Retransmit problems v2.0

[joelucid]

Remember I had problems with Ack's not getting through from gateway to client resulting in double updates received by the gateway. That problem was a side effect of sharing the RFM69HW radio and was easily solved.

Well now I have the opposite problem: I regularly lose more than 10% of all packets from sensor node to gateway. That might not seem much - but it is too much as can be seen from the log below.

Code Select Expand

06/01 13:15:33: !boot,nd:10,c:14492,ac:26645
Boot request from 10, checksum 14492
Starting to boot 10
Bootloader CRC matches
Installing new app
Serving Client
New profile: 1, timeout: 30
New profile: 0, timeout: 10
Got ack in: 2469, drop rate: 0 empty: 0
Timeout msg: 6
Boot client done
06/01 13:15:36: [10] [RX_RSSI:-72]
06/01 13:15:36: >nd:10,nr:1,pw:10,vc:340,rt:-1,l:1,rs:8,gs:2
06/01 13:15:40: [10] [RX_RSSI:-55]
06/01 13:15:40: >nd:10,nr:2,pw:10,vc:343,rt:0,l:1,gs:2
06/01 13:15:44: [10] [RX_RSSI:-55]
06/01 13:15:44: >nd:10,nr:3,pw:10,vc:343,rt:0,l:1,gs:2
06/01 13:15:48: [10] [RX_RSSI:-72]
06/01 13:15:48: >nd:10,nr:4,pw:10,vc:343,rt:0,l:1,gs:2
06/01 13:15:52: [10] [RX_RSSI:-72]
06/01 13:15:52: >nd:10,nr:5,pw:10,vc:343,rt:0,l:1,gs:2
06/01 13:15:56: [10] [RX_RSSI:-55]
06/01 13:15:56: >nd:10,nr:6,pw:10,vc:343,rt:200,l:1,gs:2
06/01 13:16:00: [10] [RX_RSSI:-55]
06/01 13:16:00: >nd:10,nr:7,pw:10,vc:343,rt:166,l:1,gs:2
06/01 13:16:04: [10] [RX_RSSI:-55]
06/01 13:16:04: >nd:10,nr:8,pw:10,vc:343,rt:142,l:1,gs:2
06/01 13:16:08: [10] [RX_RSSI:-72]
06/01 13:16:08: >nd:10,nr:9,pw:10,vc:343,rt:125,l:1,gs:2
06/01 13:16:12: [10] [RX_RSSI:-55]
06/01 13:16:12: >nd:10,nr:10,pw:10,vc:343,rt:111,l:1,gs:2
06/01 13:16:16: [10] [RX_RSSI:-55]
06/01 13:16:16: >nd:10,nr:11,pw:10,vc:343,rt:100,l:1,gs:2
06/01 13:16:20: [10] [RX_RSSI:-55]
06/01 13:16:20: >nd:10,nr:12,pw:10,vc:343,rt:90,l:1,gs:2
06/01 13:16:24: [10] [RX_RSSI:-55]
06/01 13:16:24: >nd:10,nr:13,pw:10,vc:343,rt:166,l:1,gs:2
06/01 13:16:28: [10] [RX_RSSI:-56]
06/01 13:16:28: >nd:10,nr:14,pw:10,vc:343,rt:153,l:1,gs:2

Here I update the app on the sensor moteino with 0 per mille drop rate using my bootloader at 200kbit/s. Then the node comes online and sends updates. Retransfers per mille are contained in the rt: parameter. As you can see after 15 updates we've already seen several retransmits. That shouldn't happen since the link is so much more stable at 200kbit

Is this level of retransmits normal for the stock RFM69 lib?

Joe

[joelucid]

BTW also notice the two phase RSSI which is either -55 or -72. I had seen something like this earlier but thought that problem was gone after exchanging the Pi for a Pi 2. Very strange.

[Felix]

Physics dictates that 200kbits is less "stable" - you are transmitting more data in the same amount of time. I would call it reliability not stability. That is why the lib has a stock of 55kb which is a good all around value that is proven to work and has a really low dropout rate, if any. 10% loss is not acceptable. You may have other things going on that affect this.

[joelucid]

Thanks Felix. that helped me figure it out. I had the duty cycle for listening for 200kbit boot clients at 10%. No wonder the 55kbit clients see > 10% drops. I've changed it to 30 ms : 1000 ms and now drops are acceptable.