LM3886 PCB

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi guys.

Plz kindly let me know your ideas about the Lm3886 PCB which I have designed.

Thanks alot.
 

Attachments

  • LM3886 pcb.png
    LM3886 pcb.png
    37.7 KB · Views: 463
Where is the Zobel network for the output?

Input ground area should be separated from all the rest. Input ground and speaker/zobel/decoupling ground returns should NOT share the same conductor back to the power supply (through the J1 connector)! The huge dynamic speaker return and decoupling currents will induce relatively-large and nasty-looking voltages across the distributed inductance and resistance of the ground return conductor. If that same conductor also sets the ground reference voltage for the signal input section, then the "ground" voltage at the ground end of the input resistor would be "bouncing", and a good portion of those induced voltages would be arithmetically summed with the input signal voltage! Not only would it sound worse, it is also another feedback path and might make the chip oscillate, which might make it melt, and would at least further-degrade the sound quality.

Input connectors shown will have wires that should be tightly twisted together all the way to the panel jack, which should be insulated from the chassis. Another wire should run from signal input gnd on pcb to the star ground in the power supply (stub off of trace between main filter caps in power supply). So one more wire connector might be needed. (Alternatively, connect the two panel jack gnds with a wire and run a single wire from between the jacks to the star gnd in the power supply, and then don't run one from the pcb input gnd to the star gnd.)

C740 and C742 must be connected RIGHT AT the power pins, and directly to the ground pin of the chip. Sorry but they are worse than useless, as is, since the current has to flow all the way past pin 2 of J1 and all the way back around half the board to chip pin 7! That might significantly degrade the transient response (e.g. fast attacks). They will probably have to be soldered to the pins directly, underneath the board. Or, you could use some surface mount ones, or other small ones, that would fit in between/around/under the pins, so you could solder them on the top side of the board.

C740 and C742 should be X7R ceramic caps, by the way, not film caps.

Similarly, the gnd pins of the large capacitors need to be connected as directly as possible to pin 7 (load gnd) of the chip. At the very least, the trace from pin 1 of J1 should go to the + of C8888 but THEN should go back and go between the pins of C7777, so that both of the caps' ground pins have a much shorter path to pin 7 of the chip, instead of having to make the current go back past J1 and then all the way around half the board to get to pin 7 of the chip. At any rate, you'll also want to minimize the distance from both pins of both capacitors, to the chip power and ground pins, repectively.

If possible, I would use smaller electrolytics. If you feel that you need 1000 uF, I would parallel as many smaller values as I could, to add up to the value desired. Paralleling smaller values would reduce the total ESR and ESL, potentially giving significantly-better decoupling performance than a single larger cap. And, maybe mainly, you could get more of them closer to the chip's power and ground pins.

You have no input low-pass RF filter! Add a small 220 pF film cap between chip pins 9 and 10. It will probably have to be soldered directly to the pins. (Note: When soldering directly to a pin of any semiconductor device, put a rubber band around the handles of a pair of long-nose pliers and clamp their tip onto the pin, between the chip and where you're going to solder, so they will carry most of the heat away before it goes into the chip.)

I would think very hard about possibly using a double-sided PCB, and leaving the top side as a ground plane (with a separate section for the input area). You would have to add a few stubs on the bottom layer, to get out from under caps and connectors that needed to be soldered to the top (gnd) layer, and solder short pieces of bare wire through the pads on the stubs, to connect the top layer to the stub on the bottom layer (like a "via"). But it would make the layout MUCH easier, once you got accustomed to it, and it would also perform a lot better. (You could possibly also mount radial caps lying on their sides, so you could still solder them directly to the top layer.)
 
Last edited:

6L6

Moderator
Joined 2010
Paid Member
If the heatsink contact area of the LM's package is slightly off the edge of the board, you can attach it to a flat chassis, to be used as a heatsink. Or it can be attached to a big flat-backed heatsink. As you have it now, the PCB must be under the lower edge of the heatsink, greatly limiting the mounting potions.
 
The main thing I would do is have the speaker ground return (J3 Pin 2) not go there, but rather go into the square area at near the ground connections of of C7777 and C8888. That's where the big current goes, from the power pins 1, 5 and 4 and out pin 3 and through the speaker, and it needs to return directly to the connection between the main filter capacitors that supply the current.

Actually, pin 7 is only active for mute (see the equivalent schematic, page 7 of the data sheet), very little current goes through it, it's not even signal current, and is fine in the input section's ground area. No output current goes to it. It's not a "load ground."

C740 and C742 should indeed be moved closer to the pins they connect to.

I'd also add spaces for the other "optional" components shown on Fig. 2, page 6 of the datasheet: the 220pF, the 2.7 ohm and 0.1uF from pin 3 to ground (the zobel network gootee mentioned and the ground area the 0.1uF needs to go to is in that square area to the left, near the grounds of C740 and C742), and the 0.7uH and 10 ohm between pin 3 and the speaker connection. Regrettably, Fig. 2 is for single-supply operation and has the extra bias resistors on the input and capacitor on the output that are not used with + and - power supplies, and that may make this example circuit a bit more confusing than it should be. These "optional" components are Good To Have and are recommended regardless of whether a dual or single power source is used.

And there's something to be said for single-digit parts designators, especially when there are fewer than ten parts of each type on the board. At least the connectors are J1, J2 and J3. How did you come up with, for example, C740 and C742? Maybe as a compromise, the two 100nF caps can be C11 and C12, and the 1000uF caps can be C21 and C22, so they are "numerically grouped" by function. The other two can then be C1 and C2.
 
...........Actually, pin 7 is only active for mute (see the equivalent schematic, page 7 of the data sheet), very little current goes through it, it's not even signal current, and is fine in the input section's ground area.
this is disputed in another report.
The investigator also confirmed that PIN7 did not have audio signals on it and further he described the PIN7 currents as detrimental to the signal ground.
 
benb,

Thanks for setting me straight on the function of Pin 7!

Tom
Yeah, well, maybe not so fast there, here's a poster I follow closely:
this is disputed in another report.
The investigator also confirmed that PIN7 did not have audio signals on it and further he described the PIN7 currents as detrimental to the signal ground.
Well that's interesting. Have a source and/or explanation for that? If Pin 7 doesn't have audio signal, I'd think that means it's DC only, and wouldn't have any effect on the input ground.

Or do you mean it DID have audio signals on it?

Regardless, I can't imagine it would be as significant as the speaker return current in the OP's layout.
Hi Guys,

Thanks you very much for all your reply.

Sorry can you plz let me know what you are meaning by the separation of the speaker ground from the other sections of the circuit? A picture would be very ok plz
THe picture is the PCB layout. If you look at all these traces as resistors (which they are, they're just low value resistors - there's also inductance involved, but I'll ignore that to simplify things), you can more easily see where there is a possible problem. You can visualize this better if you magine the traces along the very top and very bottom of the board are resistors. The output current from pin 3 of the LM3886 goes through J3 Pin 1 to the speaker, and then back through J3 Pin 2 to ground. This ground has significant resistance going all the way around the PCB and back to pin 2 of J1. This in itself isn't so bad, except that the ground connection to C5555 and the ground of the input connector, J2, are nearby, and essentially connected to the speaker return, which then changes the voltage at this "ground" with the current through the speaker.

Moving the speaker's ground connection to the inside square ground area where the four capacitors are grounded is a more direct path back to the source of the speaker's current (the + and - power connections at J1 and supported by the four capacitors in that area), and will not affect the ground connections on the input.

Always trace the currents, especially the BIG currents.
 
The equivalent schematic shows pin7 connected to four components.

One provides the ground reference for internal feedback/compensation, which makes it audio and compels its connection to audio ground. Being a transistor's base current limited by a 10k resistor, whatever flows there should not be much of an issue.

The second is the reference for the mute circuit, which should be DC as long as the amp is unmuted and should not have bad effects due to its linear character and low signal level.

Third and fourth are protective diodes that should only pass current when too high voltage is present at the output, e.g. due to back EMF. Current will only flow, when too much reactive energy is present, and if that situation ocurrs, having a clean audio ground or not won't make much of a difference.

I'd connect pin7 to audio ground and rather worry about the missing DC blocking cap, missing Thiele network and the limited size of C555.
 
Here's a reference (seen in another chipamp thread) to the Pin 7, or Ground pin polluting the input/audio/"signal" ground, with further down a statement that the LM4780 (which the article is about) is just a dual LM3886 and has two LM3886 chips in the package:

LM4780 Dual 60W Audio Power Amplifier

I note there's a 2.7 ohm resistor between the input ground and the output-and-power ground, and the chip's ground pin is shown as "better" going to the output-and-power ground. That is further explained in this discussion:

LM4780 Audio Power Amplifier - Glenn's Design Log

I tend to agree with pacificblue on this, on the other hand, looking again at the layout in the OP, there's no cost in connecting Pin 7 directly to the "power" ground through a PCB trace or (as looks easier on this single-sided layout) a jumper wire.
 
Those two links show the same message about the GND pin of the chipamps.
I saw a similar report by a different author.
He went into a bit more detail of the current variations (noise) that came out of that pin. That's what convinced him and me that GND PIN must not be connected to Signal Return (Signal Ground).
There is no harm if the GND PIN is connected to the Main Audio Ground. That's where all the other grounds get their reference from. That difference in tapping point is shown as "OK" and "better" in the diagram
 
According to the links in post #12 the ground pin is not a signal reference, but the equivalent schematic shows the opposite to be true. The current draw of 3 mA does not coincide with what must be expected to flow into the bases of two small-signal transistors that only drive mAs of their own. The assumption that modulation currents flow and interfere with the amplifier's sound quality is not proven in any reproduceable way on that homepage.

Maybe you can come up with that more detailed report from the different author with hopefully more convincing measurements.

Until then I can only advise to follow proven practice and use the same reference for the internal and external feedback.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.