• TI Thinks Resolved

ADS52J90: Input Overload Recovery

Prodigy 20 points

Replies: 3

Views: 49

Part Number: ADS52J90

During normal operation the ADC's analog inputs are per the data sheet:  the plus and minus inputs swing +/-0.45V, centered around VCM.  i.e., assuming Vcm = 0.8V,  the plus and minus inputs each swing between 0.35V and 1.25V.    However at the start of every image we insert a marker pulse stream into the analog signal which forces the minus input to ground and the positive input swings between ground and 1.8V.  The high and low levels are each 3 conversions long (40ns).  We are using the part in 16ch, 12-Bit mode with 75MHz conversion rate. 

These 1.8V marker pulses are within the Absolute Maximum Ratings but well outside normal operation.  However the data sheet brags of recovering in 1 Conversion Clock (5MHz overload input, 6-dBFS overload) so I'm hoping the converter will give usable data on the marker pulses and recover reasonably quickly.  We don't need high accuracy on these marker pulses. 

Can you tell me how the part will react to these marker pulses.  Assume the high and low levels are each 3 conversions wide 40ns).

Specifically, I would like to know:

1) Will this work at all, or will the converter not give useful data during the marker pulses.

2) What data code should I expect with both inputs grounded?  I expect it will be 7FF +/- some errors due to exceeding the input common mode range.

3) What data code should I expect with INP = 1.8V and INM = 0.0V?  FFF?

Thanks,  Bob

  • Bob, 

    in general it will work, how ever the exact code values may not fixed. because the waveform is not in linear region of the ADC. thus from time to time, it will vary. 

    at 1.8V and GND, ideally, you want to ADC see a 1.8Vpp signal. but this is not guarnteed ADC will treat the signal as 1Vp+0.8Vp. when VCM is 0.8V, above 1.3V, what ADC sees is not linear. Thus 1Vp may not give saturated signal as FFF in offset binary output. 

    still you can use the input impluse as a marker, but you may need to detect edge slew rate, instead of using fixed values.  our ADC recovery time means that after 1CLK , small signal will be still detected. 


  • In reply to Xiaochen (Sean)Xu:

    Hi Sean,

    I could use some clarification. 

    The set-up is: 16ch, 12-bit mode.  75MHz conversion rate

    Input signal:  INM is always grounded (0.0Vdc).  INP is a square wave alternating between 0.0V for 3 cycles(40nS) and 1.8V for 3 cycles

    1)  With INP=1.8V:  I assume we will throw out the first sample since it's probably recovering from both inputs being at GND.  I understand the remaining 2 samples may not be FFF.  How far below FFF do you think they could be? 

    2)  With INP=0.0V:  This worries me because though the differential input is 0.0V, the common mode is well outside the ADS52J90's common mode input range.  Here too I assume we throw out the first sample, but how far off of 7FF might the remaining 2 samples will be?



  • In reply to Bob Edwards1:


    since the device is out of the lienar operation range, I would say it should not be too far away from FFF in the DC offset format. may be within 100 codes. I would say that slew rate detection is still the best way for such a case. 

    we do not recommend to GND the INM pin. you can do a AC bypass to GND, not pure DC to GND.  

    since you alreday reached us via email, let us move the discuison to email. we can have more details on this part.