RMT RX Filter "signal_range_min_ns" doesn't work (IDFGH-9977)

[hamdiolgun]

Answers checklist.

• I have read the documentation ESP-IDF Programming Guide and the issue is not addressed there.
• I have updated my IDF branch (master or release) to the latest version and checked that the issue is present there.
• I have searched the issue tracker for a similar issue and not found a similar issue.

IDF version.

v5.0-dev-6416-gef4b1b7704

Operating System used.

Windows

How did you build your project?

Eclipse IDE

If you are using Windows, please specify command line type.

None

Development Kit.

ESP32-WROOM32E

Power Supply used.

Battery

What is the expected behavior?

I am using RMT peripheral to communicate with other devices over HOMEBUS. There are a lot of noises on the bus, as natural.
I am trying to eliminate the noises with RMT RX filter module but i can't achieve it.
The communication baudrate 9600, and 1 bit 104us.
For Homebus, bit 0 means "52us 0-52us 1", bit 1 means "104us 1"

I set the rx config struct as;
static const rmt_receive_config_t rmt_receive_config = {
.signal_range_min_ns = (uint32_t)18 * 1000, //18us
.signal_range_max_ns = (uint32_t)104 * 12 * 1000 //~1200us
};
rmt_receive_config_t::signal_range_min_ns specifies the minimal valid pulse duration (either high or low level). A pulse whose width is smaller than this value will be treated as glitch and ignored by the hardware.
I expect that i should not get signals less than 18us by rmt_rx_done_event_data_t.

What is the actual behavior?

I got all signals higher than 1us by rmt_rx_done_event_data_t rx data.

Steps to reproduce.

rmt_channel_handle_t rmt_rx_channel;
const rmt_receive_config_t rmt_receive_config = {
.signal_range_min_ns = (uint32_t)18 * 1000,
.signal_range_max_ns = (uint32_t)104 * 12 * 1000
};
QueueHandle_t rmt_receive_queue;

rmt_rx_channel_config_t rx_channel_cfg = {
.gpio_num = CONFIG_IO_VRF_RX_PIN,
.clk_src = RMT_CLK_SRC_DEFAULT, //80MHz
.resolution_hz = 1000000, // 1MHz resolution, 1 tick = 1us
.mem_block_symbols = 5 * SOC_RMT_MEM_WORDS_PER_CHANNEL,
.flags.invert_in = VRF_GPIO_RX_INVERT,
.flags.with_dma = 0,
.flags.io_loop_back = 0,
};
ESP_ERROR_CHECK(rmt_new_rx_channel(&rx_channel_cfg, &rmt_rx_channel));
configASSERT(rmt_rx_channel);
rmt_receive_queue = xQueueCreate(VRF_DATA_BUF_SIZE, sizeof(rmt_rx_done_event_data_t));
configASSERT(rmt_receive_queue);
rmt_rx_event_callbacks_t rcbs = {
.on_recv_done = rmt_rx_done_callback,
};
ESP_ERROR_CHECK(rmt_rx_register_event_callbacks(rmt_rx_channel, &rcbs, rmt_receive_queue));
ESP_LOGI("RMT", "Init done.");

rmt_enable(rmt_rx_channel);
rmt_receive(rmt_rx_channel, rmt_raw_symbols, sizeof(rmt_raw_symbols), &rmt_receive_config);//ready

Debug Logs.

[0]{50, 0, 470, 1}{1, 9}
[1]{50, 0, 3, 1}{1, 0}
[2]{10, 0, 562, 1}{0, 11}
[3]{48, 0, 54, 1}{1, 1}
[4]{50, 0, 3, 1}{1, 0}
[5]{1, 0, 987, 1}{0, 19}
[6]{48, 0, 262, 1}{1, 5}
[7]{50, 0, 3, 1}{1, 0}
[8]{7, 0, 772, 1}{0, 15}
[9]{48, 0, 574, 1}{1, 11}
[10]{50, 0, 3, 1}{1, 0}
[11]{11, 0, 456, 1}{0, 9}
[12]{49, 0, 262, 1}{1, 5}
[13]{50, 0, 2, 1}{1, 0}
[14]{6, 0, 774, 1}{0, 15}

More Information.

No response

[suda-morris]

@hamdiolgun

Thanks for reporting! In fact, .signal_range_min_ns = (uint32_t)18 * 1000, //18us will result in a very large register value 1440 (which was calculated in the https://github.com/espressif/esp-idf/blob/master/components/driver/rmt/rmt_rx.c#L368). However, the underlying register is only 8bit-width, so actually what being set in the register is only 1440/256 which is only 5.

We will add a check in the driver in case the user sets a value that will lead to out-of-register-range.

Meanwhile, you can reduce the RMT clock resolution by choosing another clock source, e..g .clk_src = RMT_CLK_SRC_REF_TICK, //1MHz, then we can get a valid register value after calculation.

[hamdiolgun]

@suda-morris
Thanks for your reply.
I had been realized the calculation of filter threshold.
rmt_ll_rx_set_filter_thres(hal->regs, channel_id, ((uint64_t)group->resolution_hz * config->signal_range_min_ns) / 1000000000UL);
And i've set the signal_range_min_ns according that,

rmt_rx_channel_config_t rx_channel_cfg = {
    .gpio_num = CONFIG_IO_VRF_RX_PIN,
    .clk_src = RMT_CLK_SRC_DEFAULT, //80MHz
    .resolution_hz = 1000000, // 1MHz resolution, 1 tick = 1us
    .mem_block_symbols = 5 * SOC_RMT_MEM_WORDS_PER_CHANNEL,
    .flags.invert_in = VRF_GPIO_RX_INVERT,
    .flags.with_dma = 0,
    .flags.io_loop_back = 0,
};

So,
group->resolution_hz = 1000000
config->signal_range_min_ns = 18 * 1000
filter thres = (1000000 * 18 * 1000) / 1000000000UL = 18
How did you calculated 1440 ?

Best regards

[suda-morris]

So, group->resolution_hz = 1000000

This is a trap of RMT. When setting the filter threshold, we're using the RMT's "group resolution" (one group has several channels). Which by default, equals the source clock's frequency (80MHz e.g.).

https://github.com/espressif/esp-idf/blob/master/components/driver/rmt/rmt_common.c#L166

.resolution_hz = 1000000 this configuration is set the "channel resolution".

[hamdiolgun]

Ow sorry, i'd missed it.

So,
if i select RMT_CLK_SRC_REF_TICK (group->resolution_hz will be 1MHz) as rmt tx/rx clk_src and i set tx/rx resolution_hz as 1MHz, i will reach 1us resolution again.
Then, i can set signal_range_min_ns as 18 * 1000.
real_div will be set 1 according the configuration and all that is legal. right?

esp-idf/components/hal/esp32/include/hal/rmt_ll.h

Line 311 in 3cb4db6

static inline void rmt_ll_rx_set_channel_clock_div(rmt_dev_t *dev, uint32_t channel, uint32_t div)

If i'm correct,
this gonna solve my filtering problem. (rmt rx gonna filter signals less than 18us)

Regards