AD8345 I/Q Modulator Instability

We have three manual phase shifter modules as lab equipment. These were an in-house development using I/Q modulator AD8345.  The I/Q inputs are DC levels from dual TxDAC AD9761 with 0.1% resistors; the programming is from a sin/cosine table and is fully static.  The board is 1.6mm FR4, nothing special and the RF  layout is reasonable but could be improved. 

The units function OK, but are troubled with an instability.  This appears as a ~130MHz oscillation but only at particular angle settings, maybe with full gain on the Q channel and zero gain on the I channel.  Monitoring is with a scope RF probe connected to a resistor attenuator directly after the AD8345.The oscillation remains if the RF LO is replaced with an RF short.  I have experimented with changing the supply decoupling capacitors, adding series R, changing the RF capacitors, grounding and so on, but the oscillation frequency doesn't seem affected.  The frequency can be tuned by dropping the supply voltage or sometimes by changing the I/Q angle a few degrees.  In conclusion, it seems the oscillation frequency is set by the IC, not so much by the external circuit.

In the worse case, we could make a new PCB using the ADL5370 device, but if we are doing something wrong, maybe this doesn't help.

Any suggestions?

  • Problem found.

    The gain on I and Q is sufficient to let the device oscillate internally if the applied DC differential voltage exceeds +/-0.3V on one pin.  More precisely, on one device, the level for oscillation to start was +/-0.303V on one pin on the Q channel.

    Perhaps, the data sheet could be revised to show this, for example writing the maximum modulation level with more resolution as e.g. +/-0.290V on one pin.

    On the plus side, the device was exceptionally tolerant to removal of power supply decoupling capacitors, absence of RF signal, etc.

  • 0
    •  Analog Employees 
    on Mar 2, 2013 2:51 AM over 7 years ago

    It sounded there is a spur around 130MHz( suspect oscillation internally) at specific condition( constant DC(above +/- 0.3V centered at 0.7V) at IQ inputs.

    Harmonic performance is degraded by increasing  baseband input level as expected(refer Fig 16,17 in datasheet). However, there are no spurs,which are suspected as a oscillation spur, rather than harmonics even with 0.7V +/- 0.5V on each baseband inputs(so 2Vp_p differential on I and Q. I used 1MHz CW for baseband inputs(I+/I-/Q+/Q-) biased to 0.7V, set LO (@ 140MHz, 800MHz) at 0dBm LO, set Vs at Vs=5V, 2.7V andI checked it with spectrum analyzer.

    could you specify a test condition in detail (Vs,LO Level, I+/I-/Q+/Q- input level) with screen shot, if possible? It will help me to understand the issue and to repeat it here?

  • Thank you for the interest in this.  Please see the attached.

     

    The oscillation is occuring when the differential voltage is 0.8V, outside the data sheet value of 0.6V.  Sorry, this was my error. When the differential is 0.6V, the device seems stable.  I hope to confirm this with a phase noise measurement later this week.

     

    The oscillation occurred only at precise voltages.  Perhaps when the baseband signal is 1MHz and not DC, there is no time for an oscillation to start.

     

    Would you have a comment about the suitability of AD modulators for this service, considering e.g. baseband 1/f noise?

     

    Many thanks

    Chris Gough

  • To finish this story, confirmation that the instability is gone:

    The attached is a phase noise plot for the AD8345 Modulator IC after the I and Q control voltages were reduced to +/-0.270V per pin.  The plot is generated by first measuring the absolute noise of the Agilent 8357D source, then subtracting the phase noise measurement with the AD8345 connected.  So a flat line of 0dB means that the circuit does not add phase noise or show instability.  The "grass" at lower frequencies is measurement noise only.  The noise level reduces above 50MHz separation from the carrier because the circuit has an output bandpass filter.

    The AD8345 circuit then adds ~3dB phase noise above 500kHz for 0dBm input and n*90 degrees modulation angle, slightly less when the modulation angle is 45 +n*90 degrees.  As the RF input level decreases, the noise increases.

  • 0
    •  Analog Employees 
    on Mar 22, 2013 1:43 AM over 7 years ago

    It is good to hear AD8345 showed stable. I have not had a chance to check performance(stability as well) at that condition where baseband inputs are DC driving. There are saturation at Pout with around 1.6Vp_p differential(0.7V+/- 0.4V at each BB input) at I and Q. So the stability issue at high differential dc voltage driving may be related to this. It is needed to check.

    By the way, 0.7V common mode voltage is recommended for AD8345. But it looks it is 0.8V at your data. I plotted it below for your information.