LMX2592: Calibrating Issue

Hello Mario,

Can you confirm / take a look at the following:

• For 20 MHz < PFD < 100 MHz, R0[8:5] should be set to 4'b0000.
• What are R1 and R4 set to?
• What is your SPI clock rate, and how soon after changing the frequency are you writing R0[3] = 1 to calibrate the VCO?

On a side note, what is your application and what frequency to what frequency are you converting from?

Thanks,

Vibhu

Hello Vibhu,

thank you for your reply.

1. Yes, REG 0 is set to 0x221C which means R0[8:5] set to 'b0000. By the way I do not understand the digit '4' you put after "set to".
2. Registers are set to:
   
   1. R1 = 0x0808
   2. R4 = 0x1943
3. SPI clock frequency is 4.1 MHz.
4. As far as the delay of writing R0[3] I have tried many different values ranging from a few hundreds us up to seconds.
5. The application is an upconversion (which is eventually followed by an additional downconversion). The input frequency is 20 MHz to 4 GHz which is converted fo 5.2 GHz.

For you reference here are the fulI set of register (the frequency is set to 5430 MHz)

R64    0x400077
R63    0x3F00A8
R62    0x3E0000
R61    0x3D0001
R59    0x3B0000
R48    0x3003FD
R47    0x2F08C0
R46    0x2E1EA3
R45    0x2DC600
R44    0x2C053E
R43    0x2B0000
R42    0x2A0000
R41    0x296500
R40    0x281DCD
R39    0x278204
R38    0x260044
R37    0x254000
R36    0x240048
R35    0x23119F
R34    0x22C3EA
R33    0x212A0A
R32    0x20210A
R31    0x1F0001
R30    0x1E0034
R29    0x1D0084
R28    0x1C2924
R25    0x190000
R24    0x180509
R23    0x178842
R22    0x162300
R20    0x14012C
R19    0x130965
R14    0x0E0AD4
R13    0x0D4000
R12    0x0C7002
R11    0x0B0018
R10    0x0A10D8
R9    0x090302
R8    0x081084
R7    0x0728B2
R4    0x041943
R2    0x020500
R1    0x010808
R0    0x00221C

Hello,

The "4" in "4'b0000" just to indicate 4 bits. Sorry if that confused you.

Do you see the same issue when using a different input frequency?

You can also try changing the state machine clock R1[2:0] and VCO calibration delay R4[15:8]. The register descriptions in the datasheet provide descriptions on what these bits do.

Please also take a look at http://www.ti.com/lit/an/snaa336/snaa336.pdf

Thanks,

Vibhu

Hello Vibhu,

thank you for your help. We have tried different settings according to your hints.

As for the first question, do you mean different REF in freq or different PFD? We have tried a different PFD with no positive results. Changing the REF in Frequency is not strightforward beacause the oscillator is used by other parts.

As for register changes, here is the report. Changing R4[15:8] does not produce any difference in the phenomenon. Instead, the R1[2:0] makes the behavior of spurs different. It is enough to change it from "b1000" to any other value to stop the "toggling" spur every time a calibrate command is issued. In other words, when  R1[2:0] is "b1000", as we have it now, for almost every calibrate command there is a toggle in spur (if there weren't any spur they appear while if spurs are present they disappear). Conversely, with R1[2:0] other than "b1000", for instance "b1001", the spur "status" remain unchanged after any calibrate command. So far, so good. However, this also means that if spurs were present they will be present forever and unfortunately this is what sometimes happens. In other words the improvement is quite big but it still remains the fact that if at start-up the two spurs appear they will last until the next switch-off.

I hope my description has been clear enough to understand all I said. Of course do not hesitate to ask me any further elucidation you may need.

Thanks a lot for your efforts in helping me.

Best regards,

Mario

Hello Mario,

Thanks for the detailed response. I meant to test with different input frequency to see if there is a spur on the input at the same frequency as well that is being carried over, just for debugging purposes. If this is possible please try this, or perhaps measure the performance of the input to see any similarities. Based on your description however, I believe this is likely not the case.

From what you say the R1[2:0] control makes the spurs behavior repeatable? ie You see the same spurs at the same places for different settings but this time re-calibration doesn't make them disappear. Were these spurs smaller than what you saw earlier?

Also after you adjust R1[2:0] make sure you also adjust R4[15:8] if needed, as these settings are related, the information regarding this will be on the datasheet.

Thanks,

Vibhu

Hello Vibhu,

thanks again for your support. As for changing the input frequency, it is not easy as the reference is the same for many other circuits, indispensable for checking spurs themselves, thus it would mean some HW work, ie cutting traces.

Yes, by changing R1[2:0] I can make it repeatable: with R1[2:0] = "b000" spurs appear/disappear almost at every re-calibrating. Setting R1[2:0] = "b001" spurs, if present do not change or do not appear if not present. Spurs, when present, never change: neither frequency, nor amplitude.

As far as R4[15:8], we have tried in many ways but it does not affect the behavior at all.

I saw that the article you pointed out is related to LMX2594 rather than LMX2592. As we are considering passing to LMX2594, do you think that that device, perhaps, could not suffer this problem?

 Best regards,

Mario

Hello Mario,

It seems to me that the calibration was the issue, based on the behavior you described. Does the 5 kHz offset spur disappear in both cases (no spurs and permanent spurs)? If so that confirms that the issue was calibration realted.

The article applied to both, or to be more accurate the LMX2592 has a similar calibration procedure as the document.

Thanks,

Vibhu

Hi Vibhu,

no doubt it is related to calibration since my first post. In my view, it seems that something in the chip keep switched to a signal instead of being released. Sadly, I understand from your reply you have no hint anymore to come up with solution.

Regards,

Mario

Hello Mario,

I took a bench measurement using the configuration you provided and actually, did not see any issues with calibration or any 5 kHz spurs. Keep in mind you can get better performance with a cleaner source. (I used the LMK61E2)

Also I noticed that the configuration is for a 5340 MHz output and not a 5430 MHz output, can you confirm?

If you are using TICSPro it might be worth a shot starting from the TICS Pro default and configuring the part as needed.

Can you share what you see on your end?

Thanks,

Vibhu