High Speed communication over DC power lines ?

[nurquhar]

I have been mulling over what might be involved in getting the PIC's UART's to communicate over a common DC power line supply.

ie Say I have 4 PIC circuits run off a single 24VDC PSU via local DC-DC step downs. The PIC's UART would be hooked upto a modem IC that couples to the 24v line in some way.

I saw an older Topic in which the TDA5051A modem IC was discussed and proved to work. However the max speed was 1200 bps.

Is it possible to get the speed upto say 250kbps or something close ?

I have pair of Ethernet power line adpators which I use so the principle is not new .

To get the speed up I guess the transmission carrier wave is going to have to be pushed into the MHz or GHz range.

Might it be possible to use a Radio transceiver IC in some non standard way ? ie instead of hooking it up to a tuned ariel circuit its just hooked to the 24v line via a capacitor or something

[temtronic]

Have to ask, what kind of distance are you talking about ? Inches, feet or miles ?
I currently have remote SCADA stuff running on solid copper, using +-60 VDC for about 22 miles.
The trade off is speed, further you go, the slower it goes...

So do you really need the speed(250KB) ?

[nurquhar]

The application idea is individually addressable RGB LED units all strung on a two wire daisy chain. So upto 100m would be ok.

High speed is desirable to be able to do at least 25 updates per second of all the units on the string. There could be a 100 units on the line. I think DMX512 is somewhere close to 250kbps speed and uses it own RS485 type half duplex multidrop.

[temtronic]

So you really only need one way communication, that helps out.

DMX512 should work ,use low power rs485 chips( Maxim), good for 128 units on the bus.

Simple PIC12F683 ( 8 pin) good for RGB controller,lots of people use them for that.... even me.

You'll have to watch out for I2R power drop, design for worst case(100 units, fully on, with 2/3s near the far end(say 80-100M).

[SherpaDoug]

So you have only 2 wires for both data and DC power?

You could use a power H bridge driver to send a high power RS485 type signal over the wires. Use bridge rectifiers to extract the DC power and RS485 receivers to extract the data. If the signal will be 24V peak to peak, check that the RS485 receivers can take the voltage, or divide the signal down for those chips.

Be careful about the EMI this will generate, especially if this is for a commercial product. Using good twisted pair wire and a LC hash filter at the transmitter will help a lot.

I have done this before, but it was using shielded twisted pair and for use under seawater so EMI was not a problem for me.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[nurquhar]

Dear SherpaDoug

Yes I like your H-bridge driver idea as the concept is simple. I take your point about the EMI, square edge switching of a current of an Amp or two is bound to be noticeable. Especially at a few kHz.

Unfortunately my "underground" site will be a railway tunnel and not a trench in the atlantic. So to make it work I would need to kill the EMI emissions. What does a "LC Hash filter" circuit look like ?

Would using transistor rather than MOSFET H bridge switch help "soften" the "square" edges

What sort of bit rates were you able to get upto in your application ?

[SherpaDoug]

For a hash filter you want to use wideband hash chokes such as these:
http://www.bourns.com/data/global/pdfs/5200_series.pdf
with decent ceramic caps. I would start with a pair of chokes from the H bridge outputs, then a cap across. If necessary add a second stage. Make the LC cutoff frequency at least 10x the serial bit rate. You might find an H bridge driver with slew rate control. Either bipolar transistor or MOSFET H bridges can be made to work well or poorly. Neither is inherently good or bad for EMI.

It might be good to put a terminating resistor at the tail end of the line too. Use an R to match your best guess of the line impedance, and a series cap so the R doesn't consume DC. Make this RC time constant about equal to the LC constant of the hash filter. If possible, twisted pair wire will help a lot.

I only needed 1200 baud, and I got that with no problem.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[nurquhar]

Dear SherpaDoug

Is this the circuit you are describing ?

[SherpaDoug]

Yes, that is just what I was proposing.
_________________
The search for better is endless. Instead simply find very good and get the job done.

[ckielstra]

I like this discussion because I was thinking of a similar home brew project.

Instead of differential signalling my first thoughts were to use the technique as used in power line modems. Here a high frequency carrier is superimposed on the power supply lines. At the receiving end the high frequency is filtered out and then decoded.

The power line modem chips are available from multiple vendors, just do a search for 'Homeplug chipset'. This is a solution which has proofed to be working and has little EMC emission problems.
Biggest disadvantage as I see it is the price for these chipsets, starting around $11-USD

What do you people think would be the advantages of the RS-485-like solution when compared to the high frequency carrier?

[temtronic]

Obviously there's a huge price difference in RAW parts as you pointed out...not to mention PCB layout,R&D time,etc.

other stuff...
The 'carrier' version requires more 'overhead' from a programing standpoint as well as a bigger PIC for the 'brains',hence more $$$.
Also there is the time and labour to get a unit 'up and running'.Learning curve for the 'carrier' chip(s) is greater than simple 'serial' communications .RS-485 chips,etc. are second sourced,carrier stuff is usually one source.

A lot could depend on required throughput,speed, number of devices, etc. It's a tradeoff in time,resources(man hours),chips,etc.

I currently have remote SCADA equipment running 20+miles on solid copper,immune to EMI,hackerproof,etc. but at low speed.

[nurquhar]

I am quite enthused by the A/C switch power/signal concept. My only real concern is the EMI and developing in the "dark". What I mean by this is; are there any cheap/simple methods of measuring the EMI given off by a system. Having done a bit of reasearch on this about 15yrs ago I think that for proper EMI assesment you had go to a specialist company who hand an EMI measing chamber. Probably something beyond the resources of one man band.

Should I just get my cheap AM trany radio out and see how wild it goes ? Are there any "home" EMI measring techniques ?

[temtronic]

if this is going to be a commercial product,then I'd contact the local government department in charge of 'valdating' the product for the market.The equal to the USA 'FCC', CDN 'DOC' for the UK. getting 'type' approval may cost you more than what you can sell the prouct for ! In Canada building and selling 'home brewed' modems does require proper testing by the Feds($$$$) but I can use preavvoved, over the shelf, external modems freely !!!
Back to the original post, are you in control over the type of wires used for this system or do you have to use others? Need to know the specs of wire size and distance before proceeding.There's an old Motorola Application AN1089(?) that uses their Addressable UART(MC14....) in a classic 'signalling over the power line' technique,actually bidirectional.25 years ago I did a 'proof of concept' using them over 300' ( spool of wire).
The same idea would be real easy for a PIC !

[temtronic]

Search and download the Motorola datasheet for MC14469, page 6 of 10 shows how to get 'rectified power from dataline'.....

I've used a version of this for the past 30 years....sigh, time flies...

[Ttelmah]

Many things can be 'legal', but still cause radio problems. Current systems being used for data over power lines, will typically interfere on several of the amateur radio bands. As soon as you start sending signals along nice long bits of wire, you are inherently building a transmitter. Being careful to choose frequencies/encoding to reduce problems like this increases initial cost, but can reduce problems in the future.....

Best Wishes