ADC filtering, induktor resistance

Go To Last Post
9 posts / 0 new
Author
Message
#1
  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

The LC-filter to the AVCC pin is recommended in all datasheets. Recommended values are 100nF and 10µH. But what about the inductor DC resistance? Wrong value here will have great impact of the efficency and/or the frequency responce. Also, there is a size matter, the smaller size of the inductor, the higher DC resisistance.

Does anyone have any recommendations of the maximum DC resistance of the inductor?

My favorites:
1. My oscilloscope, Yokogawa DLM2024.
2. My soldering iron, Weller WD2M, WMRP+WMRT.
3. JTAGICE3 debugger.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

I think the DCR is not really important, usually the DCR is in the range on 0.1 to 1 ohm for 1810 style inductors. The AVCC pin does not draw that much current that ohmic losses will play a major role like in switching regulators. At 300uA and 0.5 ohm resistance you lose 0.15mV. The main task of the filter is to filter out high-frequency noise and the exact frequency response is not critical. It's more important to make sure the inductor does not saturate because it does not work as an inductor anymore, but at these current levels saturation is no issue.

Maybe you could try different inductors and see the effects of the DCR.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

A 1810 inductor is usually a "supertanker" in my point of view. A 1008 is BIG and a 0805 just fine. 0603 is to prefere but is out of question for this range of inductance.

But thanks for good advice, of course the losses will be very smalla as no other circuitry draws power. I think the best would be to make a trade off between the two components so their size is almost the same. That leads to a smaller inductor with rather high DC resistance and a capacitor of for example 1µF in 0805. I will check the datasheet for current conmsumtion of the ADC.

My favorites:
1. My oscilloscope, Yokogawa DLM2024.
2. My soldering iron, Weller WD2M, WMRP+WMRT.
3. JTAGICE3 debugger.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Hi,

efficiency: The DC resistance has to low enough to ensure proper Power supply.
(my opinion: a voltage drop of less than 100mV is good)

frequency response: Its the frequency of the power supply, not of your analogue signal. So here is the lower frequency the better. In my opinion the worst case is: 100Hz of an unregulated mains supply (does anybody use this?) . Noise from an linear regulator output should be no problem. Noise from a switch mode power supply in burst mode (no continous switching) can be problematic. Here check your noise frequency and use an appropriate filter.
The more important is the noise on your VRef-Pin, because this will immediately change your ADC output.

In some designs with linear regulated power supply i used only a 10 Ohms resistor instead of the inductor and it always had equal performance as in datasheet.

Good luck

Klaus
********************************
Look at: www.megausb.de (German)
********************************

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Typically the power supply will be reasonable good filtered but of course noise from there probable will have great impact on ADC noise. But I think noise from the MCU itself will be a more typicall source for noise into the ADC. To reduce that noise as much as possible the ADC VCC pin should have its own power supply wire drawn directly from the power supply node filter capacitor.

ADC Vref if of course absolutely mandatory to have well filtered.

But above all else I think it is very strange that Atmel haven't made one single note about recomendations of the maximum DC resistance to the ADC VCC pin. For size reasons I will try to use a 0603, 220 nH, 2.5ohm inductor and a 220 nF capacitor.

My favorites:
1. My oscilloscope, Yokogawa DLM2024.
2. My soldering iron, Weller WD2M, WMRP+WMRT.
3. JTAGICE3 debugger.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Could i use 33nH inductors instead of the suggested 47nH ???

I have a lot of 33nH lying around

There is something with the resonance freq here
http://8515.avrfreaks.net/index....

I will be running 14.xx or 16 Mhz

/Bingo

/Bingo

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

if 10uH is recommended (first post) then 33nH is a bit too low. It´s factor of 300!

Klaus
********************************
Look at: www.megausb.de (German)
********************************

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

Some datasheets & app notes say 10 uH and others say 47 nH. I think I saw 100 ohms in one paper. The STK500 uses a ferrite bead.

Atmel gives us no clue how they came to any of these. If we assume that the ADC is most sensitive to noise at its sample frequency (ca. 200 KHz), then maybe we can guess what we should be doing.

10 uH has about 12 ohms of reactance. DC resistance probably negligible.
47 nH has about 0.06 ohms. DC resistance negligible.
100 ohms, well...
Ferrite bead ("1000 ohm" BLM21A102S) has maybe 50 ohms at 1 MHz (200 KHz is off the graph), DC resistance 0.45 ohms.

I can't find the AVCC current consumption, but don't forget that it also supplies the pullup current for a port.

  • 1
  • 2
  • 3
  • 4
  • 5
Total votes: 0

It is important to focus on what the purpose for the L and C network really is.

IF you use the power supply as a reference, then power supply noise becomes reference noise. This, in turn, leads to noise in the ADC value even when the signal being measured is very clean.

So, what is the predominant noise on the power supply? In close physical proximity, it is the processor, itself. Thus, the filtering that you use should depend on the clock frequency. For a given L and C, the corner frequency (radian frequency) is 1/sqrt(L*C). The lower you make this relative to the clock frequency, the greater the noise reduction will be. You can estimate it by determing the reactance of the inductor at the processor clock frequency and the reactance of the capacitor at the same frequency and pretend that they are resistors forming a voltage divider. This will estimate how much the power supply clock noise voltage is reduced before getting to AVcc.

That said, there are some practical limits. There is little or no point in reducing the noise on AVcc below one bit worth. Further, as you increase the L and the C, they tend to be less and less effective at higher frequencies. Thus, there tends to be a "sweet spot" in the L and C values and where that spot is depends on the clock frequency!

Yes, it is a little disconcerting that Atmel does not provide more concrete examples of appropriate components. On the other hand, it may be quite difficult to provide "guaranteed to work" examples. After all, how important this is really depends on how much power supply noise you have and that depends on how good a layout you have and how effective your bypass caps are.

With respect to the initial question about inductor resistance, suppose that the resistance is a whole 1 ohm. Then 20 milliamps flowing through it would generate a 20mV drop. In any real supply, you will have more than 20mV variation in the supply voltage (that's only 0.4% out of 5V!) and is worth about 4 LSBs in a 10-bit conversion. What would be more problematic is if something that takes substantial current, like a LED, is on one of the pins supplied by AVcc. Then, you would have a changing voltage drop, depending on whether or not the load current is on. That gives you a shifting reference voltage, depending on the logic state of one of those supplied output pins.

Basically, you are going to have to go back to your "electrical fundamentals" and choose the values, yourself. And, don't box yourself in by insisting on the smallest possible inductor packages!

Jim

 

Until Black Lives Matter, we do not have "All Lives Matter"!