How to stable the PLL over wide temperature range(-40~85℃)

I want to design a PLL operating over wide temperature range:-40~85℃.

The key is to design the VCO operating over wide temperature range:-40~85℃.

I used a TCXO for VCO reference frequency. Then how to make sure VCO is operating over wide temperature range:-40~85℃.

For the VCO calibration, I calibrated the VCO at 25℃ and stored the calication configuration in the flash. When the VCO start up at 25℃, VCO could start

up normally. But if the VCO start up at -40℃ or 85℃, it could not start up correctly.

How to solve the problem? Could anyone give me any suggestion?

Thank you very much in advance!

What kind of VCO ? Integrated or Discrete ?

Hi,

My thoughts:

* you don′t need to stabilize PLL
* you don′t need to stabilize VCO. It′s frequency is stabilized by the loop.
* you just need to stabilize TCXO.

I wonder why you talk about VCO calibration. It simply is not necessary. Or can you tell us why?

You just need to ensure that it starts up at all, but the frequency is of minor interest.
If there is a start up problem of the VCO, then show us the complete VCO circuit, with all necessary informations. PCB layout, too.


Klaus

Integrated, not discrete.



The VCO I used is RFFC2071.

Yes, I should not focus on the stablity of the PLL which has a closed loop to fit the frequency.

Because, I want the VCO to operated at one exact frequency point, I should give the configuration to the VCO.

The configuration is calibrated at 25℃. Then when I start the VCO at -40℃, the VCO could not start up correctly with the configuration

calibrated at 25℃.

Hi,

going through the datasheet I found folowing:
"If the device is left in the enabled state for long periods, it is recommended that VCO auto-selection and coarse tuning (band
selection) is performed for every 30°C change in temperature. The lock detect flag can be used to indicate when to perform
the VCO calibration, it shows that the VCO tuning voltage has drifted significantly with changing temperature."

Is this the problem you are talking about?

If this can′t be done automatically, then I recommend to use a temperature sensor and calibrate the VCO at different temperatures.
Then store the calibration values in a table.

Klaus

For an Integrated VCO, -40..+85 C working temperature needs a specialized process such as military grade or harsh industrial grade.
I don't think the standard processes ( TSMC, UMC etc.) are capable to work consistently between these temperature limits.Especially the process temperature characteristic is extremely important for such circuits as VCO.
In fact, if the VCO shows up consistent negative resistance under ALL temperature and process variations, it's necessary but not enough condition.Nobody guarantees that the VCO will absolutely work in ANY case.The process should be proven with thousands measurements otherwise 99 circuit works, one fails..

Yes, I have read this. This method is used when the device has started up, and when the temperature changes, we need do the VCO auto-selection

and coarse tuning.

But when I start the device at -40℃, the file dose not tell us. So maybe we should use the method as you discussed, using a temperature sensor and

calibrate the VCO at different temperatures. Then store the calibration values in a table.



I think this problem has been solved by the industry. For the motor grade transceiver, the components could operated

at the temperature range of -40~85℃.

I think they have added some circuits to ascertain the device to operate at the temperature range of -40~85℃ correctly.

But I don't know the exact method.

But you shouldn't forget one importing point.
"may operate" does not mean "will operate", instead it means the component may resist to these limits without any fail.
But the case of VCO is absolutely sensitive in terms of process variations and surely temperature.The components which have been used in VCO has to maintain their typical characteristics as long as temperature variations.
I remember that I designed a VCO with 5 calibration step to maintain the drift by temperature.

VCO startup is a different problem than VCO stability.
If you want dead-reliable starting then you probably
have to add functions like a sweeper or something, that
looks at lock detect and if no good, force-pull the VCO
across the tune range so the PLL can lock as it travels
past. An explicit startup circuit of some sort.

Hi,

I agree, something like this is important for a self running VCO.
But it is not necessary in a feedbacked VCO of a PLL.

If the feedback loop is designed correctely... it just needs that the VCO is running.

Klaus