Differential ADC Question

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

I have a voltage to sample that will only range between about 2.5 to 3.5 volts total span. So I thought that rather than using a single ended ADC on either a 0-5V range or 0-2.56V range, it would make more sense to try using a differential ADC for increased resolution. Is there any way to physically connect one of the differential ADC inputs to the internal 2.56V ADC reference on chip? There is no mention of this capability in the datasheet, but I just thought I would ask if anyone knew if this was possible anyway. Or must I use a voltage regulator at 2.5 or 3V as one of my differential inputs to difference against the voltage being sampled? Also, the datasheet states that when using a gain of either 1x or 10x on the differential ADC that the resolution will only be 8 bits... why is this? I was trying to go to the differential ADC because I thought the 10-bit single ended ADC's was too coarse... but going to differential drops your resolution to 8-bit?@? Also, the datasheet states that there is a "2.7 - Vcc differential ADC voltage range" with no further explanation. What does this mean? It seems to contradict a later equation given for calculating the differential ADC output of:

"ADC = (Vpos - Vneg) * 512 / Vref"

which has a range of values from -512 to +512 counts... or 1024 counts of total resolution... which is 10-bit... which seems to contradict the earlier statement in the datasheet that differential ADC is only 8-bit?

It is amazing to me how confusingly written the datasheets always are... If anyone would like to help interpret these statements for me I would be very appreciative...

i.e.

1) What does the 2.7V - Vcc diff voltage range imply?
2) Is the diff resolution 8-bit or 10-bit?
3) Can I tap into the internal 2.56V ref as one of my differential inputs?

Thanks again,
James

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

Well, not all ADCs are implemented identical, so you should probably
mention which AVR you are using.

What IMHO should be possible is to use the 2.56 V internal reference
as the reference to your ADC. That way, it will be connected to AREF
where you have to add a bypass capacitor. Then, use AREF as one of
the inputs of the differential ADC (external connection).

I think you'll have to experiment a bit. Also, don't take the
2.56 V figure as set into stone: it suffers from quite a bit of
initial tolerance, so you have to factory-calibrate that reference.

Jörg Wunsch

Please don't send me PMs, use email if you want to approach me personally.

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

1) I am not shure. If you give the part name, and page in datasheet, maybe we can help you more if we get the context.

2) The differential channels will also measure with 10-bit resolution. I believe they are trying to say you can expect 8-bit accuracy. (Noise or other errors may affect the two least significant bits of the result).

3) This can be partially answered by some careful reading of the datasheet for your part. For example, from the ATmega16 datasheet figure 98 (and also the text) you can find out that when you select the internal 2.56V reference it is also output on the AREF pin for decoupling. In theory you could connect the AREF pin to a differential input pin. However, I cannot find anywhere in the datasheet any guarantee that the internal reference will maintain its accuracy, if you load it. So you will have to measure the load stability of the AREF pin yourself...

----
CPL

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

> However, I cannot find anywhere in the datasheet any guarantee that
> the internal reference will maintain its accuracy, if you load it.

The analog input resistance is specified with 100 MOhm. I don't think
this would qualify as "load", connecting your DC meter would cause
about the same "load" to the AREF pin. ;-)

Jörg Wunsch

Please don't send me PMs, use email if you want to approach me personally.

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

Okay, thanks for your comments... it's the ATMega 1280 that I'm working with. I just have a sensor whose output will only move around in the 2.5 - 3 volt range... so I was hoping to use the differential ADC to get maximum resolution on that... but then when I read the line about expecting only 8-bits of resolution with the differential ADC was kind of discouraged... I already thought 1024 counts was kind of low... but 256 is quite a bit worse. Anyway, I didn't know you could actually tap into the AREF pin for a voltage reference... I thought that pin was just to "supply" a voltage, so that's nice to know if that's the case. I'll experiment more with it and see... Thank you very much for your answers & suggestions...

Sincerely,
James