RMT RX Jitter

[IBims93]

I am currently trying to use the RMT to receive a 6 Mbit/s Biphase-Manchester (every bit starts with a change in value, zeros have an additional change in the middle of the bit-time) Datastream.
This boils down to detecting whether the signal changed after a short (83,3333...ns) or a long (166,666...ns) delay.
With the Clock of 80 MHz a short delay should be 6,666... ticks and a long delay should be 13,333... ticks.
I would expect to see counts of 6 and 7 for a short signal and counts of 13 and 14 for a long signal.
However there is a high level of Jitter in the received values.
Below is an example of what the data looks like coming from the RMT.

Code: [Select all] [Expand/Collapse]

1. {{{6,1,7,0}}},
2. {{{6,1,7,0}}},
3. {{{6,1,12,0}}},
4. {{{14,1,8,0}}},
5. {{{7,1,7,0}}},
6. {{{6,1,10,0}}},
7. {{{15,1,8,0}}},
8. {{{7,1,10,0}}},
9. {{{15,1,9,0}}},
10. {{{7,1,11,0}}},
11. {{{10,1,6,0}}},
12. {{{12,1,9,0}}},
13. {{{7,1,11,0}}},
14. {{{14,1,9,0}}},
15. {{{6,1,11,0}}},

GeSHi © Codebox Plus Extension

The distinction isn't as clear cut as I would expect.
And even worse sometimes there is count of 10. 10 is exactly on the border between what can be considered long an short and is basically impossible to decode.

I am using the rmt-driver integrated in ESP-IDF.
Initialization looks like this:

Code: [Select all] [Expand/Collapse]

1.     rmt_config_t config = RMT_DEFAULT_CONFIG_RX(CONFIG_EXAMPLE_RMT_RX_GPIO, RMT_RX_CHANNEL);
2.     config.rx_config.filter_ticks_thresh = 4;
3.     config.rx_config.filter_en = true;
4.     config.clk_div = 1;
5.     config.mem_block_num = 2;
6.  
7.     ESP_ERROR_CHECK(rmt_config(&config));
8.     ESP_ERROR_CHECK(rmt_driver_install(config.channel, 1000, 0));

GeSHi © Codebox Plus Extension

Afterwards I just use the "xRingbufferReceive" function to pull available data from the buffer.



Maybe someone in this forum has done something similar and figured out why there is that much Jitter in the received data. I am kind of at a loss. The APB-Clock should be fairly stable and the datastream looks almost perfect when viewed with a logic-analyzer or a scope.

PS: Sorry for any spelling/grammar mistakes, English is not my native language.