AD7655 relationship between internal MUX and accuracy


Our customer is measuring accuracy of AD7655 and please let me know your advice relationship between internal MUX and accuracy.

Our test resul shows  large difference  (about 50~100LSB) between channel A (INA1 and INA2) and channle B (INB1 and INB2).

So, he has verified as below .

Input: AGND

Power supply: 5V

REF: AD1582

Interface: test  using both SPI and parallel

Number of test: 500  ( test is 2pattern which remains A0 = low  or high )

Sample: 3 samples

Test  result is below

Channle A (IINA1 and INA2) : about  100LSB +- 10LSB

Channle B (INB1 and INB2) :almost 0LSB

( for your information,  

in the case of input 1.0V ,  ( output of ADA4666-2 voltage follower is connected to INA1, INA2 , INB1 and INB2) ,

channle A  = 13296LSB (averaged value for 500 time),

channelB = 13230LSB (averaged value for 500 times)

I think that the internal difference between Channel A and B is internal MUX of AD7655.

Does internal MUX affect converted result as above  ?

I would like to know if above appearance is in specifaication or AD7655 characteristcs .

And we are checking his circuit design , layout and timing diagram.

Please let me know your advice



  • 0
    •  Analog Employees 
    on Jan 8, 2016 2:39 AM over 5 years ago


    The zero error specification on this device is 0.25% of full-scale.  This corresponds to approximately 163 LSBs. As well, I was able to speak with a member of the development team. He stated "there is a large offset difference between ChA and ChB. Something like -100 vs. +50 lsb or vice-versa."

    The code difference between the multiplexer channels is to be expected.

    -- Ryan

  • Hello Ryan,

    Thank you for your answer.

    I understand that AD7655 has large offset differece between channels

    I received additional data from the customer and could you give me your advice about below last question  ?

    Below data shows large gain error.

    Below data is averaged value for 500 times test

    Input 2.5V  : difference between channels =  about 203 LSB  ( channelA:33176   channelB:33400LSB)

    Input 3V: difference between channels =  about 250 LSB ( channelA:39759   channelB:40015LSB)

    Input 4V: difference between channels =  about 400 LSB ( channelA:52687   channelB:53071LSB)

    Input 4.5V : difference between channels =  about 500 LSB  ( channelA: 59215  channelB:59727LSB)

    I think that above reason may be design around reference supply

    but  I would like to know if there large gain error difference between channels.

    Please let me know your advice.



  • 0
    •  Analog Employees 
    on Jan 8, 2016 7:21 PM over 5 years ago

    Hello Terumasa,

    The AD7655 has a gain error specification of +/- 0.5% FSR.  This corresponds to +/-327 LSB.  The worst case would be where channel A has a +327 LSB gain error and channel B has -327 LSB gain error or vice versa.  In this case, you could expect upwards of a 650 LSB difference between A and B with a full-scale input.  However, this would be worst case. A typical device should have a gain error about half this much.  This only accounts for the gain error.  One would also have to factor in the offset error to get the expected value of a given input voltage.

    With that being said, it's always a good idea to minimize the resistance on the reference node.  This node must be able to properly regulate the reference load.  Any unperceived change in the reference value will manifest as a gain error.  Given the average reference current draw of 180 uA, the output resistance of the reference source must be kept to less than 200 mOhms to properly regulate the reference current assuming a 2.5 V reference is utilized.

    -- Ryan