High speed data transfer in adverse conditions - how's it done?

[Mark R]

I was thinking about how a modern CT scanner works, and simply can't work out how the data gets off the detector and to the rest of the system.

A contemporary scanner may have 60 kpixels on its sensor. Each pixel is digitized at full 24 bit depth, and the sensor operates at 3 kfps.

The problem is that the detector is mounted on a rotating gantry about 6 feet in diameter - which can spin at up to 180 rpm continuously (so no cables). The data needs to come off in real time - and must run along side power input to the gantry (3 ph, 400 V, 200 A).

So, your mission, should you choose to accept it, is how do you reliably transfer 5 Gbps over a rotating connection, along side a very heavy duty power supply?

The enabling tech is just referred to as 'slipring technology'. But as far as I can see, you need more than just common-or-garden sliprings to carry a 5 Gbps signal without trashing it.

 

[TuxDave]

Send the signal differentially so that the common mode noise will cancel out at the receiver? There are numerous papers on differential receivers accepting common mode voltages up to the 100's of volts.

If not then I guess you have no choice but to modulate the data on a carrier frequency higher than the interfering frequency (60Hz interference?) and filter/demodulate at the receiver.

 

[Mark R]

The precise scenario I'm thinking of... Pic

Any ideas how you get the data across that moving connection? Surely, you can't do this electrically - the noise from the brushes will kill things. And, it'll be random, so you couldn't simply reject a common mode voltage. The tangential velocity of the moving part is something like 40 mph, so you'd get loads of really high frequency noise - not what you want when you need a clean 5 Gbps connection.

And your data is already 5 Gbps - what sort of carrier would you need if you wanted to transmit a modulated version? 60 GHz? Does anyone routinely use such high frequencies?

I'm totally baffled.

 

[MrDudeMan]

why not use RF transmission?

Text

 

[Matthias99]

Originally posted by: Mark R
The precise scenario I'm thinking of... Pic

Any ideas how you get the data across that moving connection? Surely, you can't do this electrically - the noise from the brushes will kill things. And, it'll be random, so you couldn't simply reject a common mode voltage. The tangential velocity of the moving part is something like 40 mph, so you'd get loads of really high frequency noise - not what you want when you need a clean 5 Gbps connection.

And your data is already 5 Gbps - what sort of carrier would you need if you wanted to transmit a modulated version? 60 GHz? Does anyone routinely use such high frequencies?

I'm totally baffled.

If you're allowed to have the data connection come out what would be the "z" axis in your picture, you could attach an optical cable to the rotating part and have the connector at the other end be able to rotate (the optical connector doesn't have to be rigidly coupled). That's a horribly unclear explanation, I know, but there's a way to do it mechanically. Then you just use multi-mode fiber to carry the connections; 5Gbps is definitely doable over a few FC links.

Otherwise, I'd think you would almost have to use some sort of wireless link. I don't think you could get a good electrical connection to something moving that fast, and the power levels involved would generate large amounts of EMI.

 

[CycloWizard]

Originally posted by: Matthias99
If you're allowed to have the data connection come out what would be the "z" axis in your picture, you could attach an optical cable to the rotating part and have the connector at the other end be able to rotate (the optical connector doesn't have to be rigidly coupled). That's a horribly unclear explanation, I know, but there's a way to do it mechanically. Then you just use multi-mode fiber to carry the connections; 5Gbps is definitely doable over a few FC links.

Otherwise, I'd think you would almost have to use some sort of wireless link. I don't think you could get a good electrical connection to something moving that fast, and the power levels involved would generate large amounts of EMI.

:thumbsup:

Our department runs lots of OCT experiments. Most of them involve many very thick wires (they're pretty dated), but high-speed CT can be performed with multiple wireless connections or axially-mounted cables.

 

[Mark R]

I believe I have found an answer, albeit somewhat detail impaired.

Schliefring brochure (See pages 13 & 14).

 

[Kelemvor]

I work somewhere that makes and trains on CT scanners. Woudl you like me to go ask some of the people how they work? If so, let me know exactly what you want to know.

 

[Mark R]

Originally posted by: FrankyJunior
Woudl you like me to go ask some of the people how they work? If so, let me know exactly what you want to know.

I'd be very interested to find out.

Essentially, the question is, 'How do the slip-rings permit transfer of data at gigabit speeds?'. Is the transfer optical, electrical, or capacitative? And How can you maintain signal coherence and quality over such a large slip-ring?

 

[Kelemvor]

OK, I got some info but it's over my head. I've put it in a Word document. If you want to PM me with an email address I can send it to you.

 

[tafate]

is it possible that it functions just like a mic cable where instead of sensing on and off it flips one of two parralel lines (that doesnt carry anything except interference) so that it would be the inverse, then combines them and the inverse cancels out the interfence in the signal line and just leaving the pure signal?
or it might work like a sata drive

 

[Kelemvor]

I gave the info I have to Mark. If he wants to summarize it and type it in here, I'll let him do that as I honestly have no idea how they work. I just happen to work with people who do.

 

[Mark R]

It's capacitatively coupled across the moving join.

Essentially, the circumference of the rotating ring and a stationary detector form a parallel plate capacitor, across which the signal can be transmitted. (Alhtough capactitors can't conduct DC, they can conduct AC). Very careful attentional to impedance matching of the conductor on the ring circumference and proper termination are essential to maintaining signal coherence at all points on the ring.

Although it didn't mention anything in the document, I wonder whether you would get any Doppler shift in your signal - the linear circumferential speed on a modern CT scanner can get up to about 40 mph.

Differential signalling (like the microphone example above) is used to obtain good signal-to-noise.

I'lve drawn up a simple diagram which should illustrate the salient points: ct3.png

 

[Mark R]

Just thought this might interest some people and just illustrate the problem I was talking about.

Pic

Video (Quicktime)