Post Go back to editing

AD9172: DC test tone (NCO only) and output power

Category: Hardware
Product Number: AD9172

Hello Engineer Zone,

We're using an HTG card implementing a DAC conversion fonction based on an AD9172 followed by TCM1-83X+ and SSMC 50-ohms RF output port, therefore very similar to the AD9172-FMC-EBZ reférence design.

We have read carefully the following posts:

We have configured the device to activate the calibration NCO (instead of the Main NCO) and use it in DC test-mode NCO-only (post 2 helped us a lot because the datasheet was not clear about it), therefore producing a digital sinewave (center frequency fc=1.8 GHz, fDAC=5GHz, 1250 Msps) directly at the input of the D/A converter, and mesuring the analog signal at the RF output (spectrum, fc+/-500MHz):

  • for a DC level = -6 dBFS = 0x3FFF (register 148/149), we mesure the test tone at -10 dBm and SFDR in the 1 GHz band at 60 dB
  • for a DC level = 0 dBFS = 0x7FFF (register 148/149), we mesure the test tone at -6 dBm and we have clearly saturation (test tone is 2 dB compressed and high-power tones appear in the spectrum)

We do not understand why we are so low in output power, reasoning below with ajustments vs. the posts mentioned previosuly.

Ideally the AD9271 should output 7 dBm at full-scale

At 1.8 GHz the internal impedance is not 100 ohms différentiel but 50 ohms (post 1), that reduces the output power to 3.5 dBm

The TCM1-83X+ is not a 2:1 (100 diff to 50 ohms s/e) but a 1:1 (50 diff to 50 ohms s/e) impedance ratio transformer with an insertion loss of 1.2 dB : we should get a bit less power 4.0-1.2 = 2.8 dBm at its output

The connection between the SSMC port and the spectrum analyser account for less than 1.5 dB loss, we thus expected to measure 1.3 dBm

We are short of -7/-9 dB, where do we go wrong?

Thank you for your support,

Eric

  • Hi  ,

    This question is acknowledged. Please give us time to check.   will respond to your concern soon.

    Best regards,
    Peevee

  • Hello Eric,

    Can you try removing the inductors L3-L6 between the dac output and the baluns? These are not required since the balun has a DC bias path to ground. At lower frequencies they drop the overall output impedance and output power on the board. On our eval board (which has these inductors) I see the same appx -6dBm output at 1.8GHz Fout using 4915MHz Fdac. You will also see slightly improved power by increasing Fdac which pushes out the sinx/x rolloff response (I see about 1.6dB increase in power going from 4915MHz to 11796MHz). As you read in the other posts, the datasheet equations are an idealized response that don't consider frequency effects or PCB losses. While 50 ohms is a good approximation of the dac outputs' internal impedance around 1.8GHz, it is not exact, and doesn't include inductive/capacitive effects which will further degrade matching with the PCB trace and balun.

  • Hello,

    thank you for this explanation, would you happen to know how many dB we should gain by removing L3 to L6?

    Eric

    Edited: damned! just measured at lower output power (-20 dBm) and got 3 dB more output at 1300 MHz (-500 MHz, -17 dBm) and 3 dB less output at 2300 MHz (+500 MHz, -23 dBm) - due to these inductors as well?

  • Hello again,

    I downloaded the S-Parameters from Murata for the inductor (S2P): Z(inductor) = 7-j280 @ 1800 MHz (0.3 pF capacitor in series with 7 ohms resistor)

    I downloaded the S-Parameters from Analog Device for the DAC0 +/- (S2P): Z(int) = 36-j15 @ 1800 MHz (5.9 pF capacitor in series with 36 ohms resistor)

       

    I downloaded the S-Parameters from Mini Circuits for the TCM (S3P): Z(diff) = 52+j5 @ 1800 MHz and Z(s/e) = 44+j12 @ 1800 MHz (that's a -0.9 dB loss in a 50-ohms system)

    I simulated the HTG circuit seen from the DAC output: the load seen by the DAC differential output is Zload = 54+j0 @ 1800 MHz

    I simulated the HTG circuit seen from the RF output: the load seen at the RF single-ended output is Zout = 29+j13 @ 1800 MHz

     Again all this leads to 0 dBm more or less in RF output, I still do not understand nor explain how come we both measure -6 dBm!

  • Hello,

    I've checked on a different eval board and am seeing -4.5dBm at 1.8GHz with Fdac=5898.24MHz with L6 and L8 still in place. I then switched to a better quality cable to my analyzer and am seeing -3.5dBm. I then doubled Fdac to 11796.48GHz to push out the sinx/x roll-off and measured -2.5dBm. The eqns do not consider the sinc rolloff -- can you increase Fdac closer to 12Gsps to improve the power output and try a shorter cable to your spectrum analyzer?

    We have a simulation of the eval board including the 9172 die model, simplified laminate model, simplified dac to balun PCB traces, TCM1-83x and SMT components, and combined EM sim of the rest of the PCB that correlates very well with this measurement:

    I don't think the model used for this simulation is available, but we do have an archive folder for Keysight ADS with a full model for the AD9081/2. I believe one could replace the S-param files in the archive with the data for the AD9172, and proceed to use/adapt the rest of the model (PCB traces, balun performance etc) as described in this app note. Let me know if you would like the ads zip file. 

  • Thank you very much for your efforts.
    Let me digest all of this, I'll check a few things as you recommended, I'll and come back.

    Eric