AD7682 - reference drive

Here is the starting point of our circuit:

For AD7682, we need to use an external reference voltage as we have a non-standard voltage (~1V).  So we would like to use an op-amp to drive the AD7682’s reference. It appears – if I read correctly, that there are two ways of supplying an external reference:

Driving REFIN with an external reference like illustrated in figure 32 from the datasheet:

Driving the REF pin with an external reference like illustrated in figure 36 from the datasheet.

Both methods seem to present similar problems to the op-amp driver: It has to be heavily decoupled with from 10uF to 100nF. This is a challenge for most op-amps as instability can occur. Incidentally the circuit in Figure 36 is likely to lack a snubber network to stabilize it – at least according to the datasheet for AD8605 (page 14 fig. 40). Anyway,

  1. What are the advantages and disadvantages of driving REF versus driving REFIN?
    I suppose the drift is better when not using the internal buffer (Driving REF), but how much, appears not to be clearly documented.
  2. Can you suggest a suitable opamp buffer which is low power (<50uA and <2.2V,  AD8605 is >2.7V)?
  3. Driving REFIN: Is it ok to do with just a few hundred pF of decoupling and avoid any snubber complexity (as opamp will be inherently stable with low capacity)  or is it better to go with 100nF and add sufficient snubber circuit to stabilize the circuit? Problem with the latter is that most op-amps are not spec’ed for this type of operation – not even AD8605! Adding a series resistor to the opamp output will deteriorate the output impedance.

Managing of reference power: About power consumption, the datasheet says on page 22

But unfortunately Table 10 on page 25 is unclear to me:

     4) What combination of Bit5..3 disable the internal buffer? In other terms what is the CFG[5..3] setup suggested to operate in modes: a) Driving REF (fig. 36) and b) Driving REFIN (fig. 32) respectively?

I hope you can guide us with at least some of the points in this rather lengthy question.

br henning

  • 0
    •  Analog Employees 
    on Apr 14, 2014 8:00 PM over 6 years ago

    Hello,

     

    If you provide an external reference voltage to REF, we recommend that you use a buffer such as AD8031 or AD8605. If you provide an external reference voltage to REFIN pin, you don't need to use an external buffer.

    In either case, the REF pin must be decoupled with a 10 μF (X5R, 1206 size) ceramic chip capacitor as recommended in the datasheet. The 0.1 μF decoupling cap is enough for the REFIN pin. The REF pin has a dynamic input impedance and should be decoupled with minimal parasitic inductances. This is done by placing the reference decoupling ceramic capacitor as close to REF as possible, ideally right up against, the REF and GND pins and connecting them with wide, low impedance traces. Make sure that they follow the reference decoupling and layout guideline from the datasheet.

    Since you are using an external reference option (either REF or REFIN) anyways, you will need to set REF[5:3] bits as "01x" or "11x" depending on your temp sensor requirement.

    Regards,
    Maithil

  • Hello Maithil,

    Many thanks for your prompt reply.

    Unfortunately neither AD8031, nor AD8605 are working at 2.2V.

    1) Do I correctly understand that when using the REFIN, we can use either one of the two methods illustrated below?

    2) If this is the case, can you tell what is the REFIN input current and in particular the input current _variation_with temperature? This information is necessary for the design of R1 and R2 in case A.

    thanks for your assistance.

    br henning

  • 0
    •  Analog Employees 
    on Apr 16, 2014 12:52 AM over 6 years ago

    Hello Henning,

    Your first interpretation is correct as to what you explained.

     

    The REFIN is the output from the internal band gap. The data sheet states that the typical reference current drain at 250 kSPS and 5V is 50 uA.  The current drain scales with input voltage, so below 5V, it will be likely less than 50 uA.  Unfortunately, we don't have the data for this over temperature.

     

    The rule of thumb is that to avoid conversion errors, the average current required at a particular throughput should not cause the reference voltage to drop by more than 1/2 of a LSB. 

      

    Regards,

    Maithil