• State TI Thinks Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 1 answer
  • Subscribers 21 subscribers
  • Views 628 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LMX2592: Spurious or output crosstalk

Charles Robidart
Prodigy 155 points
Part Number: LMX2592
Other Parts Discussed in Thread: LMX2594

Dear Sirs;  In using the LMX2592 synthesizer we are trying to control both outputs separately as our application does benefits from both outputs to being used simultaneously.  We have many measurements where setting the output level (current from the drives to the load) changes not only the output level (expected) but produces spurious products in the other output.  Symptoms include:

1) the effect seems to be output frequency dependent, but not predicable.

2) If we turn off the output not being used the spurious in the desired channel goes away.  Not a good option for us.

3) The effect is definitely related to the unused output drive setting, i.e. increasing the drive level makes the spurs higher level.

4) The effect appears asymmetric.  The A output suffers much higher spurs with the setting of the driver of the B output than vice versa.

Is there a sequence problem?  I expect some crosstalk between the outputs, but this very excessive, sometimes the spurs are as high as the desired signal.

  • 0
    TI__Mastermind 27581 points
    Charles,

    There will be crosstalk between OUTA and OUTB if the frequencies are different.

    Realize that if both outputs are tied to the same Vcc supply, then they can talk to each other through the pull-up. Theoreically, the pull-up choice should have some impact; for instance inductor pull-up should be better at high frequency.

    Datasheet lists this feedthrough about -26 dBc. If that's about what you are seeing, then I would not be shocked. For this number, I think that we just measured it on our EVM; I don't think that we experimented with different bypassing or pull-up components on the outputs.

    Regards,
    Dean
  • 0 in reply to Dean Banerjee
    Prodigy 155 points
    Dear Dean: I appreciate your very prompt response. Your response caused me to realize I did not state my problem adequately or accurately. I am using the outputs of the LMX2289 at the same frequency. Per the block diagram on page 1 of the data sheet, I am setting the multiplexers just ahead of both output buffers are set to take the inputs from the "core" mux. I believe this makes the output frequencies the same. If I examine the output frequency of "A" and I set the output driver controls on both outputs to non-zero, I see the desired signal and spurious in the output of "A". If I raise the driver control on both A and B the desired signal level on A rises and the spurious become more pronounced (higher dBC) and greater in quantity. If at this point, A is say "40", and I control "B" to "0" the spurious on A go away entirely. I do not believe this is cross talk. The spurious frequencies appear to be quite random, and large in number. Thanks for your help, Charles.
  • 0 in reply to Charles Robidart
    TI__Guru* 81580 points
    Hi Charles,

    I think it would be better if you could provide some measurement plots as well as the configuration in TICS Pro format.
  • 0 in reply to Noel Fung
    Prodigy 155 points

    Dear Mr Fung:  I am still trying to make progress, but I am still having trouble.  One very important (I think) piece of progress is that we are using the Eval PCB in parallel with our circuit.  The important piece of information is we have been able to load our register values into the evaluation PCB and the evaluation PCB does NOT show the problems.  It is gratifying to know the "code" we use operates the eval PCB properly.  From this piece of news we are launched on trying to determine what the differences are between our physical implementation and the physical implementation of the eval PCB.  We have three implementation differences we can easily spot.

    1.  The output circuits on pins 20/22 and 18/19 do not match between our circuit and the evaluation board.

    2.  The PWB stack-up and layout are different.

    3.  The power supply/ bias circuit details are different.

    Our first path of investigation is the output circuit differences.  The plots below show the result of our circuit set to 900MHz output at LMX2592 IC pins indicated for output buffer settings of "50" for both the A and B output.  (For reference the data is taken at our circuit location RX2:B2)

    The plots below show the result of our circuit for IC pins indicated for output buffer settings of "0" for both the A and B output.  Please note there is some off air ingress in the 2.4 GHz area due to local Wifi.

    As you can see when the output buffer settings are low, 0, the output spectrums are reasonable.  This difference has been seen across multiple circuits and at different settings for the output frequency.  When the configuration of the LMX2592 used to generate the about plots is applied to the evaluation PCB, the output looks the like output for settings of "0", but holds for multiple output buffer settings.

    I note the following and ask:

    1.  What is the spectrum sensitivity of the "A" output when the "B"  output is un-terminated?  By this I mean the output pull ups on both ports of A and B are all inductive (18 nH).  "A" side outputs are terminated by a 50 ohm load capacitively coupled.  But the "B" outputs have no other termination other than the pull up inductors.

    2.  On your eval PCB the A outputs (AP and AM) are pulled up with 50 ohm resistors and then AC coupled into another 50 ohm resistor, and then terminated again by any 50 ohm test equipment.  The B side (BP and BM) are pulled up by 18 nH inductors, but then are AC coupled into 6 dB "T" attenuators before going into 50 test equipment.  When we termainate all unused outputs in 50 ohms and measure each output individually (4 in total) we find about a 2 dB difference in the output levels.  I had predicted about 4 dB, without taking into account any peaking due to the inductors.  Is this about what is expected for low settings of the buffers?  Low implies <30.

    Thanks for your assistance and insights.

  • 0 in reply to Charles Robidart
    TI__Genius 9920 points
    Hi Charles,
    Your plots are showed. Could you change the image format? Or use a zip file.
    Impedance matching always help to reduce reflection and radiation.

    Regards,
    Shawn
  • 0 in reply to Charles Robidart
    TI__Mastermind 27581 points

    3348.High Frequency Open Collector Outputs.pdfCharles,

    1.  We do not have a lot of characterization data on this.  If you have an output running, but the outputs are not terminated, we typically see much higher output harmonics.  Now in terms of how much output B crossatlks to output A, this is likley PCB board specific.  When output B is running, high frequency energy couples from the OUTB pull-up direct to the supply, and this energy can couple and show up at the RFoutA supply.  We know that this coupling depends on pull-up component, power supply decoupling, and OUTx_PWR setting, but don't have much quantitative data or charges to show on this.

    2.  Theoretically, if the output buffer has infinite output impedance and you have a 50 ohm test equipment, then then an ideal inductor should have twice (6 dB) higher power than the resistor.  But then the inductor is followed by a 6 dB pad, so you might expect something similar.  That being said, the inductors might not be perfect at high frequencies and there are a lot of assumptions.  Also, the resistive pull-up has a DC voltage drop accross this that also reduces output power.   I don't know where you got your 4 dB prediction, but you measure 2 dB and this seems reasonable.

    The LMX2594 is a similar device with open collector outputs and I am attaching a detailed document that discusses these open collector outputs that should give some insight into the lMX2592 outputs as well.

    Regards,

    Dean