RF RANGING PERFORMANCE REDUCED FROM 90M TO 40M - NEEDED HELP

[Anirudh_THABJUL_SB]

Hello Everyone,

I am Anirudh THABJUL, an Electronic engineer working for a company called "SECOND BRIDGE", We are based on Paris.

In our applications, We are measuring distance between two 10 layer PCB's through 802.15.4 protocol. To achieve this, We are using mainly

1. IMX6 Processor (Main processor)
2. SAM4E (Secondary processor)
3. RF215 + Antennas for RF ranging section

Our issue is when we just measured distance between two PCB by putting the IMX6 in ideal state we are able to range upto 90M (Refer the attachment1) but when we activate the IMX6, ranging is reducing to 40M (50 % performance reduction).

Can anyone explain this RF interference phenomenon? If possible please suggest us the possible ways to solve this problem like some reference sites or some reference documents?

Thank you.

Regards,
Anirudh THABJUL

[Kilrah]

Likely EMI from the processor. Might need shielding, power supply filtering improvements, PCB layout improvements...

Also don't post duplicate topics.

[Anirudh_THABJUL_SB]

Hello Kilrah,

Thank you for your reply. I have removed other duplicate post. I am sorry i am new to this group.

We ahave mentioned

"Might need shielding, power supply filtering improvements, PCB layout improvements"

Can you please share some more info if you have like some online links or document attachments about power supply filtering improvements
?

Thank you.

[VK5RC]

Does the RF share any power components e.g. +/- voltage rail or ground with the microprocessor?
I have seen some designs where the RF has its own linear regulator and isolated ground as well.

[Anirudh_THABJUL_SB]

Hello VK5RC,

Thank you for your reply.

Yes the RF share the Positive and negative power component tracks. So is there a way to fix this issue on existing prototype itself.

[Edsoncan]

Do you have spectrum analyzer?
If is it possible that the EMI can affect the TXO frequency?
Are you using a pcb antenna or a cable antenna?

 

[bitslice]

(not an RF person)

Something I've noticed in other people's RF designs is that every chip or semiconductor is linked to the power rail via an inductor, presumably to filter out power rail noise.

I'd probably stick a uP inside a metal can

Another technique I've seen is to use a spread spectrum design for the uP clock

[Psi]

Try with the IMX6 running at a few different clock frequency, it may simply be your IMX6 clock is a harmonic of the RF frequency.

[Anirudh_THABJUL_SB]

Unfortunately, We don't have spectrum analyzer with us. We are using four ceramic antenna like in the image attached below.

[Anirudh_THABJUL_SB]

Try with the IMX6 running at a few different clock frequency, it may simply be your IMX6 clock is a harmonic of the RF frequency.

Hello Psi,

Thank you for your reply. We will definitely try this technique.

[Ice-Tea]

Sounds a lot like you're coupling digital noise from the uC into your RF. Have you measured the power supply noise at the RF part with/without the uC active?

[Edsoncan]

You calculated the mca? (micro channel)
Your  antenna traces are so thin...

[Anirudh_THABJUL_SB]

Sounds a lot like you're coupling digital noise from the uC into your RF. Have you measured the power supply noise at the RF part with/without the uC active?

Not yet, We will try to calculate it. Can you please explain me what can we conclude exactly after all the noise calculations?

[Anirudh_THABJUL_SB]

You calculated the mca? (micro channel)
Your  antenna traces are so thin...

Do you mean be impedence matching? If yes, Yes we have maintained track widths as per the calculations proposed after imepedence calculations.

[Edsoncan]

I am sorry my english is not good.
I refer  Coplanar Waveguide  https://en.wikipedia.org/wiki/Coplanar_waveguide

[Ice-Tea]

Sounds a lot like you're coupling digital noise from the uC into your RF. Have you measured the power supply noise at the RF part with/without the uC active?

Not yet, We will try to calculate it. Can you please explain me what can we conclude exactly after all the noise calculations?

Calculate? No, measure. With a decent scope. The PDN (Power Delivery Network) to your RF part must be as clean as possible. If your digital part shares the same network, it will inject noise on the power rail, which will end up in the RF part and which is most likely degrading your RF noisefloor.

[nctnico]

I second the suggestion of filtering the power supply to the RF section. A spectrum analyser may be a cheaper than a multi-GHz oscilloscope and you can connect a 50 Ohm active or passive probe to a spectrum analyser to look at the spectrum present in a PCB trace.

I'm also missing lots of vias to stitch the top copper pours to the inner ground plane. The top copper pour which is supposed to be the coplanar waveguide ground may be acting as an antenna as well.

[Gribo]

Few more comments regarding your design:
Your RF traces are long and travel in unnecessary bends and twists. Keep them as short and as straight as possible.
Add shielding vias to the ground on both sides of the RF traces, for 2.4GHz, roughly 1cm between each via.

Try soldering an SMA connector instead of your ceramic antenna, and see if you get the required output power, with a spectrum analyzer.

 

[Ice-Tea]

You may also want to check if you can read out stuff like noisefloor, RSSI, SNR from you device and check with/without the uC active.