the MAG. No the MAG is NEVER on the PSU Zero Volts.
Post10 shows the power connections split up into 4 single wires. The Zero Volts should be joined and the V+, V- and Zero volts must be close coupled as a power triplet to pass over the MAG on the way to each amp PCB.
Similarly the Signal Grounds are split creating a hum loop as soon as the Source to Amp Interconnects are plugged in.
1. What is that between the reservoir caps in the first image?
2. Why connect speaker return to power gnd, isn't it better to connect it directly to main ground?
3. Could you explain what those antiparallel diodes do exactly?
I see that this way you make use of twisting the V+/V-/Gnd wires going to amp boards, right now I just twist V+/V- together.
Its looking like your 2 is what I am suggesting. In your pic the signal ground is shaped like a letter 'T' - I'm saying cut the top left connection (IN-) to that 'T' and give that IN- node only its dedicated wire to MAG. The other connections to signal gnd to remain as is.
<edit> On your PSU wiring pic, why those particular two wires to MAG? Seems to me you do want two wires but not those, rather you want a wire from the -ve and a wire from the +ve. As it is you have fairly major charging currents (those for the 2nd cap) running through MAG. None too pretty.
Last edited:
Hmm, I'd have to dig up that post where I first saw those two separate wires, it's somewhere in that huge discussion about PSU quality... but then again, I'm the one usually wrong here
so I'll definitely change that.I can understand removing the 3.3uF, but remove the bypass 100nF and snubbers?! Why? Isn't bypassing large capacitors a proven method, and don't snubbers remove the diode switching noise and help with the psu impedance?
I made some changes to the PSU:
- removed one of the wires going to MAG, connected first reservoir ground to second reservoir ground
- moved the snubbers to the transformer tappings
- removed the 3.3uF caps
- moved the 100nF bypasses directly to the pins of second reservoir caps
Can't say I hear any difference. The slight hum is exactly the same as it was, it is still a bit louder in the left channel (with some very slight buzzing), but that's basicaly irrelevant, can't be heard even at 1m from the speakers.
I still have to try the signal / nfb ground change, and decoupling MAG from chassis ground, so I'll post again.
PS - I have lots of changes going on (P2P, grounding, LDRs, paralleling Simics...) but changing the gain and adding diodes in series with V+ and V- make the greatest positive changes. Here's a link to diodes trick
- removed one of the wires going to MAG, connected first reservoir ground to second reservoir ground
- moved the snubbers to the transformer tappings
- removed the 3.3uF caps
- moved the 100nF bypasses directly to the pins of second reservoir caps
Can't say I hear any difference. The slight hum is exactly the same as it was, it is still a bit louder in the left channel (with some very slight buzzing), but that's basicaly irrelevant, can't be heard even at 1m from the speakers.
I still have to try the signal / nfb ground change, and decoupling MAG from chassis ground, so I'll post again.
PS - I have lots of changes going on (P2P, grounding, LDRs, paralleling Simics...) but changing the gain and adding diodes in series with V+ and V- make the greatest positive changes. Here's a link to diodes trick
It's just a CMC (common mode choke)...in your amp psu you have an RC filter, you can make it a bit better if you place another R (like 0R1/10W in front of first filter capacitor so your PSU become an RCRC filter. Don't put more then 0R1, 'cos there is already dynamic impedance of diode bridge + resistance from transformer secondaries + transformer leakage inductance reactance.
No, in theory at least, all input, feedback and output ground should be referenced to the same point - power ground as close as posible to amp supply pins/decoupling capacitors.
It's a ground separator, protecting you from killing yourself and potential ground loops from the rest of you system (house wireing).
Do exactly as I show you before, with lots of wire twisting of equal length for left/right channel, stay away with wireing from transformer, or place wire perpendicular to transformer radius (if it's toroidal), do not parallel line carring high current with signal lines, make them cross if you have to. If you follow this rules, you will not hear any noise or buzzing from your speaker any more.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.