PCB Mistakes/Optimization(?)

[OmarAvelar]

I'll start off by saying that I have been reading this forum for some months and I must say it's pretty nice, and I have been learning a lot.

I took the plunge and decided to build my own amplifier based off the LM3886. The schematic was built from bits from here and there.




Well...what am I really worried about is the PCB (after all, it's my first try) and I would like to ask if some of you guys could give me some pointers and check the layout looking for mistakes or improvements


Pin configuration of the opamp on board is:
TOP ROW(2,4,6,8,10)
BOTTOM ROW1,3,5,7,9,11)


P.S: Do i need the real Eagle version to use extra libraries, like the one for the LM3886?

[markiemrboo]

Originally posted by OmarAvelar
I'll start off by saying that I have been reading this forum for some months and I must say it's pretty nice, and I have been learning a lot.

I took the plunge and decided to build my own amplifier based off the LM3886. The schematic was built from bits from here and there.




Well...what am I really worried about is the PCB (after all, it's my first try) and I would like to ask if some of you guys could give me some pointers and check the layout looking for mistakes or improvements


Pin configuration of the opamp on board is:
TOP ROW(2,4,6,8,10)
BOTTOM ROW1,3,5,7,9,11)


P.S: Do i need the real Eagle version to use extra libraries, like the one for the LM3886?

Hi,

A couple of the simpler things that I have noticed:

Pin 1 and 5 should both be connected to the +ve rail.

Pin 8 (mute) should at minimum, I believe, be connected to the -ve rail via a series resistor. I use 15k.

The 22uF capacitor model looks very small. If you're intending to use a non-electrolytic here, as the model used might suggest, it might be worth finding the proper dimensions and using a more suitable model. Then again, perhaps you can get high quality, non-electrolytic caps of that large of a value in a package that small?

From what I gather it would be wise to move the power supply capacitors closer to the chips pins. It is also usually suggested to bypass them with smaller 0.1uF (100nF) caps as close to the pins as possible.

You seem to have missed out the feedback resistor, 22k. It should be going before the 680R, and should just go from that pin (pin 9) straight to the output pin (pin 3).

The input cap of 4.7uF, you have confused with the feedback cap (which may very well not be necessary) of 4.7pF. Huge difference in every possible way The 4.7uF cap should use a package a lot bigger. If you're using the 4.7pF feedback cap, this should be in parallel with the feedback resistor. i.e from pin 9 straight to 3.

Grounding arrangements also look like they would probably need to be changed. You have the low level grounds (input stuff) connected to the power ground. It would probably be wise to keep these separate.

Not much of a concern right now, but I would say it seems like you have spaced everything out quite a bit. Could be considered a good thing, but I am sure you could compact it down a fair amount if you desire. Get it all correct first though, of course

I use the LM3886 model / symbol / whatever you call it with the free version just fine. I can't remember where I got it from now though.

[impsick]

what a nice guy, taken the time to review your layout!

yeah just read Marks reply a couple time through you'll get it. And i wouold suggest just using the National schematic for the 3886 on there site. Its probably a bit more simplified. Looks like you might be mixing up the 3875, and the 3886. They are different chips so stick to there schematics.

As far as eagle libraries, go to the site and download the ones you need. then you just extract and put in the program file. Your fine with the lite edition
oh there is a download lbr file at eagle called "lmopamp". that has what you need

[OmarAvelar]

Originally posted by markiemrboo

Pin 1 and 5 should both be connected to the +ve rail.

I read that I need to use a jumper If i want to wire them together, because apparently it will cause some kind current inside the package, but I didnt quite understand so I left it like that. Do you happen to know anything about this?

Originally posted by markiemrboo

Pin 8 (mute) should at minimum, I believe, be connected to the -ve rail via a series resistor. I use 15k.

lol, thanks, but what happens if I put a different value resistor?

Originally posted by markiemrboo

You seem to have missed out the feedback resistor, 22k. It should be going before the 680R, and should just go from that pin (pin 9) straight to the output pin (pin 3).

I should have mentioned I was planning on soldering it from the back of the board

Originally posted by markiemrboo
The input cap of 4.7uF, you have confused with the feedback cap (which may very well not be necessary) of 4.7pF. Huge difference in every possible way The 4.7uF cap should use a package a lot bigger. If you're using the 4.7pF feedback cap, this should be in parallel with the feedback resistor. i.e from pin 9 straight to 3.
[/B]

I will check on this.


Not much of a concern right now, but I would say it seems like you have spaced everything out quite a bit. Could be considered a good thing, but I am sure you could compact it down a fair amount if you desire. Get it all correct first though, of course
[/B][/QUOTE]
I will try. I really appreciate all your effort, thanks

Originally posted by impsick
what a nice guy, taken the time to review your layout!

yeah just read Marks reply a couple time through you'll get it. And i wouold suggest just using the National schematic for the 3886 on there site. Its probably a bit more simplified. Looks like you might be mixing up the 3875, and the 3886. They are different chips so stick to there schematics.

I only added a zobel network, but I forgot some stuff mentioned above , thanks for the library.

[OmarAvelar]

Has RF onboard, along with the feedback capacitor (which should be 4.7pF), and arranged better, power ground isolated from signal ground , I think it looks better and hopefully made no more mistakes

[markiemrboo]

Originally posted by OmarAvelar
[B]
I read that I need to use a jumper If i want to wire them together, because apparently it will cause some kind current inside the package, but I didnt quite understand so I left it like that. Do you happen to know anything about this?

I'm not sure about this. I would have thought a trace and a jumper were pretty much the same. But what really matters is that you know pin 1 and 5 are supposed to be connected

lol, thanks, but what happens if I put a different value resistor?

It is explained in the datasheet somewhere. I think pin 8 needs something like 1mA in order to deactivate the mute circuitry. It is a case of using 'ohms law'.

Since my memory is pretty poor, coupled with the fact my maths isn't great, I have had a hard time trying to remember the equations, let alone which equations do what. I find this calculator very helpful: http://www.the12volt.com/ohm/ohmslaw.asp

From that you would want to use 'R = E / I', to work out what value resistor you would need to give 1mA. If you have +/-28v power supply rails (E = 28v) and you want 1mA (I = 0.001A), this would mean a 'minimum' resistor of 28,000R, or 28k (R = 28 / 0.001).

I give a tiny bit more just to make extra sure mute stays off. I've seen 10k used here, so it shouldn't hurt

I will try. I really appreciate all your effort, thanks

No problem.

Originally posted by OmarAvelar


Has RF onboard, along with the feedback capacitor (which should be 4.7pF), and arranged better, power ground isolated from signal ground , I think it looks better and hopefully made no more mistakes

I think I may have confused you even further


The feedback resistor (22k) / capacitor (4.7pF) is there now, but in the wrong place, though they are correctly in parallel to each other. They should be moved to before the 680R. One side of them should connect to pin 9, and they should just go straight over to pin 3 (output).



Pin 7 should be connected to ground. Input or power ground? Not sure. I do seem to remember reading somewhere it was better to connect this to input ground.



The mute resistor between pin 8 and -ve still seems to be missing, but I presume you just left this for now as you're not sure of the value.



You seem to have the input resistors in series. The input should go to the input pin via the series 1k resistor, but the 22k resistor here should be going to the input ground. i.e joined to pin 7 and the end of the 22uF capacitor. These three returns would form the input ground point.

You also seem to be missing the 4.7uF input capacitor. It is not strictly necessary, but if the input has any DC it should remove / significantly reduce this. DC at the input will get amplified and end up as much larger DC at the output. To an extent, DC at the output is "A Bad Thing".



You appear to have moved the output Zobel on to the power rail ground.... to... somewhere. This should be from the output pin (pin 3) to ground. You got this right the first time

I think some might suggest that this be off of the board, soldered directly to the binding posts. Some might also suggest that the ground return from this be separate from even the power ground. I honestly don't know what the best way to do it is. I can tell you that it seems to be a bad thing to have this Zobel returned to the input ground. I personally haven't had any problems returning it to the power ground, and just having the input ground and power ground separate.



It appears you might also have the -ve and +ve rail mixed up. You have labelled the trace going to pin 4 +ve and the trace to pin 5 -ve. Pin 4 should be -ve and pin 5 (along with pin 1, as previously mentioned) should be +ve.



I have attempted to draw a layout to try and demonstrate the changes. I think it might help, or at least I tend to find pictures and diagrams easier to understand. To be perfectly honest I don't know what makes a technically 'good' layout (taking in to account RFI and all that), and I don't know whether the separate traces from the power supply capacitors to ground should be 'starred' as I have (I do think it looks quite cool), but it should, I hope, be correct



I should also note that I have used a polarized electrolytic for that 22uF feedback-to-ground cap. I used a polarized electrolytic here myself without any apparent ill-effects. It kept output DC offset in the mV's, didn't explode, and didn't seem to have any effect on sound quality. It might be wrong, and I expect most audiophiles would cringe at the thought of an electrolytic there, but it worked.

If you don't want it there, you can just stick a little jumper in instead. I don't use it anymore, as I have just accepted that a little DC offset is OK. I would be OK with up to about 70mV, and you should be able to achieve that very easily even without the cap there.




Hope this helps

[Minion]

If going by the Datasheet the Mute pin on the LM3886 has to be connected to Ground through a 100uf Cap not a resistor (maybe a resistor will work but I have never tried it but I know that the 100uF cap does work).....

[markiemrboo]

Originally posted by Minion
If going by the Datasheet the Mute pin on the LM3886 has to be connected to Ground through a 100uf Cap not a resistor (maybe a resistor will work but I have never tried it but I know that the 100uF cap does work).....

The datasheet has a 100uF cap going from pin 8 to ground and the resistor going to the -ve rail.

The cap, as far as I know, is for a small delay. Slowly turning the mute off in order to avoid any turn on (/off?) noise.

I have been fine without the cap so far, no noises even without it

[jaycee]

Just to show off, my attempt Not quite a gainclone as I have a seperate capacitor bank of 2x4700uF per rail, but...



Schematic here. In case you were wondering, the output zobels (10R + 100n) are across the speaker terminals. The board is roughly 90x42mm.

[DJ Exprice]

What about Russ White's SuperSmallSingleSided LM3886 monoblock that's about the size of a credit card cut in half? Search the forum for the Eagle files.


(PS: Eagle is THE best PCB design software to ever exist.)