I am trying to do DMX512 (protocol), I hope to help me, time is very urgent

[tony55723]

Hi，

DMX512 protocol is USITT（United States Institute for Theater Technology）Prescribed agreement

I want to follow this agreement to design

============ DMX512 protocol：

Baud：250K

Break：88us

MAB：8us

============ Data frame contains：

Start：1 bit

Data：8 bit

Stop：2 bit

Parity：None

I know that this design will still have errors.

I hope to give me some help, the project is urgent, thanks

Another stop bit need 2 bit, is it necessary to operate directly with registers?

（But the SDK is full of library operations？）

[Jørgen Holmefjord]

Hi,

Only UARTE peripheral support two stop bits, and this is only supported on nRF52840 and nRF52810 ICs. Unfortunatelu, there seems to be no support for setting this in the drivers. 

You can set two stop bits by writing the register directly after initializing the UARTE peripheral:

NRF_UARTE0->CONFIG |= (UARTE_CONFIG_STOP_Two << UARTE_CONFIG_STOP_Pos);

The UARTE peripheral have EasyDMA support, meaning transfers can happen in the background while CPU is busy handling BLE/softdevice events. If you frequently need to handle UART events with strict timing requirements, this will not work together with the softdevice.

Best regards,
Jørgen

[Jørgen Holmefjord]

I'm not familiar with the protocol details of DMX, but if it have strict timing requirements, an external MCU might be a better solution.

[awneil]

PM Bubblynet said:

this form...

Tell me about it!!

PM Bubblynet said:

Please define "strict" timing?

Exactly. These are just "finger in the air" feelings - not detailed timing analyses

Anyhow, it sounds like you are the DMX expert here - so you're probably the best one to try it and see!

The point is that the BLE stack does have some very strict timing requirements - so whatever you do for the DMX must not mess them up - or, maybe more important, must not be messed up by the stack grabbing the CPU.

So the thing with a separate MCU is that it's the safe option.

PM Bubblynet said:

its not a big deal if we miss few packages here and there

I think any issues are going to be with individual bit timings - rather than whole packages?

PM Bubblynet said:

similar timing requirements that any asynchronous serial protocol

But they are usually handled in hardware - by the UART.

From the initial postings, it looked like the nRF52 UART was not (quite) going to do what DMX requires?

PM Bubblynet said:

if and when a DMX to Bluetooth translator should be on our roadmap

If you're looking for help with that, drop me a PM ...

[PM Bubblynet]

I wouldn't call me a "DMX expert"... I am rather just a follower of our dark lord and savior all hail Google! :D

I went quickly though the UART docs https://www.nordicsemi.com/en/DocLib/Content/SDK_Doc/nRF5_SDK/v15-2-0/hardware_driver_uart cant see a reason why UART wouldn't be able to handle DMX? for example I am curious about:

Jørgen Holmefjord said:

Only UARTE peripheral support two stop bits, and this is only supported on nRF52840 and nRF52810 ICs. Unfortunatelu, there seems to be no support for setting this in the drivers. 

it seems odd that only the baby brother and the bigger brother "supports" 2 stop bits? its just me or this is part of the protocol / formatting layer and nothing to do with UART as a driver / hardware layer?...? it seems this would be just part of the formatting library of the telegram...?

a standard 250kbit/s UART shouldn't be enough? and even then this should be enabled 2 or 3 times per second... maybe? just in case we need resources for something else? again I apologize if I am missing something basic / obvious here... I should stop being lazy and study stuff before asking :D

[awneil]

But you clearly know more about it than me!

I'm just going from the opening post - but then it's not entirely clear that the poster really understood that.

Yes, the actual data frame looks to me like just a standard UART frame.

The potential problem areas seem to be in the initial "break", and the "mark after break". Possibly also the "inter-frame time".

The OP's code did that with inline delays - that is likely to get disrupted by the BLE stack ...

[PM Bubblynet]

hm.... that seems a bit weird idea to implement the "break"... anyways, found this: https://www.nxp.com/docs/en/application-note/AN3315.pdf

3.1 Software Implementation Because the protocol consists of an untyped byte stream produced by standard UARTs, the implementation can be easily adapted to work with any microcontroller supporting baud rates up to 250 kbps.

I would put my 5 bucks on the nRF52832 can do this... but I cant jump now to prototyping... need to finish a couple of things before holidays, will see after that how it goes...

[awneil]

If that is the case, then I'd agree with you.

It would also seem sensible that the spec writers would base their protocol on something that can be easily done with standard hardware ...

[Alan]

I've been playing around with DMX.  I have DMX transmit pretty much working. One way to handle the initial Break - Mark is to change the baud to 115K and transmit a 00, then change it back to 250K.  I have a 513 byte DMX buffer and it needs to get sent in three pieces because of the DMA 255 byte limit.  If the SoftDevice interrupts in the middle it shouldn't be a problem because the DMX spec only requires that a frame gets completed within one second.  From past experience, many DMX devices will work fine if you just send 1 stop bit, but not all.  So if you need to be able to transmit to any DMX device you need 2 stop bits which means the nRF52832 is not an option.

Doing a DMX receive looks like it will be more difficult.  It looks like the driver supports notification of a BREAK which can be used to initiate receive of the 513 byte frame.  The problem is if the SoftDevice causes a delay in the start of the receive you could start receiving data in the middle of the frame.  Also, due to the DMA 256 byte limit the receive needs to be split into multiple pieces which makes for more opportunities for errors.  

Is there a good way to detect if a large delay has occurred or is there a way to chain several DMA operations together to allow a longer stream of uninterrupted data?

[Jørgen Holmefjord]

Other serial peripherals have a EasyDMA array list feature (TWIM, SPIM), but this seems to not be an option for UARTE. However, the .PTR and .MAXCNT registers are double-buffered in the UARTE peripheral. They can be updated and prepared for the next RX/TX transmission immediately after having received the RXSTARTED/TXSTARTED event. This should allow you to setup transfers for up to 510 bytes. If you need more than this, my recommendation would be to use one of the more recent ICs that support longer EasyDMA transfers than nRF52832 for UARTE peripheral. nRF52810 have 10-bit MAXCNT register, while nRF52840 have 16-bit MAXCNT register.

[Alan]

Thanks, that gives me two great options.  One more question, nRF52810 sample projects are all pca10040e (emulated) projects. Would a project that is targeted specifically for the pca10032 have any differences.  Can I use a pca10040e project for the starting point?

[awneil]

pca10032 is the nRF51 (Cortex-M0) dongle - so that's no use for any nRF52 (Cortex-M4)  projects!

http://infocenter.nordicsemi.com/index.jsp?topic=%2Fcom.nordic.infocenter.gs.ses%2Fdita%2Fgs%2Fdeveloping%2Fchips_and_sds.html

[Alan]

My mistake, I should have asked when developing for the nRF52810 is a pca10040e (emulated) project a good starting point.  I'm just concerned about the word emulated and is there something about the emulated projects that should be changed for a production build.