A Properly Isolated DMX Interface

[JEC]

I thought folks might be interested in a new design I've been kicking around.

It was designed for driving chains of pixels, but may have other uses as well.

Attached is a double-size layout of the new 'DMX Offset Machine.'

Basically, you use the DIP switches to set the system's start address. Then, the processor reads the incoming DMX stream, ignores everything received *before* the start address, then generates a *new* data stream containing only the remainder of the DMX packet. This new stream is sent out the door on a CAT5 jack for driving pixel strings.

Make sense?

The gist of it is that you can hard-code your chain of pixels to always have addresses 1, 4, 7, etc... Then just use this circuit board to set their 'proper' start address. Thus allowing you to use the system in multiple locations.

Featuring...

* XLR-5 connectors for DMX in and through, all 5 pins in parallel
* DMX input is optically and galvanically isolated from rest of circuit (per the USITT spec)
* DIP switch for selecting start / offset address
* RJ-45 output for pixel string drive
* Status and power LEDs
* Screw terminals for +12V input, which drives this circuit + the pixel chain
* Board size is ~ 3.5" x 2.5"

Schematic and board layout attached.

Thoughts?

[sparky3]

it's a good idea and would be great for any hard coded light fixtures. i have 5 rgb led color wash fixtures that are each hard coded channels1,2,and 3 and i want to be able to give them there own addresses. i have a working dmx reciever put together using a 16f876 right now i just displays the channel values but i haven't set down and done any coding for the transmitting side of it yet.

do you think it would be a good idea to put a end address selection in the code so that if you are starting at channel128 and only need to channel 158 it won't work through the rest of the 352 or 353 channels?

[JEC]

do you think it would be a good idea to put a end address selection in the code so that if you are starting at channel128 and only need to channel 158 it won't work through the rest of the 352 or 353 channels?

Yep - the general idea is to keep the lightning out of the rest of your rig.

The idea of an end address is an interesting one. In theory, the gear you're using shouldn't care if there is one data byte or 512 in a DMX packet - and should function identically in both cases.

[sparky3]

your right it should not matter now that i think about it, on the circuit that i put together i set the start address and it recieves the next 3 channels bust out of the recieve loop and then displays the data on a lcd for four channels.
i use basic micro for a compiler and had to pound assembler into my head to get the data from the dmx stream and then put in the middle of a basic program.
by the way thanks for the code samples i used part of your program,microchips program, and some of N5TLE program to get mine together. with out them i would still be stuck.
i still consider my self a newbe when it comes to programing and that is why i asked. i guess you will just be dumping the rcreg into the txreg once you hit the staritng address? basicly?

the thought lightning never occured to me.

Rob

[Trepidati0n]

My only thought woudl be to stay away from the XLR-5 connectors. I believe there is a "soft" spec which specifies DMX on an ethernet cable. If I did have to use XLR-5..I would much rather make a dongle cable.

Also..a thought would be to see if you can put the footprint for the RJ45 connector under the XLR connector. Therefore the "owner" could chose which one they wanted installed...

[wbuehler]

JEC

Do you currently offer this board or plan any Co-Op in the future?
I just made an oder request for some Pixels and would find this board useful.

Thanks

Bill

[JEC]

@Trepidati0n: Good thought - I'll revise the board layout to include DMX over CAT5.

@ Bill - yes, one I run the prototype and make sure it's working properly I'll make these available as well. Probably just PCBs at this point.

[Trepidati0n]

More thoughts...to make this more "friendly" to other DMX devices, you may also wish to add a standard DMX output besides the pixel output. That would be 4-RJ45 connectors (2 for the main chain, 1 for pixel, 1 for standard expansion).

This way people could basically create a splitter by cascading the boards with short ethernet cables (the 1'foot versions).

[wbuehler]

Awesome

Thanks

If you need some help testing a board I would like to volunteer

Bill

[JEC]

More thoughts...to make this more "friendly" to other DMX devices, you may also wish to add a standard DMX output besides the pixel output. That would be 4-RJ45 connectors (2 for the main chain, 1 for pixel, 1 for standard expansion).

This way people could basically create a splitter by cascading the boards with short ethernet cables (the 1'foot versions).

You want 2 jacks in parallel (in and through) or an isolated output with a separate driver? If the latter, it would probably make most sense to copy the output directly from the input, skipping the processor. Yes?

[Trepidati0n]

Potato/Potato....

If you plan on selling these as a finished product (like the pixel), I guess it depends on the latency when passing through the pic. If you are just a byte behind..I would say parallel.

If you plan on just selling bareboards...then go with the option to skip the processor.

[JEC]

Potato/Potato....

If you plan on selling these as a finished product (like the pixel), I guess it depends on the latency when passing through the pic. If you are just a byte behind..I would say parallel.

If you plan on just selling bareboards...then go with the option to skip the processor.

System latency shouldn't be more than a full DMX frame (20 mS).

I think my preference would be to have DMX in, DMX through, DMX isolated out and 'parsed' pixel drive/power combo out.

Don't know what I'll be selling, to be honest. Let's see how much interest there is. Worst case I'll run a stack of extra boards and publish a parts list. Best case I'll have the local assembly shop put them together for me. I can't imagine there being enough volume to justify overseas production & shipping.

But maybe I'm wrong.

This design is all through-hole, so it shouldn't take more than 15 minutes or so to assemble from scratch.

John

[wbuehler]

Nice

I to would like the option of the board providing power to the Pixels

Thanks

Bill

[P. Short]

Bill,

Are you talking about adding a circuit to generate +12V for powering the pixels? Each pixel takes about 120 mA max, so eight of them is 1A, and so forth. The power adds up fairly rapidly as you add more pixels. The expense and complexity of DIY power circuits grows when the power is above 2A or so, to the point where it might be more economical to buy a pre-made power supply (if not a PC power supply, then one of the Phihong or similar open-frame switchers).

That said, it might be useful to add holes for additional diodes and a large filter cap so that you could power the board and a small number of pixels by just adding an external transformer. Alternatively, add the pacs for a power jack so that you could use a wall-wart power adapter for the same reasons.

--

Phil

[JEC]

Bill,

Are you talking about adding a circuit to generate +12V for powering the pixels? Each pixel takes about 120 mA max, so eight of them is 1A, and so forth. The power adds up fairly rapidly as you add more pixels. The expense and complexity of DIY power circuits grows when the power is above 2A or so, to the point where it might be more economical to buy a pre-made power supply (if not a PC power supply, then one of the Phihong or similar open-frame switchers).

That said, it might be useful to add holes for additional diodes and a large filter cap so that you could power the board and a small number of pixels by just adding an external transformer. Alternatively, add the pacs for a power jack so that you could use a wall-wart power adapter for the same reasons.

--

Phil

Yep - in my mind, I imagine feeding this board with a 12v, 10A or so switching power supply, probably from a computer. $20 - $30 buys a *lot* of power these days.

[wbuehler]

****Edited******

I see that JEC posted back looks like we can add an exteral power supply and have the board feed the pixels. If that is if JEC goes this route? :D

[Trepidati0n]

May be I incorrectly understood JEC's post about diffferent output options such as "Pixel drive/power combo out" I was thinking that by drive/power that he meant data and power to the pixel? If this is the case it would be nice to power the pixel from the board without the need of suppling an external power supply.

Bill

There is no reasonable way a wall supply can do this IMO. Wall supplys are good for a couple watts...not 10's of watts. If you do find a wall supply like this...the cost tends to go up a bit (beyond that of a PC PSU)...and regulation isn't very good.

[P. Short]

That is what the schematic in the first post on this topic shows. So unless he has changed it...

--

Phil

[wbuehler]

OK that clears things up if the port is made a combo port then the board can help by passing power from the external supply to the pixel via the Pixel RJ45 connector. This would be nice since I would not need any other adapter to inject or have to splice into the RJ45 cable to feed power to the Pixel.

Thanks for the help

Bill

[P. Short]

What is the design limit on the current pass-through for the pixels? Just for the record, how much current will the traces on the PCB support?

--

Phil

[Wombat]

I think this idea is great. I was thinking of getting some pixels and the DMX splitter sounds great.

Is it possible to have do a Vixen-485/232 to DMX conversion on it as well bypassing having to buy a dongle for the first one? This would make it a universal DMX board


Power feeding to the board sounds good so long as you are only cascading up to about 10 of them.

Wombat

[JEC]

What is the design limit on the current pass-through for the pixels? Just for the record, how much current will the traces on the PCB support?

--

Phil

Traces on the current run, between RJ45 jacks, are .016" wide on 1 oz copper. There are 3 CAT5 wires for power and 3 for ground.

Limiting factor seems to be the CAT5 cable itself. When in doubt, double-power a chain (data + power on one end, power on the other end).

I wouldn't want to push more than 2-3 amps continuously through any amount of that cable. Depending on your lighting style, average power used by a pixel might be in the 60-80 mA each range.

[JEC]

I think this idea is great. I was thinking of getting some pixels and the DMX splitter sounds great.

Is it possible to have do a Vixen-485/232 to DMX conversion on it as well bypassing having to buy a dongle for the first one? This would make it a universal DMX board


Power feeding to the board sounds good so long as you are only cascading up to about 10 of them.

Wombat

I'm not quite sure what this would entail, but I'm happy to consider it. Can you elaborate and/or sketch something on a napkin for me?

[JEC]

Revision attached. No new board layout until I'm happy with the feature set.

I took out the MAX1480 chip (it's great but expensive in small quantity) and replaced it with an isolated DC-DC converter / optocoupler / RS485 receiver combination. Saved about $4 in the process. It's a 4 pin SIP package, very neat.

Current features:

* 5 pin XLR in & through jacks, USITT standard
* DMX over CAT5 in & through jacks, USITT standard
* Isolated / buffered DMX over CAT5 output, USITT standard
* PIC + DIP switches for parsing DMX input
* CAT5 output jack which merges DMX and pixel power - 'pixel' pinout

As mentioned above, the is designed to be fed by a +12v *switching* power supply, sufficiently beefy to drive your pixels. Figure 130 mA peak per pixel, and don't chain more than 32 pixels off one output (it will load the DMX transmitter too heavily).

------

Thoughts pending:

* Wombat's RS-232 suggestion.
* User could omit the PIC altogether, jump RB2 and RB5 together, address the pixels as normal and use the system as a simple optosplitter / power injector.



John

[DynamoBen]

Wombat's suggestion would require hardware but more importantly software. Some type of Vixen to Serial protocol would need to be created or the hardware would need to use an existing serial protocol. Food for thought.

[JEC]

Wombat's suggestion would require hardware but more importantly software. Some type of Vixen to Serial protocol would need to be created or the hardware would need to use an existing serial protocol. Food for thought.

Ah - okay. Happy to include the electronics and a programming port for the PIC. Beyond that, I don't think I can be much help. Just a regular RS-232 receiver? What sort of pinout and connector do folks prefer?

[wbuehler]

Cool

Let us know when we can order

Thanks

Bill

[DynamoBen]

Ah - okay. Happy to include the electronics and a programming port for the PIC. Beyond that, I don't think I can be much help. Just a regular RS-232 receiver?

FYI a PC's RS-232 port can't operate at 250K baud (DMX baud rate). You would need to create custom firmware for the device for this to work.

[NogginBoink]

Zoiks, that MAX1480 is an expensive part! It would triple the cost of an RGB pixel!

I did a little searching and couldn't find any isolated RS485 drivers for less than a few bucks.

Does such a part exist? Would it be economical to construct such an interface from discrete parts?

[JEC]

Zoiks, that MAX1480 is an expensive part! It would triple the cost of an RGB pixel!

I did a little searching and couldn't find any isolated RS485 drivers for less than a few bucks.

Does such a part exist? Would it be economical to construct such an interface from discrete parts?

Yes.

Usually I buy 1480s in sticks of 50, so the price gets reasonable. And I buy from Maxim, not a distributor.

That being said, there is an $8.40 way to achieve the same thing. I've been meaning to post it but have been swamped. Involves an isolated DC-DC converter in a 4 pin SIP package, an SN71576 and a 4N47 + glue.

[mrpackethead]

How about using the MXL1535E chip.. I was just looking through various options from maxim.. Never used on, but its listed at $4.58(1kprice).. Data sheet certainly looks like it coudl be the goer.. Its quite a lot cheaper than its cousin the 1480..


hey, and its a soic.. :-) (opps that won't be ok with the anti-smd brigage )


http://www.maxim-ic.com/quick_view2.cfm/qv_pk/4302

[JEC]

How about using the MXL1535E chip.. I was just looking through various options from maxim.. Never used on, but its listed at $4.58(1kprice).. Data sheet certainly looks like it coudl be the goer.. Its quite a lot cheaper than its cousin the 1480..


hey, and its a soic.. :-) (opps that won't be ok with the anti-smd brigage )


http://www.maxim-ic.com/quick_view2.cfm/qv_pk/4302

Looks like you just need to provide the transformer and a couple capacitors to make it run, yes?

[NogginBoink]

A transformer is ~$4.50, the chip is ~$11.25 in onesies. Then there's the capacitors and zeners. We're back to a really expensive solution. :(

[JEC]

A transformer is ~$4.50, the chip is ~$11.25 in onesies. Then there's the capacitors and zeners. We're back to a really expensive solution. :(

I'm at ~$8 with the DC-DC isolated converter, '75176 and 4N37 @ qty 100.

Peanuts on a $4K moving light. More substantial on an $9 pixel. :(

Maybe that's why many folks totally ignore the 'galvanic isolation' bit of the official spec.

[P. Short]

I think that a 4N37 (assuming that is not just a typo) is too slow for DMX speeds, a 6N137 would be a better choice.

--

Phil

[JEC]

I think that a 4N37 (assuming that is not just a typo) is too slow for DMX speeds, a 6N137 would be a better choice.

--

Phil

Sorry - you're right. I was thinking of something else while typing. 6N137 it is. Don't forget the decoupling cap!

[NogginBoink]

OK, I'm still wrapping my head around this whole isolation concept (reading AN-727 from Analog Devices now), but it appears that Analog's ADuM5240 has an isolated power supply and isolated data bridge built in, with very few external parts needed (only bypass caps), at $5.72 in onesies from DigiKey.

A hundred of 'em for $3.86/ea. I don't think it would require any additional components other than the '75176.

What function does the 6N137 have in your design? I don't see such a part in the schematic in the first post; and I don't quite fully grasp what the thread's morphed into... :)

EDIT: It looks like we'd need the ADuM5241 or 5242, not the 5240. Same prices, tho.

[mrpackethead]

A transformer is ~$4.50, the chip is ~$11.25 in onesies. Then there's the capacitors and zeners. We're back to a really expensive solution. :(

I'm at ~$8 with the DC-DC isolated converter, '75176 and 4N37 @ qty 100.

Peanuts on a $4K moving light. More substantial on an $9 pixel. :(

Maybe that's why many folks totally ignore the 'galvanic isolation' bit of the official spec.

Still $8 on a $9 pixel is probelmatic.. I think we are back to the isolating the string of pixels concept....

[DynamoBen]

The "gold standard" for DMX isolation is a combination of 75176, 6N137, and some type of power isolation (usually DC to DC transformer). Nearly all pro and entry level DMX equipment uses this combo. BTW 6N137 isolates data so at worst you will take out the 75176, and 6N137 but not your microcontroller.

[JEC]

Still $8 on a $9 pixel is probelmatic.. I think we are back to the isolating the string of pixels concept....

Yep. Did I post that schematic? I've been meaning to. Hang on a bit...

[Trepidati0n]

Anybody use something ilke this yet?

http://www.analog.com/en/prod/0,2877,ADUM5241,00.html

[JEC]

Anybody use something ilke this yet?

http://www.analog.com/en/prod/0,2877,ADUM5241,00.html

Oooh! That's even nicer.

Still need a '485 receiver, but Digikey only wants $5.27 for one.

[Trepidati0n]

I think you be getting gouged

http://www.analog.com/en/subCat/0,2879,767%255F830%255F0%255F%255F0%255F,00.html

[NogginBoink]

Anybody use something ilke this yet?

http://www.analog.com/en/prod/0,2877,ADUM5241,00.html

That's what I posed three posts up! :)

[Trepidati0n]

Knew I saw that somewhere. We actually use those in my field. They work great and are small. As a FYI...there is an isolated RS485 transceiver with built in power on the way...guestimate price is $5 in volumes of 1k. But still...one part :)

[NogginBoink]

I think you be getting gouged

http://www.analog.com/en/subCat/0,2879,767%255F830%255F0%255F%255F0%255F,00.html

The products we're talking about are in a different table, here:

http://www.analog.com/en/subCat/0,2879,767%255F827%255F0%255F%255F0%255F,00.html

[Trepidati0n]

I thought when he responded he was mention about $5 for the RS485....guess I need to learn to read. Since i'm not included in this conversation...I'll stay out now. Cheers.

[P. Short]

I see two difficulties with that part:

1) Very low DC output currents. It may not matter if you put all of the power consuming parts on the same side of the part as the power source, but would be a problem in many circumstances. For instance, the true-blue 75176 receiver cannot be powered through the part (although there are higher-priced, lower power alternatives around).

2) If I read the spec correctly, the component only passes data transitions through the data transformers, with moderately fancy logic to convert the edge pulses back to DC levels. Look carefully at the section in the spec titled "DC Correctness and Magnetic Immunity" It would be a very good part for manchester encoded signals and a lot of other encoding types, especially at high frequencies, but not so good for NRZ.

--

Phil

[Trepidati0n]

2) If I read the spec correctly, the component only passes data transitions through the data transformers, with moderately fancy logic to convert the edge pulses back to DC levels. Look carefully at the section in the spec titled "DC Correctness and Magnetic Immunity" It would be a very good part for manchester encoded signals and a lot of other encoding types, especially at high frequencies, but not so good for NRZ.

You are not reading the spec correctly. It doesn't require a constant refresh. What is happening is when you send and edge to the device it flips how it is chopping the power across which changes the output. on the other side We had the same concerns at first when we read the spec until we realized what was really meant.

10 mA should be more than enough for a modern RS485 transceiver. As log as you aren't putting on a super heavy load (i.e. stick with the 32 device limit), it should be more than enough.

[JEC]

Knew I saw that somewhere. We actually use those in my field. They work great and are small. As a FYI...there is an isolated RS485 transceiver with built in power on the way...guestimate price is $5 in volumes of 1k. But still...one part :)

Who is the manufacturer? That's really, really attractive.

[P. Short]

What is the data rate and encoding scheme that you are using?

--

Phil

[NogginBoink]

2) If I read the spec correctly, the component only passes data transitions through the data transformers, with moderately fancy logic to convert the edge pulses back to DC levels. Look carefully at the section in the spec titled "DC Correctness and Magnetic Immunity" It would be a very good part for manchester encoded signals and a lot of other encoding types, especially at high frequencies, but not so good for NRZ.


You and I must be reading it differently. I read the section:


The decoder is bistable and is, therefore, either set or reset by the pulses, indicating input logic transitions. In the absence of logic transitions at the input for more than 1 μs, a periodic set of refresh pulses indicative of the correct input state are sent to ensure dc correctness at the output.

to indicate that the decoder does ensure the correct DC output.

However, given my track record of being wrong lately, don't put much stock into my interpretation.

[NogginBoink]

Whoa.. the ADM2483/2486 are isloation devices with the RS485 transcievers built in. How cool!

$5.72/1
$3.86/100

http://www.analog.com/UploadedFiles/Data_Sheets/395735992ADM2483_b.pdf

AN-727 says "These devices reduce part count to one device, plus decoupling."

Sounds juicy to me.

[JEC]

Whoa.. the ADM2483/2486 are isloation devices with the RS485 transcievers built in. How cool!

$5.72/1
$3.86/100

http://www.analog.com/UploadedFiles/Data_Sheets/395735992ADM2483_b.pdf

AN-727 says "These devices reduce part count to one device, plus decoupling."

Sounds juicy to me.

Plus a second power supply, if I'm reading page 17 correctly... Nice price!

[P. Short]

OK, it looks like I mis-read it.

Part of that comes from my bias against wonder, all-in-one fancy parts available from only one vendor. My inclination is always to use bland, generic parts (especially if space is not a problem), so that there is less difficulty purchasing them in the long run.

--

Phil

[NogginBoink]

Whoa.. the ADM2483/2486 are isloation devices with the RS485 transcievers built in. How cool!

$5.72/1
$3.86/100

http://www.analog.com/UploadedFiles/Data_Sheets/395735992ADM2483_b.pdf

AN-727 says "These devices reduce part count to one device, plus decoupling."

Sounds juicy to me.

Plus a second power supply, if I'm reading page 17 correctly... Nice price!

Yeah, I just saw that. Rats. I was thinking it had the power supply built in like the others. I wonder how far off they are from releasing a device with iCoupler, isoPower, and RS-485 transciever in the same package.

This is a new arena for me. Does no company create a single-chip isolated power supply solution?

[DynamoBen]

I'm with Phil on this, all-in-one is OK but what do you do when they quit making it? I like 75176/6N137 combo, tried and true. Besides I have 75 of each on my bench. ;)

[JEC]

This is a new arena for me. Does no company create a single-chip isolated power supply solution?

Which brings us back to the MAX1480 I usually use. But it's expensive. And singly manufacturered. I've used hundreds of them without a single fault.

28 pin wide DIP (only) plus about 8 resistors and a decoupling cap.

I'm waiting for an order of the DC-DC converters (NKE1205SC from Mouser). 4 pin SIP package, fairly small footprint. Should be here Friday or so. 12V in, 5V out @ 200 mA. Add a 6N137 and 75176 and it's done.

I don't think there's a cheap way to do this. But I'm delighted at $8, because it's $4.50 better than what I've used in the past. And there's no bulky transformer to buy.

As a front end to a string of pixels, I think it's a great design. As a front end to a single pixel, not so practical.

But that's just me.

[P. Short]

I've been wondering about what the point of isolating a single pixel is anyway. It doesn't have an independent connection to the AC power line, and it's power comes in on the same cable as the data signals. So all you would be doing is providing galvanic isolation between several wires in the same cable.

--

Phil

[Trepidati0n]

The big point of isolation in DMX is that you have to assume the person using it is an idiot or extremely limited on really how electronics work. Therefore to limit headaches when setting up multi thousand light setups (like concerts)...isolation just removes a couple of variables that if gone wrong..can cause headaches that are very hard to figure out.

[DynamoBen]

Isolation is primarily there to protect the equipment from outside environmental conditions like lightning, and attached equipment failures. When you start daisy chaining things you can end up with a really high repair bill if one thing in line gets zapped. Most of the times I've swapped out damaged chips it was due to lighting, a 75176 is hard to kill.

[P. Short]

DynamoBen,

I'm not sure what you are trying to say.

--

Phil

[DynamoBen]

I think we may be talking about two different but related items. The only type of DMX device that needs true galvanic isolation is a transmitter. Most receiving devices have no need for that level of isolation since the line is “protected” by the transmitting device. However a receiving device will still use an opto to protect the electronics from induced spikes introduced on the line.

Does that make more sense? So full protection for a transmitter (this includes splitters) and “partial” protection for receivers.

[Trepidati0n]

Isolation is primarily there to protect the equipment from outside environmental conditions like lightning, and attached equipment failures. When you start daisy chaining things you can end up with a really high repair bill if one thing in line gets zapped. Most of the times I've swapped out damaged chips it was due to lighting, a 75176 is hard to kill.

Going to have to partially disagree with you. However your experience with consumer lighting and bad things and me dealing with aircraft in ilghtning storms might result in that difference. I still stand by my previous statement.

[mrpackethead]

Have we missed breaking ground loops as a reason for isolation?

[RJ]

None of my stuff connects the ground from the DMX as recommended. Should be no Ground loops. I see the isolation as a protection on the transmitter just as a means to protect the expensive equipment attached to it. (Computer, etc) there are a lot of people out there running the enttec open or the homemade version and there is no isolation in it. Doesn't mean I am advocating this I am not I just am putting it into focus.


RJ

[DynamoBen]

Have we missed breaking ground loops as a reason for isolation?

RJ is right as a rule of thumb pin 1 on a DMX transmitter is tied to Ground, however Pin 1 on the DMX reciever just passes thru to the next connnector.