Conversion error for AD7657

Dear Sir/Madam,

 

As for your ADC AD7657, why the miss conversion is happen in sometimes(not every time)?

Do you have the similar case for this miss conversion in the past?

 

We are just proceeding to verify for convert error why it is sometimes happen.

The way of verify is doing conversion may time continuous (we say this "roop back") .

So we found the 1 miss conversion happen after 100 times conversion as following. But  some samples are not happen on 1M to 10M times.

Almost the data converted were no problem they were correct data.

 

Our confirmed

1)   1)  Voltage of Vref pin==Stable seems no problem

2)   2) Range pin==fixed to GND this means 4*Vref seems no problem

3)   3) Component layout external cap. of 3-REFCAPs==Put on the close IC pin seems no problem

4)   4) We just use the parallel I/F, use all 6ch.

5)   5) We caught the Unusual result as following chart.

   This means the reason is digital setting not analog reason.

6)   6)  We confirmed this trouble in our thermostatic oven 0C to 40C also, but it is brought the same result.

7)   7)  We found a couple of AD7657s for this miss conversion. Not only 1 AD7657.

8)8) What should we confirm our condition? Please give us your good advise.

8)     Thanks Kaos

input voltage Calculated value Trouble data
DEC HEX case-1 case-2 case-3
DEC HEX Voltage DEC HEX Voltage DEC HEX Voltage
10V 8191 1FFF - - - - - - - - -
9V 7372.8 1CCC 7680 1E00 9.38 - - - - - -
7.5V 6144 1800 5119 13FF 6.25 - - - - - -
5V 4096 1000 2047 07FF 2.50 - - - - - -
4V 3276.8 0CCC 3584 0E00 4.38 - - - - - -
2.5V 2048 0800 1023 03FF 1.25 3072 0C00 3.75 4096 1000 5.00
1.25V 1023 03FF 4096 1000 5.00 - - - - - -
0V 0 0000 4096 1000 5.00 -4097 2FFF -5.00      
-1.25V -1024 3C00 0 0000 0.00 - - - - - -
-2.5V -2048 3800 -1024 3C00 -1.25 0 0000 0.00 -4097 2FFF -5.00
-5V -4096 3000 -2048 3800 -2.50 - - - - - -
-7.5V -6144 2800 -5120 2C00 -6.25 0 0000 0.00      
-9V -7372.8 2333 -7681 21FF -9.38 - - - - - -
-10V -8192 2000 -8000 20C0 -9.77 - - - - - -
Comments
1) The lower 8 bit of trouble data seems to be 00 or FF.
2) The upper 8 bit of trouble data seems to be shifted 1bit as for calculated value.
3) When the input voltage is 10V or -10V, we cannot confirm trouble conversion data.