Audio circuit grouding philosophy and layout with 1MHz SMPS

[Jester]

I'm working on the ground layout for an audio circuit. The bandwidth of this circuit is limited to 2400Hz.

The device is powered from a single AA battery, the U5 boost converter is used to provide +5V for both the Op amps etc. as well as the U4 boost converter constant current source. U4 is actually providing constant current for 5 of these circuits (not shown in the schematic).

My questions:

1) Should I run separate ground connection traces from U4 and U5 back to the negative terminal of the battery or simply connect them to the ground plane?
2) I expect the 5V bus will be somewhat noisy because both the U4 & U5 choppers are running at 0.5-1MHz. Should I regulate or add more filtering to the +5V bus?
3) I can likely lay this out in 2 layers, however a 4 layer board is an option, will the full ground plane of a 4 layer board help significantly?
4) Should I void the ground plane under the L1 and L2 inductors?

Thank you

 

[FvM]

In case of doubt, a solid ground plane won't be bad.

The LED driver circuit can't work. The IR-LEDs have only 1.2V forward voltage each, so a simple series resistor from 5V would be the appropriate way to supply it, or a buck converter if minimal power consumption is a must.

 

[Jester]

In case of doubt, a solid ground plane won't be bad.

The LED driver circuit can't work. The IR-LEDs have only 1.2V forward voltage each, so a simple series resistor from 5V would be the appropriate way to supply it, or a buck converter if minimal power consumption is a must.

There are actually 10 LED's in series, not shown on the simplified schematic, but mentioned in the text.

 

[FvM]

There are actually 10 LED's in series, not shown on the simplified schematic, but mentioned in the text.

I see.

Back to the ground problem. As said, a continuous ground plane is usually a good idea. For a very sensitive circuit, you should avoid battery or switcher current flowing through the analog circuit part. A separate power ground with a single connection point to analog ground can be a way to assure this. But if you split grounds, there must be supply filter inductors between the power and analog section and bypass capacitors on both sides of it.

 

[schmitt trigger]

How far away are the converter's switching frequencies? Although the 0.5 to 1 Mhz switching is far away from the audio range, I would be concerned that intermodultion products (beating) actually fall within the audio passband.

 

[Jester]

I see.

Back to the ground problem. As said, a continuous ground plane is usually a good idea. For a very sensitive circuit, you should avoid battery or switcher current flowing through the analog circuit part. A separate power ground with a single connection point to analog ground can be a way to assure this. But if you split grounds, there must be supply filter inductors between the power and analog section and bypass capacitors on both sides of it.

Is a 100nF ceramic adequate for decoupling on either side of the inductor?
Taking into consideration, the switching frequency of the two choppers, (500kHz and 1.2MHz) what inductor value would you recommend for the power feed to the analog section?

Thank you

 

[schmitt trigger]

what inductor value would you recommend for the power feed to the analog section?

Thank you

As large as possible, which at the same time has a self resonant frequency much higher than the highest switching frequency.
I know, it is a tradeoff.

 

[Jester]

What would be better?

https://www.digikey.ca/product-detail/en/SRU2009-100Y/SRU2009-100YCT-ND/3741544 10uH, f(sr)=50MHz

or

https://www.digikey.ca/product-detail/en/744043101/732-1111-1-ND/1639172 100uH, f(sr) = 11MHz

 

[schmitt trigger]

I would use the second inductor.

In a pi-configuration (capacitor-inductor-capacitor) using surface mount ceramic caps, it should give you excellent rejection.

 

[Jester]

I would use the second inductor.

In a pi-configuration (capacitor-inductor-capacitor) using surface mount ceramic caps, it should give you excellent rejection.

I simulated it with 10uF, and 100uH, see attached.

Two more questions if you don't mind for the inquisitive mind.

1) Is it adequate that I put the PI filter only in the +5V power feed and simple maintain a common ground between the quiet and noisy sides, just making sure that the grounds from the noisy side take a separate path (from the quiet audio ground) all the way back to the battery, ie. the battery -ve terminal will be the star point?

2) You mentioned earlier to make sure the self resonant frequency was well above the highest switching frequency, is there a rule of thumb for what is safe, so for example in this case we have highest freq =1.2MHz, and f(sr) = 11MHz so roughly a decade apart, would an inductor with f(sr) of 2MHz have been too risky?

Thanks again.

 

[schmitt trigger]

A decade apart is a good rule of thumb, but of course there are harmonics...
That is why I suggested the Pi-filter.

BTW, when you perform the simulation, set the vertical scale in dB.

 

[Jester]

schmitt trigger,

Thanks for taking the time to answer my questions.