Calling all clever folk (basically everyone except me then...)
I am working on a panel in a boiler room. They have these modules that communicate with some analogue temp sensors (0-10v) and transfer that to a main controller. That looks to be controlled by PICs.
This information is sent over a ribbon cable within the control panel. Bad idea there off the bat if you ask me.
They are unsurprisingly suffering with 0v not being 0v. Therefore the sensors are showing fluctuating values.
They are also suffering ground noise on their outputs to 0-10v valves, making them 'twitch'.
I have been asked if I can come up with a solution, which doesn't include ripping it all out, as that isn't an option.
One of these ribbon cables (run alongside all the 230v singles!) carries 6x 0-10v signals and a 12+ DC + GND pair. This one isn't too bad on the noise front, but it's not great.
The other appears to have 24v AC pair and 10x 0-10v signals in it. This one suffers greatly.
I have suggested stripping the 24v AC pair off the ribbon cable and trying to move that away from the signal lines and also moving the ribbon cables out of the cable management (away from the 230v).
My other thought wa maybe some ferrite coils over the ribbon cables at each end? I have had random luck with these before, but no idea how you work out the required impedance/frequency.
Any ideas or guidance would be great. Not getting paid for this I should add, its the boiler room in a local museum #Zerocash
you can put a shielded ribbon cable in place for the one that doesn't self-contaminate. Ground the shield at the signal source end, usually "in the field".
the cable with the self-contaminating 24VAC needs work. As a first attempt, you might simply separate the wire pair with that signal, and try shielding the rest of the cable, leaving both connection ends as is. This would involve wrapping the remaining wires with foil, and adding a ground connection to the foil at one end. But it's really not ideal, it would be stopgap.
The better solution is to remove that pair from the equation. snip the offending wires at the source end(being careful to only cut the two). Now, shield that cable.
Run the 24V signal as a separate cable, via a separate route.
To this, you might be able to add RC filtering at one end or the other, by creating a small plugin board and filtering the signals appropriately there. Total cost unlikely to exceed $100, if you know what you're doing.
Good luck. These problems are vexing.
Ribbon cables must have a ground wire in each side of the signal wire. Are the connectors the positive insulation displacement type? If so, they may have been installed improperly and at least one wire is not making a firm contact with the wire.
If noise is still a problem, there are ribbon cables with twisted pairs, but that is quite rare.
You can quickly add a shield to the cable with a braided wire shield. Ground only one end of the shield.
I would suggest making up copies of the cables.
Then do any modifications/experiments on the copies.
If you then come up with a solution, all is well and good. However if you don't come up with a solution then you can always revert back to the original configuration.
Only if the designer intended it that way. If the equipment wasn't intended to be 'exposed', only within a chassis for example, then using half the wires for ground wouldn't be likely.
Be very careful with "must have". It might be a good design practice, but it's not a given.
Thanks for all the ideas. I don't get the impression this was all that well thought out.
The ribbon cable is cores 1&2 as the 24v AC, and then the signal lines 3 upwards.
Looks to me like the DC ground is literally attached everywhere at every board.
The first board has a varistor across the 24v AC supply. These look like simply PC type 20 way ribbon cables. Certainly not what I would put in a 230v control panel full of noisy contactors.
The ribbon cables control 10v relay which in turn operate the 230v coil contactors. There is a lack of any suppression anywhere.
Tomorrow I might try and draw out the system on paper and see if I can get a plan together.
I did find some screened ribbon cable. Maybe I will try a length of that as a replacement lead.
Ideally, I would rip it out and start again... but that is just a can off worms and there is no money attached to this.... so no!
Hmm, the plot thickens
In the cabinet I found a load of old drawings and I think this is the ribbon cable (J19 = Ribbon connector)
I drew the colours on as it was difficult to work out what went where. But, the drawing shows the connections different to the pins on the PCB. If you test out the PCB pins with a multimeter, they are arranged as the (poor) drawing attached, not as the schematic drawing shows.
The blue line goes off and on this board is simply the ground connection to 8x leds that are controlled by the 8x signals. But that blue GND isn't connected on this board.
So... yea. Not sure how the GND gets to this modules 'leds'. I assume it is connected on one of the other boards. This 20 way ribbon seems to be the common method of interconnection.
Not being super user of AC, is that all they are using to 'smooth' the supply. I believe the LM46001 is a programmable step-down IC.
I assume the ES2B-13 diode is preventing those 1000uf caps going pop.
So options.
Mew leads is actually a nightmare. They are all over the place and the screened cable is damned expensive.
Make a small 'intermediate' PCB with a ribbon cable male + female connector with some further noise suppression onboard? No idea what, but Google is there for a reason.
How are you measuring the difference in 0V points.
If power is being transferred along the ribbon cable, the gauge of the wire may be too small for the current.
Good idea, and the 12V as well.
Use a decent cable for power.
My cheapo oscilloscope has been the test point.
It's a HORRIBLE thing to work on. A right mess.
I actually think a lot of the interference is in the signal lines. Not the ground.
The ground isn't great, but not really sure how to test that efficiently. The runs are pretty long between these panels and the resistance seems a bit high on some of the runs between the grounds of various boards.
I would not have used the ribbon cables at all, but I certainly would have run the 24v supply lines in something a little more robust.
My thoughts are that they have noise on the 0-10v signal lines induced from the 230v cables that go all over the place and not so much the ground (as they originally told me they were suffering with).
I think they have incorrectly identified the issue as a noisy ground, when actually its the 0-10v signal that is fluctuating.
I will continue to poke in the cabinet (I am an electrician by the way, so I know what I am 'poking')
Trace the grounding, ensure there aren't multiple paths(there likely are). That, in itself, can introduce significant noise, even without a 230VAC noise source.
Without the ability to introduce filtering, no shielding, and without rerouting both the 24VAC and 12VDC, you're reduced to rearranging the ground to see if it magically fixes the problem.
Good luck.
Well every ribbon cable effectively has two sets of 3 grounds
Not sure I understand that logic. I though maybe one was for the AC and the other for DC, but they seem to be joined at random places.
Still think some ferrite coils over the ribbon cables may help (even slightly) as a simple assistance, but not sure what frequency/spec to try
Probably, you're beat before you start then. A "designed system" would more likely have one star ground point for analog and digital, and the AC side would be completely isolated.
You can fritter around the edges, and you might get lucky and make a change that results in a significant improvement, but it could take weeks to get there - or it might happen today. Ferrites won't hurt, so go ahead.
Good luck.
Yes. This is a freebie job, so I am limited in what I can achieve and how far I want to go.
It's clearly a poorly designed system.
I am not sure of the benefit of them calling all the 2 pin connections to the boards 24v AC.
There appears to be nothing that requires AC and I assume all the boards are half wave rectifying that 24v AC (I see no bridge rectifiers anywhere, just a single diode on one side of the AC inputs to the boards with elect caps after that).
It might be worth reading the AC ripple on the 1/2 wave rectified DC.
Measure the DC with DMM in AC mode.
Could be insufficient filtering or stuffed electrolytic capacitors allowing mains ripple on the "DC"
