UART question/WROOM-32

frenkR
Posts: 9
Joined: Fri Nov 27, 2015 11:05 pm

UART question/WROOM-32

Postby frenkR » Sun Oct 30, 2016 10:50 pm

hi,
on ESP32 pinout(images from web), it is specified that TXD0=GPIO1, RXD0=GPIO3, RXD1=GPIO9, TXD1=GPIO10. Latter two are also used for SPI access to flash.
in ESP32 doc (esp_wroom_32_datasheet_en.pdf) on page 9, there is specified that UART0/UART1 can use arbitrary GPIO pin. Is this correct?
Rgds,
Frenk

User avatar
kolban
Posts: 1683
Joined: Mon Nov 16, 2015 4:43 pm
Location: Texas, USA

Re: UART question/WROOM-32

Postby kolban » Sun Oct 30, 2016 11:26 pm

Although it is pretty technical stuff, this is the document you are likely going to want to study:

http://espressif.com/sites/default/file ... ual_en.pdf

It is called the ESP32 Technical Reference Manual. Specifically, look at Figure 6 on page 28. From there we see that there are two "mapping" elements. A thing called the IO_MUX and a thing called the "GPIO matrix". From this I believe that there are "logically" 256 distinct peripheral control signals that can be mapped to 40 physical pads (pins). Of the possible 40 pads (numbered 0-39), five are not exposed (24, 28, 29, 30, 31) giving is 35 specific pads (pins).

Now if we look at table 16 on page 34, we see the 256 possible signals ... which for example include 16 in (U0RXD) and 16 out (U0TXD). Great ... that then seems to say that I can use the matrix to map signal 16 in to an arbitrary input pad and 16 out to an arbitrary output pad. But then things take a turn for the "ahh" ... when we look at table 17 on page 38. This seems to show that U0RXD is tied to pad 3 while U0TXD is tied to pad 1. So how can we explain the contradiction? Are the functions to pins fixed or not?

And unfortunately I don't have enough knowledge/skills to answer that ... but I'll stick my neck out and make a guess in the hope that better folks than me can correct.

I think that in the ESP32 there are two types of function/pad mapping. A matrix mapping using a GPIO matrix and a "faster" direct IO mapping known as "IO_MUX". My belief is that one CAN map any function to any pad using only matrix mapping. However, should one wish/need, one can bypass matrix mapping and expose functions on very specific pins to get "maximum" performance/throughput.

Again ... making it up as I go along.... you can map U0RXD and U0TXD to arbitrary pins using GPIO matrix ... however somehow "performance" or "quality" will be less than using "direct IO" through IO_MUX which forces particular functions to specific pins.

You will see I am being VERY woolly on the concept of "why" one would use matrix mapping vs direct IO.... and I have no idea. I don't know what the benefits may be in one fashion or another. Hopefully this post can be used as a basis to say "yeah/neigh" to some concepts and ideas ....
Free book on ESP32 available here: https://leanpub.com/kolban-ESP32

frenkR
Posts: 9
Joined: Fri Nov 27, 2015 11:05 pm

Re: UART question/WROOM-32

Postby frenkR » Sun Oct 30, 2016 11:50 pm

thank you for a very useful hint/link/thinking. Have a lot to study. thanks again:)

ESP_Angus
Posts: 2344
Joined: Sun May 08, 2016 4:11 am

Re: UART question/WROOM-32

Postby ESP_Angus » Mon Oct 31, 2016 12:38 am

Thanks Kanflo for posting a great summary. :)

Re: "performance tradeoffs", the tradeoff from routing a signal via the GPIO Matrix is a small amount of latency (2 clock cycles of the peripheral APB clock, at the default 80MHz clock this is 25 nanoseconds) and a lower maximum signal speed of 40MHz. The latency refers to the time between when a signal voltage appears on the pad, and when the signal reaches the peripheral (in this case the UART) to trigger an interrupt or update a register. Or vice versa, the time between updating a register and the pad transitioning to the new state. This gap becomes 25 nanoseconds longer if you're using the GPIO Matrix. This is really only something you need to worry about if you're routing very high speed signals and/or need to save every last nanosecond of latency. In all other cases, using the GPIO Matrix gives you much more flexibility.

One other important thing to keep in mind is that the serial bootloader uses UART0 RXD0/TXD0 as the usual/default method to load images into flash, and these pins are fixed (as there's no opportunity to reprogram the GPIO matrix on an initial reset.) For this reason, it's much easier if you keep the UART0 pins in the default pin locations.

The same advice applies to keeping the SPI flash pins in their default locations (as shown in the module schematics, etc.) It is possible to use the EFUSE feature to reconfigure the SPI flash pins if you need these elsewhere, but it's always going to be easier to keep the defaults.

Who is online

Users browsing this forum: No registered users and 144 guests