Project

General

Profile

Solved: LED node (v2) losing most data packets

Added by lanroth about 6 years ago

Hi there,

I recently got a couple of LED nodes (v2) and assembled them according to the great guide here:
http://jeelabs.org/2012/12/28/assembling-the-led-node-v2/

They work great except they both lose lots of data packets. They’re both running the rf12demo and I’ve been testing with the ‘t’ command. About 1 in 30 packets sent via a JeeLink are received.

The JeeLink is about 4 meters away through a couple of walls.

When I swap the led node for a pre-assembled JeeNode v6 everything works as expected: sending a ’t’est packet using the rf12demo works >95% of the time.

The LED nodes are configured exactly the same as the JeeNode.
I tried swapping the microcontroller in the JeeNode with the one in the LED node with no effect.

From this I can only conclude that it’s a hardware problem probably due to me messing up the assembly of both LED nodes.

All testing was done on bare LED nodes, i.e. with no other hardware attached to them except the USB BUB which was used for power.

I made an antenna cable by measuring the one on the working JeeNode.

Any advice would be greatly appreciated,
Steve


Replies (10)

LED node (v2) losing most data packets - Added by martynj about 6 years ago

The BUB is jumpered to provide 5V or 3.3V power (different jumper from the 5V or 3.3V logic signal)?
Is the packet loss roughly symmetrical?   JeeLink <-> LEDnode ?

RE: LED node (v2) losing most data packets - Added by lanroth about 6 years ago

Hi Martyn,

I’ve not changed the BUB’s jumpers from their default setting, but I can’t remember what that is.

I think packet loss is about the same in both directions, but will test and report back this evening.

Thanks for your prompt reply :-)
Steve

RE: LED node (v2) losing most data packets - Added by martynj about 6 years ago

When the 12v supply is absent, the logic power on the board can be derived from the FTDI connector. Since there is a diode isolator before this voltage reaches the 3.3V LDO,  the BUB (or similar) needs to provide 5v on FTDI Pin 3.  If there is only 3.3v on FTDI Pin 3,  the LDO is out of range and the logic voltage level sits ~2v, causing many problems as you would expect.

RE: LED node (v2) losing most data packets - Added by lanroth about 6 years ago

I measured the voltage on FTDI pin 3 and it is 5v and also tried testing with the LedNode plugged into 12v but to no effect.

I can also confirm that sending and receiving are both lose about the same number of packets.

I’ll try moving the JeeLink close to the LedNodes and see if that improves the situation.

Thanks for your help with this,
Steve

RE: LED node (v2) losing most data packets - Added by martynj about 6 years ago

Reducing baud rate will help if the signal is marginal

RE: LED node (v2) losing most data packets - Added by martynj about 6 years ago

Steve, your reception problem is fixed?

RE: LED node (v2) losing most data packets - Added by lanroth about 6 years ago

Hi Martyn,

No, not yet. I tried moving the JeeLink to within a meter of the LedNode but this didn’t seem to help. I didn’t get much time to experiment though.

I’ve got a couple of spare hours tomorrow evening so I’ll try dropping the baud rate to see if that improves things.

Cheers,
Steve

RE: LED node (v2) losing most data packets - Added by martynj about 6 years ago

Hmm - something more fundamental I think. Try nudging only the LEDnode channel frequency in smallish steps (say 50KHz) i.e. F=1600 /- 5,/- 10 ….
in the Frequency Setting control word.

RE: LED node (v2) losing most data packets - Added by lanroth about 6 years ago

Hi Martyn,

Thanks for the suggestion. I wrote a sketch that set the LedNode frequency to ever diverging values from the default.

The packet loss rate dropped to around 3% when I used: rf12_control(0xA632). If my calculations are correct (and I’m doing this from memory, so they might not be!)
That’s a frequency of: 868+(0xA632-0xA640)*0.005 = 867.93Mhz

It’s working great now, your guidance has been invaluable on this, many thanks.

I’ve not had to fine tune the frequency on the RFM12B before and wonder how common it is?

Cheers,
Steve

RE: LED node (v2) losing most data packets - Added by martynj about 6 years ago

Steve - good, my suspicions were correct. Now you know the approx deviation, you can fix it with the crystal cap bank setting and go back to standard frequency settings. The old style metal-can crystal is usually /-12ppm, but runs a tad low with the default cap bank setting; the flat SMD is a tighter tolerance (better than/- 10ppm) but runs a tad high.

Usually these errors are much less than the range of the receiver automatic AFC adjustment and are not a problem. You can get a significant irreversible ‘jump’ out of factory calibration with the SMD type if it gets ‘overcooked’ during soldering (at module manufacture time and/or attaching to the motherboard). Your numbers show ~80ppm delta, that’s a lot and probably too much for the cap bank to correct (adding pf lowers the frequency, so try setting minimum cap bank value).

    (1-10/10)