[sdiy] PWM noise in analog ground in Optigan-like system

[Neil Johnson]

Hi,

Tom Wiltshire wrote:
> Am I right in thinking that the motor is powered off this same  
> power supply?

*ouch*

> If you've got the TIP120 switching the motor on and off at the PWM  
> rate and the motor is on the same power supply as everything else,  
> I'm not surprised you've got noise. It's a high(ish) current load,  
> and switching it rapidly on and off is going to put spikes on the  
> power rails. I don't think the problem is specifically the PWM,  
> since I've used PWM lots and not had problems with it bleeding  
> through to the output - rather the problem is the load.
>
> I've seen references to spikes caused by the inductance of motors  
> when the power is turned off, and diodes across the motor to help  
> prevent this. I understand the gist of it - a motor is dynamo once  
> the power is turned off, so if it's still moving (which it is)  
> it'll generate voltage, but you'd need someone better versed than I  
> with motor control electronics to tell you if that is the issue.

Its worse than that:

1/ Assuming its a DC motor with brushes you're going to get a large  
amount of very wideband noise from the commutator-brush mechanism  
(think spark-gap transmitter).

2/ Common 'ground' is a great way to couple noise between high- 
current and low-voltage circuits.

3/ The reverse-protection diode will protect the transistor, but  
consider now that the Drain is bouncing up to one diode drop above  
the supply rail at the PWM frequency, and that any energy in that  
spike is pushed onto the supply rail.

4/ These narrow high-frequency spikes are going to capacitively  
couple into everything inside the box.

> Could you try an oblique strategy, like change the PWM frequency to  
> be outside the audio range? If you could get it up to 30KHz, all  
> your problems might magically disappear. That'd be my (typically  
> hacky) approach!

Well, you could, and the effectiveness of any filters will increase,  
but then so will any capacitive coupling of the noise into the audio  
path, and then any non-linear components (intentional or otherwise)  
will in turn help to produce more audible noise.

Really, physical separation is your best friend here.

Cheers,
Neil
--
http://www.njohnson.co.uk