AD8302 Accuracy

Hello Analog,

I have two sine wave signals, say A and B. Each has identical frequency (maybe out of phase) anywhere between 5-100MHz and amplitude over 60dB range (1:1000).

However, at any specific time, the maximum difference in amplitude between signal A and signal B is less than 14dB (5:1), and signal A is always greater than, or equal to, signal B.

 I need to determine with great accuracy the amplitude ratio, and thought of using the AD8302.

If I understand the datasheet correctly, the maximum error at 100MHz over magnitude rate of 0 to 14dB is around 0.1dB (from TPC3, page 6, datasheet).

I think this translates to about 0.3dB in gain measurement accuracy.

Firstly, is my analysis correct?

Secondly, could you recommend an alternate, more accurate way of achieving this measurement.

If I may, another question.

For the AD8306, the log linearity curves show a sort of undulating pattern across the input range.

Is this repeatable and therefore able to be calibrated out in post-processing?

Thank you.

David

Parents
  • +1
    •  Analog Employees 
    on Apr 29, 2021 3:03 AM

    Greetings David,

    Yes indeed, AD8302 does seem like it might be a good fit for this application. Working down to 5 MHz may require some bigger capacitors in certain places on the schematic. This LF usage topic has been covered in other EngineerZone threads. Please advise if help needed in this area, or consider starting a new EZ thread on LF usage if necessary. 

    Please keep in mind that AD8302 specifications require that both input channels always be within the specified amplitude range: 224uV to 224mV rms, or -73 to -13 dBV, or -60 to 0 dBm when terminated in 50 Ohms. Reference datasheet page 2. 

    About the accuracy: TPC 3 shows around 0.1 to 0.2dB in the middle of the range, increasing over temperature to approx. 0.5dB, and increasing further as signals approach the ends of each detector's usable range. Remember, the T of TPC stands for Typical! Individual results may vary. Also look at TPC 24, showing the affect of phase causing a slight additional error. 

    If one signal input is consistently higher than the other, consider an attenuator on that higher-level channel, to help bring the two channels closer together in magnitude. This way, when both signal levels go up and down together, it becomes possible to keep the response more centered in the design range of the device (-73 to -13 dBV as noted above). This is illustrated in TPC 12.

    To get better accuracy, consider averaging of the result, which could be in firmware, software, or by using the MFLT pin. For even better accuracy, two separate detectors may be necessary, of newer design with the best log-linearity performance. For the ultimate accuracy, you may need a mixed-signal solution with high-resolution ADC.

    AD8306 like many log detectors will have some ripple in the log-linearity conformance plot. The ripple peaks magnitude and location on the curve can vary with temp, frequency, input signal harmonic content, Vcc, part-to-part, maybe other factors. It's probably too much to attempt to remove with calibration, and we would not recommend it. Instead we present the graphs which state the error and advise that the application must accept the error, or find another part. But we always recommend at least a 2-point calibration to account for production variation of slope and intercept shift from part-to-part.  

    Hope that helps!    -Bruce H. 

Reply
  • +1
    •  Analog Employees 
    on Apr 29, 2021 3:03 AM

    Greetings David,

    Yes indeed, AD8302 does seem like it might be a good fit for this application. Working down to 5 MHz may require some bigger capacitors in certain places on the schematic. This LF usage topic has been covered in other EngineerZone threads. Please advise if help needed in this area, or consider starting a new EZ thread on LF usage if necessary. 

    Please keep in mind that AD8302 specifications require that both input channels always be within the specified amplitude range: 224uV to 224mV rms, or -73 to -13 dBV, or -60 to 0 dBm when terminated in 50 Ohms. Reference datasheet page 2. 

    About the accuracy: TPC 3 shows around 0.1 to 0.2dB in the middle of the range, increasing over temperature to approx. 0.5dB, and increasing further as signals approach the ends of each detector's usable range. Remember, the T of TPC stands for Typical! Individual results may vary. Also look at TPC 24, showing the affect of phase causing a slight additional error. 

    If one signal input is consistently higher than the other, consider an attenuator on that higher-level channel, to help bring the two channels closer together in magnitude. This way, when both signal levels go up and down together, it becomes possible to keep the response more centered in the design range of the device (-73 to -13 dBV as noted above). This is illustrated in TPC 12.

    To get better accuracy, consider averaging of the result, which could be in firmware, software, or by using the MFLT pin. For even better accuracy, two separate detectors may be necessary, of newer design with the best log-linearity performance. For the ultimate accuracy, you may need a mixed-signal solution with high-resolution ADC.

    AD8306 like many log detectors will have some ripple in the log-linearity conformance plot. The ripple peaks magnitude and location on the curve can vary with temp, frequency, input signal harmonic content, Vcc, part-to-part, maybe other factors. It's probably too much to attempt to remove with calibration, and we would not recommend it. Instead we present the graphs which state the error and advise that the application must accept the error, or find another part. But we always recommend at least a 2-point calibration to account for production variation of slope and intercept shift from part-to-part.  

    Hope that helps!    -Bruce H. 

Children