Continue to Site

Eng-Tips is the largest engineering community on the Internet

Intelligent Work Forums for Engineering Professionals

Register Log in
  • Congratulations The Obturator on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!

Noise Abatement

Status
Not open for further replies.
itsmoked

itsmoked

Electrical
Feb 18, 2005
19,114
Hi all!

I have been called in on a product that is having some difficulty that needs a solution.

Essentially it is an OT-Shelf video camera/display screen that gets bought, disassembled, and reassembled into a different form factor.

Well the fly-in-the-ointment is the original camera head (size is about 1/2 a walnut) was plugged directly onto the main board with a tiny ~20pin connector. the new geometry results in the camera head having to be about 6" away. The 'uninformed solution' was to "just put a cable in". A ribbon cable resulted in nasty video artifacts that rendered the unit useless.

The next attempt was to run individual wires, twenty of them. This was marginally better as with all those itty bitty wires ungrouped the system works well. As soon as you gather those wires into a bundle things go sour. The display which is fed from data received over these wires and subsequently processed into what the LCD displays starts showing horizontal lines of non-sense spread over the video image. As you squeeze the wire-bundle together this gets much worse - as you spread it out it gets much better to the point that all the defects disappear. Of course the assembled product needs them mostly bundled.

I know what I'm going to do to try to solve this but I would really like your informed input and hunches before I join battle with this thing. Note: there are no schematics or even any information available for the ICs used. I'm flying totally blind on this.

Keith Cress
Flamin Systems, Inc.-
 
Replies continue below

Recommended for you

Sort by date Sort by votes
Hi Keith;
A hunch;
How about removing one wire at a time from the bundle? If you can identify the problem wire/circuit you may go for a shielded cable for that circuit and ribbon cable the rest.
Respectfully
 
Upvote 0 Downvote
Hi Keith,

Bill has already put you on the right track (use shielded wire for the sensitive signals and unshielded for the rest).

To identify the sensitive wires, that need shielding, use an oscilloscope to probe each signal. Those who look analogue are the ones needing special treatment. Pure digital signals seldom need shields. An added bonus with shields is that they make the ground connection a lot stiffer - yes, solder to ground at both ends. Don't listen to those know-it-all guys saying that shields shall only be connected one end.

Good luck! Sounds like a nervous job. Not much space there, I imagine (no pun intended).

Gunnar Englund
--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...
 
Upvote 0 Downvote
If it is analog video, the noise itself (on the display) is usually quite informative. For example, ghosting distance might roughly match the electrical two-way length of the cable and you'd be able to guess there's a mismatch and reflections (6" is so short that this example probably doesn't apply in this case). Pulses of noise will match - timing to frequencies. To make this work, you need to be informed of the video standard - this can usually be determined when you find the H avd V sync pulses and note their timing characteristics.

If it is analog video, then it is likely that the video signal is broken out onto several pins with the H amd V sync pulses on their own wires. Concentrate on shielding the one analog video wire first, the sync pulses are pretty rugged. Keep any other digital signals (noise) away from the primary analog video.

If it is colour, then you might have R, G and B video signals, so you'd need to concentrate on those three first.

If it is some sort of proprietary all-digital interface, then you simply need to keep the pulses on one wire away from the pulses on the other wire.

PS: Our on-staff 'know-it-all guy' sometimes recommends grounding the shield at one end only on the SECOND (outer) LAYER of shielding (if appplicable).

I suspect that a roll of RG-174 should be able to sort the whole mess out - for your prototype at least.

Good luck.
 
Upvote 0 Downvote
I'm going to say this is a CCD-style setup, which means high-speed digital signals... the 20 lines says to me it's a parallel transfer (at least 8 bits, maybe 16). The CCD is going to need a clock signal, so that's going to be the fastest rise/fall line in the group. If that leaks over into other lines you'll get random noise. If other lines leak over into the clock line, you'll get missed pixels, shifted lines, etc. There may or may not be differential lines, as well, so if you can find those pairs and twist them together you'll be doing better. Noise reduction is the obvious savior here, and it may come from a simple shielding of the correct lines, as previously mentioned.


Dan - Owner
Footwell%20Animation%20Tiny.gif
 
Upvote 0 Downvote
Most cameras that small are CMOS, such as the OV9650:

The smallest package is a 22-pin deal, consistent with the number of wires indicated. These guys are pretty cool, fully self-contained w.r.t. A/D and direct digital video output. Note that the pixel and other clocks are generated off-chip, hence there are issues with ensuring clock integrity and isolation.

TTFN
 
Upvote 0 Downvote
"camera head having to be about 6" away"

Definitely a "ground both ends" thing.

Gunnar Englund
--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...
 
Upvote 0 Downvote
Wow guys thanks for all the ideas. I don't feel adrift and alone on this 'Lost Dutchman' of a ship.

Last night I just got the system functioning as all the cables to the different parts and the battery have NO indicators as to polarity. I had to painstakingly analyse everything to even figure out how to plug the mess together.

I started by finding the only two identifiable devices; 74LXC244. From that I was able to deduce which of the four battery prongs was the ground. I was also able to find the (3) lone grounds in the 'problem' cable to the camera head. Their pattern allowed me to find the same pattern on the camera head and thus plug everything together and turn it on without letting the smoke out.

The Cable has in-fact 36 wires in it!! They terminate on the main board near the 244s but then so does the 20 pin ribbon that runs the display so I am not yet sure if the 244 run the camera or the display or possibly one each. I will check that soon.

Other info. The 'Problem' looks like this. Picture a VGA screen about 2x2 inches.

Perhaps of the 640 horizontal lines that are showing my brown desktop 40 will instead be blue lines that cross the display horizontally. They mostly only span about 1/2 to 3/4 of the display, always starting from one side.

To my surprise they change dramatically with what the image has in it. Jamming my finger into the image may cause a huge multiplication of the bad lines and they may all run out to my middle-of-the-screen-area finger image. They then track my finger.

The camera head appears to have NO other IC on the board only about 40 passives. (Res and caps) so I think this is fully digital info washing across the cable. Keep in mind this handmade cable is 36 of the smallest diameter wirewrap wire you can get ~30AWG. I don't think there are any analog signals but I am still not sure. Still wondering why you would need that insane number of wires, 20 makes more sense to me but 36?

I suspect IRstuff might be right and the pixel clock is in that mix.. <shudder>

Anyway.. I will be on it again in an hour or so and will front any new info. Thanks for all your great points they are helping me gel things in my head.

Keith Cress
Flamin Systems, Inc.-
 
Upvote 0 Downvote
Grounding at both end means that your "shield" is now a ground path. Therefore, ground currents WILL flow, and can get coupled onto your sensitve signals.
 
Upvote 0 Downvote
Kieth, at #36, there may be multiple power wires in parallel.
Gunnar, I thought you were one of those advocating ground one end only, always.
Just joking friend. I know that normal grounding practice for power circuit shields doesn't apply here.
Respectfully
 
Upvote 0 Downvote
Keith,
What about an overall shielded cable, similar to an IDE cable for a PC? You could go to a custom cable shop to have the proper connectors put on.

xoxide_1929_137176413


JRaef.com
"Engineers like to solve problems. If there are no problems handily available, they will create their own problems." Scott Adams
For the best use of Eng-Tips, please click here -> faq731-376
 
Upvote 0 Downvote
Point taken melone.

waross; Yes, I think that's why I see 3 GNDs and (2)3.3V

Jraef; I'd love to use that stuff but I think it is way to big. (36)wires in a connect 0.375" long.[cry]

Here is what a survey of the signals comes up with.
Note the wires marked GND can be ohmmed out to actual ground.

The ones marked GND? Cannot be - but never show any activity!!

(12) DATA I haven't actually seen a CLK unless one of the sine waves is it.

2wbynow.jpg


I'm considering building a cable without the GND?s present to see what happens... Getting rid of 14 wires could help with a solution if they where just part of a connector for mechanical structure on the original configuration.

Keith Cress
Flamin Systems, Inc.-
 
Upvote 0 Downvote
Okay I have determined by microscope and jewelers tools that the camera head is a cmos camera.

Specifically a Micron MT9V403.

Now I have more info on the parts. Clearly it is running in full color VGA mode at 24Mhz CLK and subsequently at 70 frames/sec.

The cable should be totally impedance controlled in my opinion and is not, being made of Since the actual video defects are so specific I am still hoping that perhaps the CLK is the biggest problem. Knowing which pin it THE clock from the data sheet will now let me hop from the imaging device back to the cable and confirm which is the clock and for that matter, why there seems to be two of them. The (2)24Mhz sine waves.

Keith Cress
Flamin Systems, Inc.-
 
Upvote 0 Downvote
melone,

This is one of the "eternal truths" I was warning for. Having GND connected via a (thin) wire and return currents going through that couples noise by transformer action between wires a lot more than having GND connected via the shield. Noise current flows on all sides of the wires inside the shield and produce no net magnetic field that can induce noise in internal conductors.

A shield is also a low impedance conductor that will reduce common mode voltage between the two subsystems to a minimum compared with having one (thin) wire for ground connection.

My comment about 6" - a clear ground both ends stands. I do a lot of trouble-shooting on high-power PWM systems where signal components go close to 100 MHz and I have seen this mythical "ground one end only" cause many problems. And also some good money for yours truly.

Gunnar Englund
--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...
 
Upvote 0 Downvote
The multiple GND lines are there to ensure signal integrity between all of the data/clock lines. When creating a flat cable, you should see alternating GND and data signals, which has the effect of all GNDs ending up on one row of the connector and all data/clock on the other row. Controlled impedance won't be so much of a problem at 27 MHz as signal integrity. If the clock signal looks like a sine wave, either your scope has too low of a bandwidth or your probe is loading the driver. If you determine it's not the probe loading the circuit down, any small perterbance in the system will screw up the clock and show you artifacts, such as the blue lines you're currently seeing, so again I suggest having a look at the clock line and/or diver.

Dan - Owner
Footwell%20Animation%20Tiny.gif
 
Upvote 0 Downvote
100MHz sounds pretty fast, but what are the edge rates, i.e. rise/fall times? Edge rate, not frequency, is the driving factor in the amount of noise generated in a system. 100MHz sin wave is drastically different than 100Mhz square wave with 750ps rise/fall times, right?
 
Upvote 0 Downvote
"PWM systems where signal *components* go close to 100 MHz "

Spectrum of these signals goes well above 30 MHz. That is close to 100 MHz in my world. I am referring to MOSFET power stages. IGBT spectra seldom reach more than 10 or 20 MHz.

Gunnar Englund
--------------------------------------
100 % recycled posting: Electrons, ideas, finger-tips have been used over and over again...
 
Upvote 0 Downvote
As an update. I built a cable (six hours under a microscope) and when I was done the camera worked flawlessly... Tried their cable and it worked flawlessly too, now!!. So they brought me another main camera board proven to have the problem. We plugged it in with their cable and demonstrated the horrible problem again. We switched to my cable and it worked flawlessly... They rushed off with my cable to try it on other screwed up units. They all worked correctly except one...

The gross difference between my cable and theirs is that:
1) I used colored wire..
2) They potted the ends of theirs and mine was for my exploration so I didn't pot mine.

Q: You guys think that clear epoxy potting would increase the inter-wire coupling capacitance?

Keith Cress
Flamin Systems, Inc.-
 
Upvote 0 Downvote
Skogs,
Point of technical merit. You say:
>Noise current flows on all sides of the wires inside the shield and produce no net magnetic field that can induce noise in internal conductors.

Not true. An inner conductor has mutual inductance with the coax shield. Therefore the shield has the same mutual inductance with the inner conductor. This is an application of the reciprocity theorem. I know it is impossible to get your head around how the outer couples to the inner, but you have to believe in the reciprocity theorem!
 
Upvote 0 Downvote
Status
Not open for further replies.

Part and Inventory Search

Sponsor

Back
Top Bottom