Analyzing uC power decoupling

Question

I am analyzing the schematics of STM32F429 Discovery board, specifically power delivery circuit, here it is:

I have few questions that I couldn't get by myself:

1. Capacitors C26 and C25 are tantalum capacitors(used before ADC supply pin), they are 1 uF, but on the other hand capacitors C42 and C41 are 2.2 uF and are cheap ceramic caps(they are used for internal voltage regulator). The question is - why they use tantalum capacitors before ADC? Can I plug ceramic capacitors there or aluminium electrolytic? Tantalum are expensive and many people say unreliable, basically on any forum people say something like "stay away from tantalum caps". What are implications of using ceramic or aluminium electrolytic capacitors(or even through-hole electrolytes) in ADC decoupling?
2. The ADC power input is basically pi-network(assuming we have cap near the supply) with exception that we have not 1, but 4 capacitors. Why not just use two capacitors(big and small to cover more frequencies shunting to ground). Putting two of identical capacitors pairs seems redundant, it would make sense to use different values like Dave says.
   Not longer relevant, see comments.
3. This question is more generic one, not only for this schematic. Often when it comes to inductors people just say "Oh, and here we have inductor". I am quite fascinated how people treat inductors like you can replace it with anything. How big the inductor should be? Does "bead" means that this is not actually an inductor, but a ferrite bead? Is it really so irrelevant and you can put in anything you have laying around, no matter if it is 10 uH, 10000 uH or small ferrite bead?

Answer 1

Capacitors have inverse characteristics between C and ESR, such that C*ESR=T is constant for the same design (material, type, size, voltage).

There are many types and I find it useful to explain low ESR in terms of this reference constant, T. This time is the fastest time that you can expect to charge or discharge the capacitor. If you pulse any shorter, it behaves more like a low resistor value. Ideally it should be zero, but as caps are made smaller in C , the ESR rises and visa versa.

Considering cap technology has improved over the decades, the present norm is:

• T = 100 µs - 10,000 µs general purpose (small to huge caps )
• T = 10 -100 µs low ESR
• T = 0.01 to 10 us ultra low ESR

(problem with some ultra low ESR caps is cost in large plastics values and microphonic noise in ceramics on analog supplies , which can be "heard" by ADC's) COG/NP0 ceramics are noise free but lower k.

Now I just went to DigiKey and searched for Tantalum 1uF caps and sorted by lowest ESR, with "-" at the top of the list (unspecified)

The 1st item was KEMET T491A series 1206 SMD $0.34(1) with ESRmax = 10 Ω.
http://www.digikey.com/product-detail/en/kemet/T491A105K010AT/399-9449-1-ND/3522981

Then I checked ROHM 603 SMD same result.

There are ESR variations with V ratings.

Then I searched for Alum elect. caps and in 1uF. Here ESR that was listed, ranged mostly from 95~415Ω with exceptions down to 0.41 (0 stock and 3k MOQ) , 5 & 12 Ω.

Conclusion, you can be careful and find a lower ESR cap, but if they say Tantalum, it is for a good reason since the solid types are consistently low ESR and not more expense in general. Solid Tantalum have always been low ESR unlike Alum, so they don't need to advertise as such.

The ferrite bead raises the track impedance >50MHz (which includes DRAM current spectrum).

• If the bead spec says 50 -100 Ohms then going into "JUST" the 1 uF low ESR tant cap then this ripple noiseis attenuated 5:1 to 10:1 then further with the 0.1 uF cap.

The Analog Vss needs ripple to be low for the ADC Vref. The ripple can be measured with a short scope 10:1 probe pin and barrel or spring clip to gnd <1cm to avoid inductive ground loop in the probe >10MHz.