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.

LMX2594: Additional spectral content during automatic ramp

Part Number: LMX2594
Other Parts Discussed in Thread: , LMX2572, LMX2492

We're using the LMX2594 + Reference Pro + TICS Pro (1.6.10) to understand the ramp function within the device. While we can generate ramps, there appear to be frequencies outside the desired band.

Using the settings from the LMX2594 EVM Manual, figure 18 (generating a calibration-free ramp between 12.000 and 12.125GHz).

Settings on PLL page ...

And settings on ramp page ....

(Ramp is disabled) Produces 6.0GHz as expected (note that channel divider set at 2 to bring output within the range of Tek RSA306) ...

Ramp is now enable by simply checking the box on the TICS ramp page ...

Outputs ramp is generated (see below) however, the response is not flat, and there is an additional 5.25MHz below the 6.000GHz (the expected ramp extending between 6.0000 and 6.0625GHz).

1. Do these setting appear correct for a calibration-free 125MHz (at the VCO) ramp?

2. What might we be missing that could explain the plot below?

3. During VCO calibration, what should be expected at an enabled RFout?

Thanks ... Jim

  • Hi Jim,

    My coworker will get back to you by next Monday.

    Regards,
    Hao

  • Hi Jim,

    As you ramp back to 6GHz immediately after it reaches the peak, there will be overshoot in the down slope and therefore the lower frequency boundary is less than 6GHz.

    If you can do a frequency vs time measurement, you will get something similar to below plot.

  • Thanks Noel ...

    What I was attempting to create was the triangle waveform shown in the LMX2594EVM manual (SNAU210A, Figure 19) - I believe my TICS Pro settings match those of the example.

    I haven't seen mention of overshoot in the frequency elsewhere ... does this only occur when large steps are made to N (as is the case for the saw tooth you show)?  In the example I sent, DEN should only be incrementing +/- 2097. So please clarify, what conditions would cause the VCO to overshoot (or perhaps the question is, can the overshoot be quantified based on setup conditions ... eg the step presented to the PD?)

    And the original question #3 ... what is expected at RFout during VCO calibration? Our goal is to create a smooth ramp over a large range (ideally 1GHz, but most certainly a larger range than recommended for calibration-free ramps). Figure 11 of the datasheet alludes to the frequency discontinuities during a large range in the ramp, but I've been unable to find discussion on exactly what is being calibrated and what the VCO output may do during that period.

    ... Jim

  • Hi Jim,

    Your configuration has the RST bit checked and next ramp = RAMP0, as a result, it is a saw tooth ramp.

    In order to get the triangular ramp as shown in Figure 19 of the user's guide, you need to setup two ramp configurations like Figure 18. Unfortunately and I am sorry to tell that, I just found out Figure 18 is not entirely correct. The RST bit should not be checked. The correct configuration is as follows:

    Whenever we change the frequency of a PLL/synthesizer, the transient response of the loop will make the output like below:

    Overshoot always exist but its amplitude depends on loop filter design as well as the amount of frequency change. If the frequency change is small, the overshoot will be small. 

    In the saw tooth ramp, there is virtually no overshoot during ramp up because the frequency increment of each ramp is small. However, after ramping to the peak, the next ramp is over 100MHz, so the overshoot is big and obvious. 

    For linear ramp without calibration, there is no calibration, however, the linear ramp range is very limited. If a larger ramp is required, VCO calibration is needed. As such, we won't get linear ramp, we will get a staircase ramp like below. At the beginning of each step, the VCO is calibrating and so there is a big glitch.

    More detail on ramping with calibration is available in LMX2572 datasheet, section 8.1.7.2, please check it out.

  • Thanks Noel, turning off the reset bit (and also swapping RAMP0 and 1 from what I had) does fix the large step and subsequent overshoot.

    Because we need to create a large ramp delta (600MHz min) without the calibration discontinuities, we may be to the point of discounting the LMX2594. A few more questions ...

    1. If the device is periodically calibrated across the VCO range (eg every 20MHz) and the VCO registers are recorded (VCO_CAPCTRL, VCO_DACISET, ...), is it possible to then create the "stairstep ramp" updating VCO registers with the previously recorded values as opposed to running the calibration? (or in other words, during the VCO register updates will the VCO frequency stay stable?)

    2. Does TICs Pro support read-back in order to try the above idea? (I see the MUXOUT can be connected to the Reference Pro SPI, but haven't seen a command in TICs Pro for "read registers")

    3. Are there any other TI synthesizer offerings (PLL+VCO) that would allow a smooth delta-600MHz linear ramp?

    Thanks again Noel    

  • Hi Jim,

    1. Yes, you can manually write the CAPCTRL, DACISET,.... to bypass calibration, we call this full-assist. Depending on the step size per frequency change and how often you make the change (in Ramp mode, this is done once every fpd cycle), you may need a powerful processor to continuously program LMX to sweep frequencies.

    2. Yes, TICS Pro supports register readback. First of all, set MUXOUT to readback output, then in the Raw Registers page, click R110 then, on the right most column, click the Read Register button. Repeat above for R111 and R112.

    3. The secret of integrated wideband VCO is it has a bunch of switchable capacitors to tune its frequency, that is why VCO calibration is needed in order to pick the correct capacitor. For example, LMX2594, there are seven VCO cores and each core supports approximately 1GHz tuning range. However, each VCO core is actually divided into many sub-bands. So the frequency coverage of each sub-band is less than 600MHz. You can find the details in the following appnote.

    So the short answer is, no, all integrated VCO synthesizers cannot support linear tuning range of 600MHz. 

    We have LMX2492, which is a PLL only device but support frequency ramp function. If you have a VCO with 600MHz tuning range capability, you can use it together with LMX2492 to generate linear ramp.

  • That answers our questions Noel.

    Thanks much for the support ... Jim