Synchronization of multiple nRF24L01 transmitters

[Francesco Bernardini]

Good afternoon,

For a beamforming application in the Wi-Fi range, I am planning on using an nRF24L01 to generate the signal.

The signal would later be split into N different signals, and individually attenuated/phase shifted by means of other chips.

As an alternate design concept, I was also conjecturing if I could eliminate the signal split, and use directly N units of nRF24L01, synchronized to generate the same signal and directly connected to the phase shifting chips.

The application is at 2.4GHz, so the N signals should be synchronized to a precision of the order of the picosecond (~1degree of phase shift), or at least the same phase precision of the combined signal split apparatus, which is ~4degrees.

My question is: has anyone ever attempted something similar, or has an idea if said precision is attainable with multiple nRF24L01, and possibly some additional device to help the synchronization?

Thank you in advance and

Best regards

Francesco

[ketiljo]

You definitivly need to supply a common clock source to all nRF24L01's. The internal logic is based on a state machine so you should in theory be able to syncronize all radios this way. Note that the clock source needs to be fairly clean, e.g based in a crystal oscillator. 

[Francesco Bernardini]

Good morning,

Thanks a lot for your answer.

I guess that means that the best synchronization achievable depends on the supported crystal frequency, which for the nRF24L01 is 16MHz -> 62.5ns.

That would be large compared to the 0.41ns of 2.4GHz, but faster crystals such as ECX_L.pdf (ecsxtal.com) (up to 1.5GHz) won't work with the nRF24L01 and I'd have to find something else.

Are there ways, or specific chips, to synchronize the signals after they are generated, recovering what's lost in the crystal input phase?

I am thinking about the twin pendulum principle, considering the N signals by the different nRF24 chips are theoretically identical, but I am just speculating 

[ketiljo]

It only works with a 16 MHz crystal. If all radios are fed with the same 16 MHz souce and the PCC tracks have the same length, you will have very little timing difference since the internal logic should all be the same. 

[Francesco Bernardini]

Understood, thanks a lot for your input!

[Francesco Bernardini]

Good morning again Ketiljo,

although I closed the ticket, this morning I woke up with this other doubt: 

yes, I can use the same clock to sync multiple nRF24L01, and put extra-care in designing the traces so that they have equal lengths.

And as you said, after that the synchronization will depend on how precise is the manufacturing process and the internal circuitry behaves, across different units.

In other words, let's assume the nRF24L01 are in working status, are perfectly in sync, and generate the exact same signal: is there any benchmark with the amount of phase difference one may expect between each pair of nRF24L01 of the set?

If such a benchmark isn't available, would it be of interest for the community if I made it myself?

If you think it's better to open a new ticket for this question, I'll do it

Thanks again and

Best regards

Francesco

[ketiljo]

Francesco Bernardini said:

In other words, let's assume the nRF24L01 are in working status, are perfectly in sync, and generate the exact same signal: is there any benchmark with the amount of phase difference one may expect between each pair of nRF24L01 of the set?

This is not something that's tested as it's way out of scope for the normal use of the radio.

[Francesco Bernardini]

Understood, thanks once again!