Separate analog & digital supply MCU

[Vindhyachal.takniki]

1. I am using STM32 - 32 bit cortex M3, it has separate analog & digital supply pins.
2. Datasheet says :

VDDA and VSSA must be connected to VDD and VSS, respectively.


3. Since I have 12 analog channels connected , so avoid any digital noise to come in analog signal can I connect a inductor = 10uH & cap = 0.1uF (as filter) between AVDD & DVDD.

4. Can I do so?

Edit: Or can I connect a separate 3.3V regulator for both VDD & AVDD with common ground.
But this I think won't work as datasheet saya they must be connected together

[Paul Price]

Ok with me.

[AndyC_772]

VDDA and VSSA must be connected to VDD and VSS, respectively.

This is your clue.

Separate regulators aren't a good idea. They won't have the exact same start-up and shut-down characteristics, so during those times, there will be a significant voltage difference between the analogue and digital supplies on the chip. Depending on how the chip is designed, these differences might cause substantial current to flow through the die, which in turn can destroy the chip.

A filter between analogue and digital supplies is unlikely to cause any harm, but think very carefully about how useful it will actually be. What noise frequencies do you expect to be present on the digital supplies, and how effective will your filter be at those frequencies? If you really need the last couple of LSBs of accuracy and/or resolution, would you be better off with a separate ADC chip anyway?

[Vindhyachal.takniki]

My MCU is running at 120Mhz core frequency with prefetch & instruction buffers on.
This may cause noise as digital circuit is switching at high frequency.

You are right about problems associated with separate regulators.

I will go with 10uH inductor as filter.
Here is inductor which I will use:
http://www.digikey.com/product-detail/en/LBR2012T100K/587-2045-1-ND/1788992

[hans]

A 10uH inductor on it's own does not make a filter.

Combined with an e.g. 100nF capacitor you have made a LC filter which may or may not work well. Think about resonance points of LC circuits, and how they actually can amplify instead of dampen a (very) particular frequency.

I personally prefer sticking a ferrite bead on there, because it has a lower Q and is less likely to form a resonance point.

[AndyC_772]

My MCU is running at 120Mhz core frequency with prefetch & instruction buffers on.
This may cause noise as digital circuit is switching at high frequency.

Well, yes, OK, but you then need to consider very carefully the effect of your filter at those frequencies, as well as their likely effect on the ADC.

As there is a great deal of circuitry already on the die switching at those frequencies, noise is going to couple capacitively into the ADC from one part of the die to another, and your filter doesn't feature in that mechanism at all.

If you're using a QFP packaged device, then there may be, say, 10mm of lead frame and PCB trace between the die and your filter cap. Work out the inductance of that trace, and its impedance at 120 MHz (and higher harmonics, of course), and think about the effect this is going to have on how effective any filter can possibly be.

Also worth considering is the operating frequency of the ADC itself. What's its bandwidth? What effect can 120 MHz noise possibly have on it?

I don't have all the answers, of course, but there's a lot more to this issue than "I have noise, therefore I need a filter". Depending on your application, I expect what you need is either a simple capacitor right across the VDDA and VSSA pins, or a separate ADC.

"Heroic" attempts to filter the digital noise out of an ADC, where that very noise is generated on the same die as the ADC, are a waste of time, IMHO.

[leppie]

Depending on how the chip is designed, these differences might cause substantial current to flow through the die, which in turn can destroy the chip.

I have been burnt literally by this. Some Analog 24-bit ADC I was trying to hook up. Did the analog parts without hooking up VDD, 150 deg C in a second.