Post Go back to editing

Precision Opamp Selection for AAF (2nd order Sallen-key)- 24 channels 16 bit DAS

Hi,

I'm working on 16 bit 24 channels DAS. Here are the specs;


Architecture : Single ADC multiple channels. In each channel one AAF is used.
Resolution : 16 bit
Signal swing : +/- 10V (20V)
Input signal frequency : 200Hz max
Number of channels : 24
Total time available for 24 channels for conversion : 800us [ each channel 33us approx]
ADC sampling frequency : 4 MSPS
Line : Differential
Required 3-dB point for AAF : 200 Hz
Order of the filter : 2
Attenuation at 75Khz : -96dB (min)
Settling time: < 2us [ After every selection of channels - It seems to be small when compare to total available time period for each channel (33us). It is because, to increase the per channel sample rate by taking multiple samples of a single channel. 30us is used for sampling each channels.]
AAF architecture: Sallen key
Input allowable offset : <1.2mV
Supply : dual +/-15(min)

24 channels are connected to ADC via MUX(ADG1207). So maximum transient loading occurs during the switching of channels which are connected to max and min value of input signal. Since Sallen-key architecture is used, there may be a need of additional RC filter at the output of AAF opamp.

Can anybody suggest an opamp for this particular application.

Thanks,

Jebas.

Parents
  • Hi Oliver, Harry & Matt,

    Thank you very much for your responses.

    @Oliver

    Thanks for your comment on input bias current and the selection of Opamps. Unfortunately, the opamps mentioned by you having either higher input OFFSET and input offset current.

    FYI, I can afford only 1.2 mV of offset at the input of AAF.

    @Harry

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    "Every year ADC prices come down, so I would consider  using three ADCs or more.  Although the ADG1207 is a low capacitance mux, it still has finite on resistance and capacitance.  To get to 16 bits of accuracy, you have to allow 11 time constants to settle."

     

    I accept your point. But, RC time constant is not seems to be a big problem in my application. From the datasheet of ADG1207 the switch ON resistance(max) is around 250 ohm and the capacitance at the output of MUX is 100pF[Inclusive of three (24 channels => 3*8:1) ADG1207 output capacitance + next stage amplifier input capacitance].

                             one time constant = 250 * 100 pF = 25 ns

                              11  time constant = 11 * 25 ns = 275 ns [ Which is very much less compare to available time per channel as specified in my 1st post]

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    "96 dB at 75 kHz?  With a second order filter, you only get to ~52 dB."

    I accept this too. Only with 2nd order Sallen key filter -96dB of attenuation is not possible. It is because of finite output resistance of opamp and its dependency on frequency. The maximum stop band attenuation is given by the ratio of R1[with ref to attached image] and opamp out resistance.Say, the output impedance of opamp at unity feedback is 100 ohm and the filter resistance used is 100 Kohm (R1). The ratio between this two is 1E-03 i.e -60 dB.

    That is why, it is usual to provide an additional RC filter at the output of Opamp. I have attached [Analog_Devices_query_1.pdf] the frequency behavior of filter with and without the RC filter.

    But the series resistance used in RC filter can't be greater than 100 ohm. It is limited by the design. Because of this limitation I am forced to use bigger Capacitance.

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    " With perfect muxes, op amps, and ADCs, you still have to worry about pc board layout.  Real easy to get crosstalk and wind up with a 10 bit system.  I would strongly recommend singles or duals, not quads, to avoid routing problems and crosstalk.  I would start with a minimum of six layer board."

    Thanks for your comments. I too considered your point. But I compromised myself with the following points.

    1. Signal frequency is only 200Hz maximum

    2. Coupling capacitance (Parasitic) between each channel will be <1pF (rough approx, if wrong please correct me). At 200Hz its impedance is 796     Mohm. 

    3. It is a differential signal processing

    Please give your comments on this. I am more interested on this issue.

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    @Matt

    As I mentioned in above paragraph, I am using an additional RC filter to discard the issue mentioned by you. Please give your comments on this.

    Thanks,

    Jebas

Reply
  • Hi Oliver, Harry & Matt,

    Thank you very much for your responses.

    @Oliver

    Thanks for your comment on input bias current and the selection of Opamps. Unfortunately, the opamps mentioned by you having either higher input OFFSET and input offset current.

    FYI, I can afford only 1.2 mV of offset at the input of AAF.

    @Harry

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    "Every year ADC prices come down, so I would consider  using three ADCs or more.  Although the ADG1207 is a low capacitance mux, it still has finite on resistance and capacitance.  To get to 16 bits of accuracy, you have to allow 11 time constants to settle."

     

    I accept your point. But, RC time constant is not seems to be a big problem in my application. From the datasheet of ADG1207 the switch ON resistance(max) is around 250 ohm and the capacitance at the output of MUX is 100pF[Inclusive of three (24 channels => 3*8:1) ADG1207 output capacitance + next stage amplifier input capacitance].

                             one time constant = 250 * 100 pF = 25 ns

                              11  time constant = 11 * 25 ns = 275 ns [ Which is very much less compare to available time per channel as specified in my 1st post]

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    "96 dB at 75 kHz?  With a second order filter, you only get to ~52 dB."

    I accept this too. Only with 2nd order Sallen key filter -96dB of attenuation is not possible. It is because of finite output resistance of opamp and its dependency on frequency. The maximum stop band attenuation is given by the ratio of R1[with ref to attached image] and opamp out resistance.Say, the output impedance of opamp at unity feedback is 100 ohm and the filter resistance used is 100 Kohm (R1). The ratio between this two is 1E-03 i.e -60 dB.

    That is why, it is usual to provide an additional RC filter at the output of Opamp. I have attached [Analog_Devices_query_1.pdf] the frequency behavior of filter with and without the RC filter.

    But the series resistance used in RC filter can't be greater than 100 ohm. It is limited by the design. Because of this limitation I am forced to use bigger Capacitance.

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    " With perfect muxes, op amps, and ADCs, you still have to worry about pc board layout.  Real easy to get crosstalk and wind up with a 10 bit system.  I would strongly recommend singles or duals, not quads, to avoid routing problems and crosstalk.  I would start with a minimum of six layer board."

    Thanks for your comments. I too considered your point. But I compromised myself with the following points.

    1. Signal frequency is only 200Hz maximum

    2. Coupling capacitance (Parasitic) between each channel will be <1pF (rough approx, if wrong please correct me). At 200Hz its impedance is 796     Mohm. 

    3. It is a differential signal processing

    Please give your comments on this. I am more interested on this issue.

    -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

    @Matt

    As I mentioned in above paragraph, I am using an additional RC filter to discard the issue mentioned by you. Please give your comments on this.

    Thanks,

    Jebas

Children
No Data