Pulsed Current Draw In Audio

[ssashton]

Hello All,

I'm new here so thanks for having me! I've been watching Dave's videos for the last 6 months or so and learning a lot. 

I am a one-man business making loudspeakers. I am attempting to build an amp PCB that has a bluetooth audio module on it. I have done my best with grounding to minimize noise but I do still hear a low level tone carried on the audio signal. It really is low level, but I want to reduce it if possible.

I have measured the audio output as shown below. I believe the tone I hear is the 3.2KHz blip. The lower frequency noises are really inaudible (hearing is less sensitive at low frequency). It is coming through the ground as I can pull the audio input of the amp to gnd and still hear it.


Yellow power direct to BT module. Green 22R in line with power for current sense.

I thought this is probably pulsed current draw being de-modulated by the audio circuits. So I put a 22R resistor in line with the power feed to the BT module and put my 'scope on it. Indeed I can see a very periodic current draw! However what I didn't expect is that the period is 1.25ms or exactly 800Hz. I thought it would be much higher!


BT current draw during playback, audio muted.

So.. how do you guys think this 3.2KHz tone can be appearing in the audio band? Could it be what I am hearing is just the rising and falling edges of that period?

My plan had been to measure the frequency of the current draw and then try to measure the ESR of a few capacitor combinations to get minimum impedance at that frequency on the output of my regulator. I currently have 100nF multi-layer ceramic in parallel with 100uF electrolytic Panasonic FM series. Playing here makes some differences but I'd like to be a bit more scientific.

Also of note - I didn't expect the high frequency noise to drop when I inserted 22R with the power line, but it has. What's going on there?!

[magnusbh]

Hi, I am also new here but not in electronics.

I have been building mobile phones for a Sony many years and the two most common problems with "bumblebee" are the audio trace to the amplifier.

1, power trace to transmitter/receiver close to audio traces (signal,GND)
2, GND problem.

1, easy to see and change in the cad.
2, You will need to have a current-"free" trace fror the audio GND, do NOT share audio GND with the plane for power or transmitter. Think coax cable when you connect the ic´s. The Audio in from receiver has a GND, let this trace run in parallell with the audio signal but donot connect it to the GND plane. let it only be connected at the input GND pin. You can patch a trace in the air and test (remeber to cut the trace on the pcb).

Good luck!

BR
Magnus

[ssashton]

I have done a little investigating. Changing my regulator capacitors to Tantalum 15uF instead of Panasonic FM 100uF has reduced hiss noise floor, especially in the high frequencies. The little tone still remains.

I started to wonder if it was the LED PWM driver but it was not actually connected to a load and when I added a 1K resistor the pitch of the tone did not change. So I don't think it's that. There are also 2 SMPS regulators built in to the bluetooth IC. I tried loading each of those lines with a 1K resistor and again the pitch of the tone does not change so I don't think the noise is either of those.

I'm going to try playing with a different BT module PCB and see what I can find. I'm pretty lucky the noise is so small to be honest. I can only hear it with my ear close to the speaker and the amp at full volume with no music playing. I like to try and get things perfect though!

P.S. I ran an FFT on the current draw shown above. There was an 800Hz component as well as something at 1.4KHz IIRC. Nothing directly at 3.2KHz.

[DaJMasta]

The 50Hz hum that's visible on the spectrum may be audible in some very low noise situations and it's coming from the mains power - if you choose to address it, it is a fairly common problem and some re positioning or changing of some of the initial power supply stuff should help.  Given that it's the peak, even if it's not very audible, it's certainly a potential issue.  Just physically moving input lines away from signal level stuff can help, as well as shielding any signal lines that go over mains lines, but stuff like bigger filter cap banks in the PSU or a toroidal transformer can help too.


As for the 3.2kHz tone, have you looked at a spectral analysis of that BT module current draw?  Could be that the sharper edges of the pulses coincide with a 3.2kHz tone even though the base frequency does not, and since 3.2kHz is the 4th harmonic of 800Hz, then they are probably linked.  There's also a chance that the 3.2kHz tone is a mixing product of much higher frequencies interacting.

Try isolating the digital power supply and ground plane.  If you are working with a PCB, dividing the analog and digital ground planes can help reduce the digital related quick return currents from interfering with the more sensitive analog stuff.  Otherwise, I'd probably put a small inductor in series with the bluetooth module's power supply.  Something like main filter caps on the supply rail -> inductor -> secondary digital side filter caps, these LC filters can be a really good way of isolating noise from getting back out into a system or quieting down a noisier supply.  Bigger inductors and capacitors mean a lower rolloff frequency, and if you were only filtering out MHz+ level noise, you could probably get away with a ferrite bead instead of a full on inductor.  Of course, you want the inductor to have a low enough DC resistance and high enough current capability to supply your digital section, and you want the capacitors to be big enough relative to the inductor so that you don't get a big voltage spike when you first power up or shut down the device and small enough to prevent your supply from needing to feed a massive capacitive load (if that's an issue, like with switching supplies).  A spice simulator can give you a good idea of ballpark performance for given values, but for dozens to hundreds of uF of caps on either side and just a few mA draw for the modules... maybe 10-22uH of inductance and less than 2 ohms of DC resistance?  Just a ballpark guess.

If you look at some datasheets for low level analog designs (like with a sensitive DAC or ADC), there are some pretty good explanations of separating analog and digital ground planes and examples of decoupling power between them.  They'll also probably show you some examples where using different linear regulators for each helps with noise isolation (so, if you have a 7.5V input or something and need a 5V rail, you have a regulator for the digital and a regulator for the analog).  Then, there's also physical separation on the PCB layout, minimizing digital trace lengths, flood filling the ground plane, and even via stitching between digital and analog sections for that little extra bit of isolation.

It's a complex thing to really nail the execution, but there's a lot of sort of general procedures that can at least improve performance.