Recommended values to inductor and capacitor at AVCC & A

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

It has been mentioned before but surely it is strange that Atmel never had given any good recommendation of the values of the capacitor at AREF and the inductor to AVCC. 10µH is mentioned but not a word about DC resistance.

Read this text from the datasheet of the linear regulator TPS73028 and you will understand how much there is to know about putting in a correct value of the AREF capacitor:

Quote:
The internal voltage reference is a key source of noise in an LDO regulator. The TPS730xx has an NR pin which is connected to the voltage reference through a 250kΩ internal resistor. The 250kΩ internal resistor, in conjunction with an external bypass capacitor connected to the NR pin, creates a low pass filter to reduce the voltage reference noise and, therefore, the noise at the regulator output. In order for the regulator to operate properly, the current flow out of the NR pin must be at a minimum, because any leakage current creates an IR drop across the internal resistor thus creating an output error. Therefore, the bypass capacitor must have minimal leakage current. The bypass capacitor should be no more than 0.1μF to ensure that it is fully charged
during the quickstart time provided by the internal switch shown in the Functional Block Diagrams.

I think the topic about the output current should be the same for the AVR Vref so from that we can learn that a low leakage capacitor would be to prefere if we need to minimize the error.

The capacitor value is more complicated to figure out as we know nothing about the internal circitry in AVR so if there is something similar like the 250k resistor we can only guess. A too small capacitor will result in unnecessary noise. A too large capacitor will mean it will take longer time to charge (and decharge in case of swithing to another reference voltage). Other resistors in the AVR like reset and port pull-up are 20-60 kohm so if there are anything like a resistor in the Vref circuitry it MAY be something like this. 100 nF is usually recommended but most probably there is not any problem to increase this value but 100nF could just as well be the perfect value.

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

bengtr wrote:
It has been mentioned before but surely it is strange that Atmel never had given any good recommendation of the values of the capacitor at AREF and the inductor to AVCC.
47nH. Application note AVR042. Section 2.2 and Figure 6-1.

bengtr wrote:
Read this text from the datasheet of the linear regulator TPS73028 and you will understand how much there is to know about putting in a correct value of the AREF capacitor:
So you are extrapolating from some random, unrelated LDO datasheet to the behavior of an AVR?

Stealing Proteus doesn't make you an engineer.

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

You're wrong, application note AVR042 do not give any answer to my questions. In AVR042 they use an inductor of 47nH but in every datasheet we can see that they use 10 µH. But as I said, it is not the value in Henry we need to know but resistance or current capability. To corectly choose the inductor we need to know the peak current flow through the inductor that the ADC will consume during a conversion. The correct design flow is to start with resistance so you can calculate the voltage drop during peak current. Second you can choose the inductor value that together with the capacitor will create a filter of a certain cut off frequency depending on the noice sources in your design.

The answer to your second note is of course YES. Any source of information that you can learn something from is good information.

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

Let's see. Maximum drop between VCC and AVCC is 0.3 volts. Assume absolute worst case current of 30 ma into the AVCC terminal, that gives a resistance of 10 ohms. (note, processor itself does not take more than 30 ma in almost any case unless sourcing current).

I've gotten good mileage out of 0.1 uf capacitors, but then again, I don't switch references.

Harvey

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

bengtr wrote:
To corectly choose the inductor we need to know the peak current flow through the inductor that the ADC will consume during a conversion.
The current supplied by AVcc is not just that used by the A/D converter. It also supplies all of the current for some of the digital output circuitry - which ports depends on the specific AVR. For example, with the mega32, mega644, etc. AVcc supplies the current for PortA digital outputs.

Many of the datasheets recommend an LC filter using a 10uH inductor and 100nF capacitor. This is usually in the figure showing the recommended board layout in the vicinity of the ADC pins in the section entitled "Analog Noise Cancelling Techniques".

Don Kinzer
ZBasic Microcontrollers
http://www.zbasic.net

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

bengtr wrote:
You're wrong, application note AVR042 do not give any answer to my questions. In AVR042 they use an inductor of 47nH but in every datasheet we can see that they use 10 µH. But as I said, it is not the value in Henry we need to know but resistance or current capability. To corectly choose the inductor we need to know the peak current flow through the inductor that the ADC will consume during a conversion.

That's what your capacitors are for.

The inductor, by nature, is going to oppose the sharp, hard nature of the pulses that would be demanded by the chip. That's fine, it's there to *block* high speed stuff coming down the line. The capacitors are there to hold up the voltage during that time. If you look closely at the app note, you will see that it specifically shows the high-current loop being from the uC pins to the capacitor, and no farther.

For that reason (the inductance impeding the fast-rising, high-current bursts), a low inductance in the capacitor(s) is critical, to some extent more important than the actual capacitance.

As another mentioned, the needs of an LDO in terms of a capacitor are usually quite different than a microprocessor, as it usually affects the compensation loop of the LDO. Such is not the case on a microprocessor, where you simply need to hold up a voltage.

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

If we startwith the inductor to the AVCC. It is correct that AVCC supply not only the ADC but also any I/O that uses the same port. But for this discussion I skip that. The purpose for this duscussion is primary to find out the best filtering for the usage of the ADC.

I see several design steps here.
0. The condition for all this is that we need maximum performance of the ADC.
1. The filters purpose is to stop power line noise to reach the AVCC pin.
2. We want a filter that use small and cheap components.
3. The filter should stop noise but it should not limit the performance of the ADC by supplying to little current into the AVCC pin.

1-2: An L-C filter as suggested by Atmel may be the best filter in most aspects.
3: The first parameter I would design for is the maximum tolerable voltage drop into the AVCC pin during a conversion. That would be set by the peak current during conversion and we do not know anything about that. Most important parameter for the voltage drop would be capacitor.
To keep cost low I would not choose a larger capacitor that 1µF. There are capacitors with less ESR than ceramic but not for the small sizes and cost of a ceramic capacitor so a ceramic would be to prefere.

The next step in the design is the inductor. More inductance will give better filter, lower cut-off frequency. The "recommended" value spans from 47nH to 10µH so from that we can learn just nothing. Why not just choose the biggest value we can get? If I choose an inductor with high value it will usually also have high resistance. Typical value can be up to 10 ohms. So, is 10 ohm any problem? I think not so maybe this discussion is not so relevant as I thought. If the ADC port consume power at any of the pins, resistance may be very important though.

But what about the VREF capacitor?
/edit 081015
First note that the capacitor to the related LDO is only connected to the LDO's VREF voltage. It is not connected to the LDO's in- or output. Thats the reason that I thought it could be interesting.
/
Noise at the ADC reference voltage will certainly also result in noise is the converted signal. So I do not think it matters that a LDO may be even more sensitive as it use it in a loop. If the capacitor will be part of a filter together with som eresistance it would have been great to know. Also bear in mind the mentioned possible effect of leakage current and also maximum value to not effect the charge time to much.

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