LMX2572: About output frequency

Hi Adachi-san,

Did you mean the customer can get it locked to 812MHz, 1850MHz and 1038MHz but the performance is bad?

Which performance is not good?

One big problem with the configuration is the fpd is just 1MHz. This is not an expected use case. To get the best phase noise from the device, the loop filter bandwidth should be sufficient wide. In order to achieve this goal, fpd should be high enough to support wide loop filter bandwidth and to reduce N-divider value. 

another issue with the configuration is the following settings did not match with the fpd and fosc, this will result in VCO not being calibrated properly. 

If you can get 1038MHz but also see 4152MHz, this is normal as 4152MHz is a harmonics of 1038MHz.

Hello Noel-san,

The customer says that only 1038MHz cannot be output.

Do you have any ideas?

And I tell them about fpd.

Regards,

Adachi

Hi Adachi-san,

So the customer can see 4152MHz signal but no 1038MHz signal? If the customer set the divider to div/2, will they see 2076MHz signal?

Hi Noel-san,

They can see 812MHz and 1850MHz, but not 1038MHz. They plan to output about a dozen frequencies, and 812MHz and 1850MHz are the frequencies before and after 1038MHz. And, they said VCO leakage (4152MHz) has been confirmed under the above conditions (set to RFoutA: 1038MHz).

Regards,

Adachi

Hi Adachi-san,

That's why I asked. if they can see 4152MHz, that means the VCO has been locked to this frequency. They said there is no output with channel divider = div/4. If they set the channel divider to other value, for example, div/2, will they see 2076MHz signal?

Hello Noel-san,

The customer says that the 2076MHz signal is not output even when the channel divider is set to div/2. Can you think of any possible causes?

Regards,

Adachi

Hi Adachi-san,

I have no idea what went wrong, let's try direct VCO output to confirm the VCO is indeed locked to 4152MHz. 

Hello Noel-san,

OK. I'll tell the customer to confirm.

Regards,

Adachi

Hello Noel-san,

The customer got the waveforms of 4.152G, 2076MHz, and 1038MHz for the VCO. It still seems that 2076MHz and 1038MHz are not output.
I send you the setting screen as well.

Regards,

Adachi

Hi Adachi-san,

Can you make OUTB_MUX = 0?

Hi Noel-san,

OK. I'll tell the customer to confirm.

Regards,

Adachi

Hello Noel-san,

The customer confirmed that 1038MHz is output. However, I don't think it is desirable to have unnecessary clock output from an unused pin, is there any countermeasure?

Is it a issue with the software for the evaluation board?

Regards,

Adachi

Hello Noel-san,

 In this case, is there any way to not output OUTB?

Regards,

Adachi

Hi Noel-san,

Does this setting mean that OUTB will not be output?

Regards,

Adachi

Hello Noel-san,

I understand. Thanks to the support.

Regards,

Adachi

Hello Noel-san,

I received additional questions from my customer.

He is unable to output 812MHz properly. He suspected saturation and set the output level (OUTA_PWR) to I lowered the spurious to near zero, but the spurious did not go away.

Attached image and TCS files. (812MHz)

812.tcs

When he output 800MHz, which is a close frequency for testing, he get a clean CW.

Attached image and TCS files. (800MHz)

800.tcs

Can you check to see if there is something wrong with the setup?

Regards,

Adachi

Hi Adachi-san,

800MHz configuration:

Can they change OUTB_MUX to "Channel Divider"? I am using the latest TICS Pro, "Hi-z" has been changed to "Reserved".

812MHz configuration:

I don't see issue here.

Will the problem go away if they program R0 (click the CalibrationVCO button) one more time?

Can they readback the following registers?

Overall speaking, I suspect the root cause could be a VCO calibration problem. The fpd is unusually small, do you know why they have fpd = 1MHz? is their loop filter design for this fpd?

Hello Noel-san,

About 800MHz configuration: 

Sorry, I forgot to confirm that point.

About 812MHz configuration: 

The customer calculates 1 MHz from Equation 1 in the datasheet.

Fpd = Fosc(203MHz) * Doubler (x1) * Multipler (x1) / { PreR (1) * R (203) } = 203MHz x 1 x 1 / ( 1 * 203 )= 1MHz

Could you please confirm that this calculation is correct?

Regards,

Adachi

Hi Adachi-san,

the calculation is correct. However, people rarely use 1MHz fpd. It is because the N-divider would be very big. In this case, N-divider = 12800 @ fout =800MHz. N-divider will contribute noise to the PLL. Bigger N, higher noise. 

If we want to have low overall jitter, we will use a higher fpd (smaller N) and wider loop bandwidth. 

The only use case with a small fpd is, we want a narrow loop bandwidth. 

My question is, why did the customer choose 1MHz fpd? What level of performance they are targeting for?

Hello Noel-san,

Understood. I will ask for the targeted loop bandwidth.

How can fpd be calculated from the targeted loop bandwidth?

Regards,

Adachi

Hello Noel-san,

Thanks. I understand. I will ask the customer about fpd.

Regards,

Adachi