This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DDC112: What is manufacture valiation of output data when 22pF capacitor only is connected to IN1 terminal.

Part Number: DDC112

 Hello guys,

 One of my customers is manufacturing some products which DDC112 is used for a photo diode (PD) current sensing.

 In the manufacturing, they found that some of DDC112 (about 2%) output data showed higher current than others (about 98%) when PD is covered completely from light .

 They thought the cause might be PD manufacture variation. So they removed PD. Then DDC112 IN1 was connected to only 22pF capacitor. The capacitor opposite

terminal is connected to GND. But same DDC112 showed higher current than others too at the conndition.

 The followings are those DDC112 output.

DDC112 which outputs higher data (categolized in 2%)

output data       standard deviation
1.81527pA     0.02865
1.81748pA     0.02850
1.81800pA     0.03045

DDC112 (others- categolized in 98%)

output data       standard deviation
1.81410pA     0.02239
1.81337pA     0.02197
1.81312pA     0.02363

The followings are the result when capacitor connected to IN1 and GND was exchenged to exclude capacitor cause.

 DDC112 which outputs higher data (categolized in 2%)

output data       standard deviation
1.82209pA  0.02661
1.82182pA  0.02650
1.82303pA  0.02982

DDC112 (others- categolized in 98%)

output data       standard deviation
1.81167pA  0.02182
1.81145pA  0.02315
1.81108pA  0.02344

 The customer questions are the followings.
Q1.  Are those output data difference in normal range of DDC112 characteristics?
Q2. If Q1 answer is yes, which datasheet parameter defines the deviation range?
Q3. Is there any way to improve the deviation?
 Your reply would be much appreciated.
 Best regards,
 Kazuya Nakai. 

  • Hi Kazuya-san,

    How are you?
    Thank you for using DDC112 device.
    Yes, I will forward your question to our system engineer.
    We will reply to you in about two days.

    Thank you very much!

    Best regards,
    Chen
  • Hi Chen,

    Thank you for your reply.
    It's nice to talk with you on this E2E.

    I see. I'm looking forward to receiving your next reply.

    Thank you again and best regards,
    Kazuya Nakai.
  • Hi Kazuya-san,

    How are you?
    Sorry for your waiting.
    Our system engineer is looking at these questions.
    We will reply to you very soon.

    Thank you again!

    Best regards,
    Chen
  • Hi Kazuya-san,

    I see currents, not charge... Do you know what is their integration time? I can guess it from the data, but want to make sure things match. Once you get it in charge units, you can look at offset and ibias...

    Regards,

    Edu

  • Hi Edu,

    Thank you for your reply.
    I see. I will check it and imform you the result as soon as I get it.

    Thank you again and best regards,
    Kazuya Nakai,
  • Hi Kazuya-san,

    How are you?
    Thanks for helping the customer.
    We have one more question is:
    Could you please ask the customer what the Range was set (when they showed us the above test results)?

    Thank you very much!

    Best regards,
    Chen
  • Hi Chen, Edu,

    Thank you very much for your support.

    The customer conditions are the followings.
    1. AVDD=DVDD=+5V
    2. Only IN1 is used.(IN2 is connected to GND.)
    3. VREF=4.096V
    4. RANGE=1
    5. CLK=10MHz
    6. CONV clock L level width= 100ms
    7. CONV clock H level width= 100ms
    8. Conversion mode=Continuous

    Could I have any your comment base on those information?
    Also if you need more information, please let me know.

    Thank you again and best regards,
    Kazuya Nakai.
  • Hi Kazuya-san,

    Thank you for the conditions. Do you know which one is their exact part number? I mean, U/Y version or UK/YK? It is a bit different in the spec.

    Also, can you ask them what do they mean for output data and standard deviation. I.e., for instance, is 1.81527pA the output value of one DDC112 channel and 0.02865 (pArms?) the noise for that channel? I would think so, but just making sure...

    So, they basically listed 3 channels that look similar to others and 3 that look off, correct?

    Looking at the difference between what they call a normal channel (in average 1.81353) and an outlier (in average 1.816916), it looks like it is 0.00338pA which would translate into 7 codes. That looks really small to me... Are we reading this right?

    And how do they compute those currents? As you know, the device measures charge, so, I want to know if they are subtracting or not the 0.4% offset (output 4095) on their calculation. I.e., 1.81527e-12*100e-3/50e-12*2^20=3807 codes. Did they get that code as output (3807) or actually they got 3807+4095=7901?

    A small issue we have looking at this data is that it is taken at Range 1 (50pC FSR) but offset error spec is given in Range 5 (250pC), so, it is difficult to know if it is within spec or not. The first number above (3807) would be pretty close to the ideal 4095 and actually would be close to typical spec (+/-200). The 2nd number (7901) is quite far. I would have to check here, but maybe the Range 1 as it is 5 times more sensitive can result on wider offset spec. No idea...

    The other important thing to take into account with those long integration times (100ms) they need to watch for ibias which has a typical of 0.1pA but a max of 10pA. I suspect that this may be because it is not measured in production and they just placed a wide margin, so, I would not expect it to be like that for most of the units, but we have seen outliers in this kind of spec before in a different DDC (not DDC112). So, if that ibias varies from channel to channel by 0.1pA, it is already bigger than what they call "outlier".

    To see if ibias is the culprit, they could change the integration time and see how it varies the output. Offset does not change with integration time, while the effect of ibias gets bigger with integration time...

    Also to note that if it was ibias, the effect would get worse with temperature (roughly 2x increase for every 10C T increase)

    Nevertheless, I am thinking we still missing something as again, we are talking only about 7 codes difference in average which looks very small... So, I'll stop speculating here and wait for your answers... May also have done some math error but please feel free to double check these numbers. 

    Best regards,
    Eduardo

  • Hi Edu,

    They are using DDC112U (DW package).
    They get the output data every 0.1s(100ms) for 20 seconds. So total data count is 200. 1.81527pA is average value of those data. Also the standard deviation is calulated from those data.

    "So, they basically listed 3 channels that look similar to others and 3 that look off, correct?"
    -> Correct.

    "Looking at the difference between what they call a normal channel (in average 1.81353) and an outlier (in average 1.816916), it looks like it is 0.00338pA which would translate into 7 codes. That looks really small to me... Are we reading this right?"
    ->I think so and the customer beleives the data is correct. They want to know what is this phenomenon (data averages can be categolized clearly to two groups as 98% lower data grop and 2% higher data grop) cause. Do you have any answer?

    "And how do they compute those currents? As you know, the device measures charge, so, I want to know if they are subtracting or not the 0.4% offset (output 4095) on their calculation. I.e., 1.81527e-12*100e-3/50e-12*2^20=3807 codes. Did they get that code as output (3807) or actually they got 3807+4095=7901?"
    -> I will ask them this.

    "The other important thing to take into account with those long integration times (100ms) they need to watch for ibias which has a typical of 0.1pA but a max of 10pA. I suspect that this may be because it is not measured in production and they just placed a wide margin, so, I would not expect it to be like that for most of the units, but we have seen outliers in this kind of spec before in a different DDC (not DDC112). So, if that ibias varies from channel to channel by 0.1pA, it is already bigger than what they call "outlier".
    -> If the customer tries to measure the output data with other integration time, what integration time is better to try?
    50ms? 20ms? 1ms? Could you please recommend me which time is better? I will ask them to measure the output data when I get your recommend time.

    Thank you very much again and best regards,
    Kazuya Nakai.
  • Hi Kazuya-san,

    Can they provide a graph of the distribution (a histogram of offsets)? If I understand right, they are saying that they got a Gaussian distribution of channels, all club together very tight and then got very few channels, again, very tight, 7 codes away. I say very tight because otherwise wouldn't be able to see the two Gaussians separated only 7 codes apart. This would actually be much better than any spec we got (we talk about 100's of codes between parts), so, I think we are missing something on their explanation.

    To remove the effect of the ibias they can use much shorter integration time. Say 1ms. Whatever difference they see in offset between a reading say at 100ms and one at 1ms is due to the ibias integrated over those extra 99ms.

    Regards,

    Edu

  • Hi Edu,

    Thank you for your strong supports.

    I see. I will try to get their row output data or graph you told and I will reply as soon as I get it.

    Also I will ask them to get the output data again with 1ms intgration time. To make sure, could you tell me
    what CONV H/L width are?
    1. CONV clock L level width= ??ms
    2. CONV clock H level width= ??ms

    Thank you again and best regards,
    Kazuya Nakai.
  • No problem. Just 1ms each is good (CONVL=CONVH=1ms). It doesn't really matter as long as it is short and we know the number.

    Regards,
    Eduardo
  • Hi Kazuya-san,

    We were discussing internally and we think it is better to move this discussion off-line so that I can close the thread... Please write us to ddcxxx-support@list.ti.com

    Best regards,

    Eduardo

  • Hello Edu,

    I will talk with you off line. So I will close this thread.

    Thank you and best regards,
    Kazuya Nakai.