Help! I'm using the MCU to drive a MAX7221 LED driver. I'm also generating PWM audio. I wanted the sound to be louder so I added a simple audio amplifier, the TDA7052. It works, but there's a constant high-pitched whine.

I tracked it down to noise on the power supply caused by the MAX7221. It's supposed to be a low EMI design, and I'm using the bypass capacitors as specified, but there's still significant noise. After going through the amplifier stage it's 1V p-p!

I have bypass capacitors at the amplifier, at the power supply, and everywhere else I can think of but I can't eliminate the whine. Any ideas?

Ok, it's not 1V p-p, it's 0.1V (I had the scope on a different setting). But it's just as loud! :-)

Can you tell what frequency it is.... 60Hz? Higher? You may need to put in a notch filter. If you can identify the frequency, you may be able to tell what it is, and filter for it.

BM

It's complicated. The fundamental seems to be about 800Hz, but there are a bunch of other things going on. It depends on what LEDs are lit up at the time. The datasheet says it can be between 500Hz and 1300Hz with 800Hz being typical. I can't filter out a notch that big because I'm using actual audio in that range as well.

There's a chart that says the scan frequency is related to the supply voltage (constant) and number of digits enabled (constant), in which case it should be constant, but there are still a bunch of harmonics. Comb filter then? And how to apply a comb filter to the power supply?

Some times, it takes more than capacitors to control noise.

It sounds like the LED driver is causing some loading on the power supply, which varies according to the number of LEDs being lit. I take it that they are probably multiplexed somewhat, which contributes to that loading. The power supply in turn, is not recovering or maintaining stability on the output, impressing the signal on the supply rail.

An additional method of controlling this kind of noise is by using inductors in the supply lines. They block the higher frequencies from passing and by adding something like that between the source of the noise and the power supply might help a lot. The audio portion would not be connected to the same side of the inductor, otherwise, it will still pass the noise through.

So two inductors, one in +V and one in GND?

With a 8V supply the audio portion is drawing about 50mA, which means it's seeing a load of about 160Ω. So if I want a first-order LR filter with the 3dB point at 400Hz, I should pick LxR = 0.0025sec, or L=16uH? Something like this.

Oops, divide by 2pi. L=2.5uH?

More oops. I got that completely wrong. It's supposed to be f=R/(2πL), or L=64mH. I should just put those inline with the power rails feeding the audio stage?

Holy cow, they're huge.

A little experimentation comes in handy at times. Try this:

60 turns or #24 magnet wire around a torroid core. Easy to come by if you have some old audio transformers from old radios. Just use the transformer core and rewind it. The larger ones may be more than needed. Essentially, you are making a low pass filter that passes frequencies below the audible range and blocking those in and higher than that range.

Place it between the power supply and the source of the noise, and it only needs to be on the positive side. Since it's a common supply, you have common mode noise. The choke is the separation between the noise and the audio portion. (AC Decoupling) If you like, another choke to the audio section would not hurt.

I chose the 60 turns arbitrarily. It could be more and it could be less. Typically, the more turns, the more effective, but higher DC impedance. (but MUCH higher AC impedance)