How to get right digree? Nrf52811 direction finding

RE: How to get right digree? Nrf52811 direction finding

Dmitry — Sat, 21 Mar 2020 18:05:31 GMT

Jitter is the real fact. My pic is a bit exaggerated. Okay, if it's not an issue for you.

Do you have any success in phase detection?

RE: How to get right digree? Nrf52811 direction finding

miaojb12 — Sat, 21 Mar 2020 16:41:42 GMT

Just like your PIC mean,I don't think BLE will work find.

I think crystal will change with temperature, but it changed slowly.

In a shot time, crystal is must be same frequency,this is base of PLL working

RE: How to get right digree? Nrf52811 direction finding

Dmitry — Sat, 21 Mar 2020 08:22:05 GMT

[quote userid="71096" url="~/f/nordic-q-a/59322/how-to-get-right-digree-nrf52811-direction-finding/240947"]but now, we did not get the right phase .
we don't knoue if there something wrong with hardwear.[/quote]

I cannot say anything more with no info... it would be good if you can take two sample sets in CSV format under two different angles.

[quote userid="71096" url="~/f/nordic-q-a/59322/how-to-get-right-digree-nrf52811-direction-finding/240947"]but, In my mind, we take more data, just for reduse the fluctuations of jitter,
is not it?[/quote]

Yes, averaging works, but longer distance between samples increases error (that's what I see on my device, though you may have better crystal) - if you can get rid of it just by decreasing slot length, I see no reasons to work with such a long slots. See what I mean:

RE: How to get right digree? Nrf52811 direction finding

miaojb12 — Fri, 20 Mar 2020 17:47:07 GMT

yes, your are right aboat -360/+360 degree turnover.
we resove it .
1:Straight line fitting with 1st and 5th ant, get slope1 and slope5.
2:slope_average=(slope1+slope5)/2. maybe this slope is a nearly one with true slope.
3:while(slope_green - slope_average > 360degree/48us){
slope_green -= 360degree/48us;
}
while(slope_green - slope_average <- 360degree/48us){
slope_green += 360degree/48us;
}
4:slope_green is not turnover now .
//----------------------------------------------------------------------
we have tested about slope_average and slope_green.
slope_green is much more stable than slope_average.

dynamic range
slope_green <0.2%
slope_average <1%

but now, we did not get the right phase .
we don't knoue if there something wrong with hardwear.

//-------------------------------------------------------------------
PS:Do you mean the jitter is random.
but, In my mind, we take more data, just for reduse the fluctuations of jitter,
is not it?

RE: How to get right digree? Nrf52811 direction finding

Dmitry — Fri, 20 Mar 2020 17:10:16 GMT

Almost correct.. see, you have 10ppm crystal difference (don't look at jitter now). At 2.4GHz frequency, you get 24kHz offset - 10ppm turns into 9.6% error relative to 250kHz, in 48 usec your green line could make a -360/+360 degree turnover... Now, jitter is a random fluctuations of crystal frequency, and thus a received phase too. While jitter is negligible at bitrate level, being multiplied by RF demodulator it becomes visible. The longer distance between points, the larger error is accumulated.

RE: How to get right digree? Nrf52811 direction finding

miaojb12 — Fri, 20 Mar 2020 13:46:25 GMT

I draw a pic, wish you can understand what i mean.

RE: How to get right digree? Nrf52811 direction finding

Dmitry — Fri, 20 Mar 2020 12:49:48 GMT

[quote userid="71096" url="~/f/nordic-q-a/59322/how-to-get-right-digree-nrf52811-direction-finding/240837"]our antenna switching ic is "SKY13322-375LF",whitch's switching time is 60ns.
[/quote]

good.

[quote userid="71096" url="~/f/nordic-q-a/59322/how-to-get-right-digree-nrf52811-direction-finding/240837"]1-16 ignored(may be changed with the switch)
17-80 used for compute average phase
81-96 ignored(may be changed with the switch)[/quote]

it's better to look at the plot, hard to say exactly where slots are changed. In theory you have two samples for reference period then continuous set of 0.125us samples after 1us gap.

[quote userid="71096" url="~/f/nordic-q-a/59322/how-to-get-right-digree-nrf52811-direction-finding/240837"]3:comput frequency offset beteen receiver and transmitter with 1sth and 5th ant(whitch is same ant)[/quote]

as a variant, but

[quote userid="71096" url="~/f/nordic-q-a/59322/how-to-get-right-digree-nrf52811-direction-finding/240837"](longer slot -> more crystal jitter influence -> lower accuracy) may resoved?[/quote]

what I mean - the distance between points that you compare should be as short as possible, a jitter is really noticeable. You have 12us slots, it's too long. Of course you can average between multiple periods, but I see no reason for adding jitter error if you can work with shorter slots.

[quote userid="71096" url="~/f/nordic-q-a/59322/how-to-get-right-digree-nrf52811-direction-finding/240837"]4:use 17-80 datas of each ant compute average phase[/quote]

Maybe I didn't understand clearly, but you should apply frequency offset correction to each sample before calculating average phase.

RE: How to get right digree? Nrf52811 direction finding

miaojb12 — Fri, 20 Mar 2020 11:25:51 GMT

thanks for your reply.

our antenna switching ic is "SKY13322-375LF",whitch's switching time is 60ns.
We got 96 data for each ant.

1-16 ignored(may be changed with the switch)
17-80 used for compute average phase
81-96 ignored(may be changed with the switch)

//-----------------------------------------------------------------------------------------------

How we compute phase
1:we have 4 ants all;
2:we get 1st ant and 5th ant the same one
3:comput frequency offset beteen receiver and transmitter with 1sth and 5th ant(whitch is same ant)
(longer slot -> more crystal jitter influence -> lower accuracy) may resoved?
frequency offset <0.2%wave change,like 11.373,11.372,11.374,11.370
If frequency offset is 0, we will get num like, 12.500 beteen each data.
4:use 17-80 datas of each ant compute average phase
5:compute phase back to the time of first ant.
//--------------------
example
1th average phase:221
5th average phase:166
will get the change beteen each data is 11.373(frequency offset=22.54kHz )

2nd average phase:296 --> 9(phase of 2nd at 1st ant time)
(296-(11.373*96))=-795
-795+402*2 = 9

3rd average phase:249 -->76(phase of 3rd at 1st ant time)
4th average phase:271 -->-190(phase of 4th at 1st ant time)

//-------------------------------------------------------------

Is there any mistake?

RE: How to get right digree? Nrf52811 direction finding

Dmitry — Fri, 20 Mar 2020 09:34:03 GMT

Hi,

could you visualize your data as a plot or at least put CSV here? It's really hard to parse tons of digits in the head 🙂

From your config - all 96 samples falls under the first antenna - we won't see any phase change.

IMHO switch pattern is too complex and outstretched, you should use as short switching slot as your antenna switching IC allows, ideally not longer than 2 usec (longer slot -> more crystal jitter influence -> lower accuracy). That's why reference period is separated - you have enough time on a single antenna to calculate frequency offset, and you have no chance to do that with two samples (4us sample spacing) - at least 8 samples are needed.

RE: How to get right digree? Nrf52811 direction finding

miaojb12 — Fri, 20 Mar 2020 07:31:31 GMT

sorry,changed < first 2+16 samples is ignored。>