Single supply LM3886 layout

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
There are times when a single rail amp has its uses. Here's an attempt for an amp running on +48V (from an smps for example). That's pretty much a datasheet implementation.

A big concern was the thump on start up with an output cap. Hence the output relay for muting at start.

What do you think, any errors, improvements to be pointed out ?

The board is 8*5 cm big.
 

Attachments

  • se lm3886-2.png
    se lm3886-2.png
    44.4 KB · Views: 677
  • se lm3886.png
    se lm3886.png
    7.9 KB · Views: 674
Member
Joined 2009
Paid Member
You might want to discharge C16 faster when the supply collapses on turn-off.

It also looks like you're running the signal ground through the high-current portions of the ground plane. I suggest finding a quieter route for that.

I suggest revisiting the decoupling on the LM3886 as well.

Tom
 
Last edited:
Thanks for weighting in. I must say though that I don't get all of your comments.

- on the discharging of the c16 cap, I see the point.
- on the grounding: all signal ground points have their own groundplane (at the right of the board, in blue). This plane connect to the power ground at the output connector. What else should I take care of? The layout in the first post didn't have that plane though.
- I'm ok with revisiting the decoupling... What would you suggest to improve? Is it a question of values or layout?

Thanks for any further help.

Edit: for discharging c16 would adding a diode from the positive terminal of c16 to the 48v rail do the trick? That and replacing r7 by a 4.7v zener to protect the mosfet.
 
Last edited:
Comments rather than criticisms further to the above about the layout....

R4 could be better off connected to P on the relay (leave O disconnected). This ensures the C10 negative plate never has chance to charge to any DC voltage.
R4 at 330R also needs to be rated 1W at very least for worse case dissipation.
Using 1K instead allows you to use normal 0.6W/MRS25 resistor and should be perfectly adequate.

R12 and 13 should be reasonably substantial metal or carbon (does not matter in this position) film 2W or more whatever you can fit.
If anything goes unstable you want them to have at least a chance of coping and living to tell the tale!
Whatever you do; do NOT use wirewound in R12 position, not even the "non-inductive" type!

Mount R4, 12 and 13 about 3-5mm off the PCB to allow air to circulate and so they don't scorch the board if they get hot.

Don't think you need both C3 and C4. C4 alone should suffice otherwise keep an eye on your required treble response....

Very best regards,
Simon
 
Last edited:
You're right on c6 of course.

Indeed, r4-12-13 are intended to be 1w.

On r4, it might be a concern as it is if one plugged in a speaker with the amp already powered on. The reason I did it that way is that I will be using the amp in a powered speaker. I need to check again the delays to see if I can go up to 1k. What I could also do is have a low value r4 for startup and add a high value resistor before the relay to keep the negative terminal of the output cap at ground.

Thanks for the comments, that helps :)
 
No r4 would have no effect beside heating up a bit without reason. It wasn't a matter of r4 affecting the speaker, just that the isn't any reason to worry about connecting or disconnecting the load.

It goes without saying on c6 voltage rating... except that I apparently missed the obvious.

Edit: for the higher power resistors, I usually use the vishay pr02 serie.
 
Last edited:
Have another look at your 2n7000 switch on and off.
The 220k will give ~4.4V on the gate, if the electro does not leak and Vcc=48Vdc.
If the electro leaks half the current passing through the 2M2, then you'll only have ~2.4V at the gate.
the 2n7000 with 4.4Vgs will easily switch 50mA for the relay, but with only 2.4Vgs might struggle.

As Tom says 220k & 100uF is slow and has an RC time constant of 22seconds. You need a more sophisticated relay ON/OFF.

Q1.
why do you need RC = 220seconds for start up? (2M2*100uF)
Q2.
what time delay does your circuit need?
Q3.
Why use a big leaky electro, when a low leakage plastic (MKT) would do?
Q4.
Have you estimated electro leakage and it's effect on timing?
Q5.
Why no "loss of AC" detect?
 
Last edited:
I did also think his time delay would be 220s too but looking at the gate circuit it would only get to 1/10th supply voltage anyway as there is divider formed by R7/R11 and anyway the mosfet will turn itself on much earlier than that at a couple of volts or less .....
 
the Vgs @ 5V needs ~ 0.16 of the RC to fully switch on. 220seconds * 0.16 gives a fully on @ ~35seconds. Does he need to wait half a minute for the circuit to stabilise?
Even if the 2n700 switches @ 3Vgs, it will still be a delay >20seconds.
Somewhere between 1second and 5seconds seems more appropriate. That's why I proposed he ask himself Q2.
 
Assume the gate needs 6V.
63% of 48V = 30V. That's the voltage that will have charged up the capacitor that does not leak when 1*RC has expired.
6V is ~20% of 30V so you are looking for ~ 20% of the RC as your timer value.
Thus for 5seconds delay you need an RC of 5seconds/20% = 25seconds.
2M2 & 11uF = 24seconds
You could use 10uF low leakage electro, or if you can tolerate a shorter delay use a 4u7F MKT.
or change the circuit to use a voltage dropper so that you use more than 20% of the RC to turn on the mosFET.
Replace 220k with a Zener to protect the gate from excessive voltage.
 
I just put the delay system into LTspice to check the values. Here are the results.

Q1: with the values given the relay will fire up at about 10 seconds, not 220. It only draws 10mA btw, not 50mA. The 2n7000 will start conducting enough current at about 2V vgs.

Q2: Actually 5 sec delay is not sufficient. For 4700uF into 330r to settle under 0.5V, I need more than 7 seconds. 1V is still 6 seconds. So a simulated 10 seconds delay is not unreasonable for thumpless switching. I can always reduce the value of the 2M2 resistor if it is slower in practice.

Q3: MKT are expensive in the values I need and a 100uF/10V cap is hardly a "big" electro.

Q4: the amp will be powered by an of the shelf smps. So no isolated AC available for AC loss detection.

So, in line with post #5, I attached what I plan on using finally.
 

Attachments

  • rc delay.PNG
    rc delay.PNG
    34.3 KB · Views: 195
the Vgs @ 5V needs ~ 0.16 of the RC to fully switch on. 220seconds * 0.16 gives a fully on @ ~35seconds.

The voltage across the cap will grow exponentially, not linearly. Also, the MOSFET will start to turn on before Vgs reaches 5.0 V.

Rather than a large RC, I generally use a smaller RC with a zener diode in series with the base of a bipolar transistor to drive the relay. With a low-current zener, the base current of the bipolar will bias the zener. This will make it possible to have the relay engage some time after the supply voltage reaches the zener voltage.

The circuit shown in the post above is horrendously complicated for what it does. I think there is some value in minimizing the noise of the input reference, but the relay driver circuit is just complicated for the sake of being complicated in my view.

Tom
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.