olddumbandnew
Posts: 16
Joined: Mon Apr 08, 2019 2:47 pm

splicing FFC cable GNDs

Sun Jun 02, 2019 11:30 am

Hi,

Can anyone think of any reason that terminals 1,4,7,10 on the FFC cable of the camera can't be joined so I can use a 12 wire slip ring.

I Have FFC connectors so I was going to plug the cam into a concetor, link all the grounds (1, 4, 7 & 10) to a single wire of the slip ring then un-link them back to another FFC conector/cable (ie, 15 pole Cam FFC -> 12 pole slipring -> 15 pole FFC Pi socket).

I'm hoping to make a swivel cam using a servo (hence the need for the slip ring)

Cheers in advance

aBUGSworstnightmare
Posts: 1491
Joined: Tue Jun 30, 2015 1:35 pm

Re: splicing FFC cable GNDs

Sun Jun 02, 2019 6:56 pm

Shielding of MIPI-CSI signals....

olddumbandnew
Posts: 16
Joined: Mon Apr 08, 2019 2:47 pm

Re: splicing FFC cable GNDs

Sun Jun 02, 2019 7:37 pm

aBUGSworstnightmare wrote:
Sun Jun 02, 2019 6:56 pm
Shielding of MIPI-CSI signals....
But as far as I can see from the schematics of the Pi cam and Pi itself, they're a common ground anyway.

https://www.raspberrypi.org/documentati ... a2_2p1.pdf

https://www.raspberrypi.org/documentati ... educed.pdf

So all I would be doing is making them common ground on the loom (which the FFC cable doesn't)

6by9
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 7282
Joined: Wed Dec 04, 2013 11:27 am
Location: ZZ9 Plural Z Alpha, aka just outside Cambridge.

Re: splicing FFC cable GNDs

Sun Jun 02, 2019 8:29 pm

Inserting a ground between each of the pairs reduces crosstalk between them.
Csi2 runs at several 100 MHz, and each data or clock signal should be treated as an impedance and length matched differential pair.
You may get away with abusing the signals, or you may get random corruption. User beware.
Software Engineer at Raspberry Pi Trading. Views expressed are still personal views.
I'm not interested in doing contracts for bespoke functionality - please don't ask.

olddumbandnew
Posts: 16
Joined: Mon Apr 08, 2019 2:47 pm

Re: splicing FFC cable GNDs

Sun Jun 02, 2019 9:33 pm

I've got spare cams but not spare Pi's. :shock:
Perhaps I could ground 1 & 4 and seperately 7 & 10. I guess there is only one way to find out. (I might buy a Zero to test it - cheaper if it all goes pear shaped)

olddumbandnew
Posts: 16
Joined: Mon Apr 08, 2019 2:47 pm

Re: splicing FFC cable GNDs

Sun Jun 02, 2019 9:38 pm

6by9 wrote:
Sun Jun 02, 2019 8:29 pm
Inserting a ground between each of the pairs reduces crosstalk between them.
Csi2 runs at several 100 MHz, and each data or clock signal should be treated as an impedance and length matched differential pair.
You may get away with abusing the signals, or you may get random corruption. User beware.
Hmm...thinking about it, common grounding 1 & 4 and 7 & 10 should work. The distance change shouldn't be much (if any) different to the distance change from 1 - 10 on the Pi Zero FFP reducer cable so at 100MHz, I doubt there would be any noticible difference in anything (unless I was running at 100+ FPS)

olddumbandnew
Posts: 16
Joined: Mon Apr 08, 2019 2:47 pm

Re: splicing FFC cable GNDs

Sun Jun 02, 2019 9:52 pm

& the impedence on the FFC cables is going to be WAY more than the loom of a slip ring. I'll suck it & see. Success or fail post to follow.... :? :? :? :?

6by9
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 7282
Joined: Wed Dec 04, 2013 11:27 am
Location: ZZ9 Plural Z Alpha, aka just outside Cambridge.

Re: splicing FFC cable GNDs

Mon Jun 03, 2019 6:39 am

Impedance matching, not just impedance. At rf frequencies cables will have what is called a characteristic impedance, with typical values being 50ohm (rf transmitters) , 75ohm (composite video cable), 90, 110, or 120ohm (USB and csi/dsi use these latter values. See the compute module schematic for which as I can't remember) .
Any point where your cabling changes characteristic impedance results in a reflection of all or part of the signal. If the distance between reflections happens to be related to the wave length of the signal then you get standing waves.

Length variation may mean that the clock signal may be arriving earlier than the data, therefore the data isn't stable when latched. If you look on the back of the boards you'll often see pairs of tracks doing wiggles - this is so that the length of the pair is matched to the other pairs in the link.

And I said several hundred MHz, not just 100Mhz. In theory it can run at up to 1Gbit/s ddr, so 500 MHz link frequency. With one device I am running at 472MHz link frequency, so pretty much at the max. I can't remember the number for the camera modules off the top of my head.
Software Engineer at Raspberry Pi Trading. Views expressed are still personal views.
I'm not interested in doing contracts for bespoke functionality - please don't ask.

olddumbandnew
Posts: 16
Joined: Mon Apr 08, 2019 2:47 pm

Re: splicing FFC cable GNDs

Mon Jun 03, 2019 8:51 am

Cheers, I've got you now.

It seems doable (if not a little fiddly) for everything except the common grounds since I'd be effectively making the common ground in the middle of the whole thing. If I use short FFC at each end I can match the impedence in the slip ring loom accordingly but matching the terminals 1, 4, 7 & 10 could be a problem. I'll have a further head scratch.

I reckon it would still be doable if I knew the impedence's between 1, 4, 7 & 10 (on the Pi) and ground itself. (although would it matter....so long as the impedence between each terminal is matched?)

6by9
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 7282
Joined: Wed Dec 04, 2013 11:27 am
Location: ZZ9 Plural Z Alpha, aka just outside Cambridge.

Re: splicing FFC cable GNDs

Mon Jun 03, 2019 9:29 am

Checked the schematics - DSI and CSI are 100ohm differential pairs.

TBH this is all down to best design principles. Where designing products that need to pass QA tests, you do everything right. It also reduces issues when doing EMC testing. If you're doing a one off, then suck it and see. Abide by the principles where you can, but don't obsess over them.
You asked the question over why there were multiple grounds, but feel free to try it out with fewer.

See viewtopic.php?f=44&t=195139#p1221456 for one of my lash ups that works quite happily, however it is only running at 432Mbit/s, or 216MHz link frequency. The SMA cables were cut as accurately as possible to make them the same length but are going to be a number of mm different, and the dodgy soldering onto the camera board isn't going to be perfectly impedance matched. It works, but I wouldn't look at selling a product built like that.

I vaguely recall a very old story over the design of two CPUs by different manufacturers. One radiated like anything and ran hot, whilst the other ran significantly cooler and didn't radiate anywhere near as much. The difference was that one had used sharp 90deg bends in the tracking, whilst the other had nicely rounded corners. (I want to say this was the early Intel x86's vs Motorola 6800, but that could be totally bogus).
Software Engineer at Raspberry Pi Trading. Views expressed are still personal views.
I'm not interested in doing contracts for bespoke functionality - please don't ask.

olddumbandnew
Posts: 16
Joined: Mon Apr 08, 2019 2:47 pm

Re: splicing FFC cable GNDs

Mon Jun 03, 2019 9:53 am

Thanks for all that. I would try it with a Pi Zero (for economic safety) but thinking about it, the smaller FFC connector is a problem just from a physicam point of view. I'll give it a go - although it would seem I have a long wait before I can (4 week delivery time on the slip ring).

Wouldn't it be cool if there were 15 terminal slip rings or, even better some kind of 15 terminal FFC slipring.

I'll try and remember to post the outcome - gives me a few weeks to hone my soldering skills :D

(the alternative is to rotate the whole Pi with just a few other inputs needing to go via a slipring - I don't need power to go that route as its all solar/battery powered - it would just be a bit cumbersome)

Thanks again

Return to “Camera board”