Author Topic: About ribbon cables at fairly high frequencies.  (Read 6784 times)

0 Members and 1 Guest are viewing this topic.

Offline SpatterlightTopic starter

  • Newbie
  • Posts: 1
  • Country: se
About ribbon cables at fairly high frequencies.
« on: July 23, 2020, 12:38:55 am »
Hi, I just tried to google the answer to this, but it seems to be ungoogleable.

From a practical point of view, how high frequencies can be transferred via a ribbon cable?  Let's say the common 1.27 mm type. I made some calculations on my SPICE software on transmission lines, but I'm not sure whether a ribbon cable can be regarded as a transmission line.

Really, the frequency isn't that important, it's how the transients will propagate that limits the length, I think.

In my particular case, I need to transfer bitrates at a maximum of 12Mhz. Can someone perhaps estimate the maximum length of the cable?  I'm using it with a ground wire inserted between the signal wires.

Also, I intend to insert resistors of some 22 ohms at the TX point.  And the receiver is a MCU that has some 5 or 10 pF on the inputs.

For me, 50cm ( 20 inch ) will be perfectly alright. Actually, it's a I2S interface, but it's insensitive to jitter. The only importance is that there are no bit errors. Possible ringings must have subsided after a maximum of 30ns.

Thanks
 

Offline helius

  • Super Contributor
  • ***
  • Posts: 3688
  • Country: us
Re: About ribbon cables at fairly high frequencies.
« Reply #1 on: July 23, 2020, 12:54:39 am »
Unshielded ribbon cable may have common-mode noise immunity issues, which the twisted-pair type cable (AMP Spectrastrip) was designed to address. It's possible that the twisted pairs self-inductance limits the risetime, I'm not sure. Some ribbon cables used to connect computer busses were also sandwiched between aluminum shields.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: About ribbon cables at fairly high frequencies.
« Reply #2 on: July 23, 2020, 01:00:45 am »
I'd seriously reconsider it as a medium by a few GHz.  Full shielding would probably be needed below that point.

SATA cables are basically full-shielded ribbon cables made of better-performing materials.

As for range vs. frequency, given reasonable drivers and receivers -- PVC insulation isn't fantastic, so don't expect any miracles.  At 12MHz, and average logic type transceivers, it's probably 10s of meters, maybe a bit over 100.  Further than you would ever want to run a ribbon cable (for a number of reasons)... ;D

12MHz, that's comparable to the bandwidth of old school hard drives -- cool fact, the ST506 interface used single-ended (TTL level) control signals, chained between one or two drives and terminated with a resistor pack in the control board and in the last drive; and differential (RS-422 level) data signals (from/to the head amplifier, point-to-point -- no split cable allowed), terminated with 120 ohms or so.  It was high tech in 1980. :D

I would be more concerned about the EMI implications of the arrangement -- if you're doing a one-off and don't have any particularly sensitive equipment nearby, you probably won't mind the emissions, but the susceptibility may still be an issue.  At the very least, assign every other wire to ground, and connect them to ground at both ends.  This gives a typical about 100 ohm impedance for the each signal line inbetween, so yeah a little source termination resistance is desirable (most logic devices have a 50-100 ohm source resistance, so a little extra to make up the difference is about right, or more can be used to slow the risetime further).

50cm is short.  Short enough that inter-integrated circuit (I2C, I2S, SPI, etc.) signals are very possible, even in a full commercial setting, without a shielded enclosure, testing to EMI standards.  In addition to interleaved grounds, it might also need some ferrite beads, and cautious filtering at the transmitter and/or receiver.  RS-422 signalling would be wise, but may not be required.  (I would strongly recommend RS-422 and/or shielding for anything much longer.)

Tim
« Last Edit: July 23, 2020, 01:02:27 am by T3sl4co1l »
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Online jbb

  • Super Contributor
  • ***
  • Posts: 1265
  • Country: nz
Re: About ribbon cables at fairly high frequencies.
« Reply #3 on: July 23, 2020, 01:11:20 am »
You can probably do it with LVDS. It’s differential which will improve radiation and noise rejection.
 

Offline langwadt

  • Super Contributor
  • ***
  • Posts: 4857
  • Country: dk
Re: About ribbon cables at fairly high frequencies.
« Reply #4 on: July 23, 2020, 02:10:29 am »
12MHz, that's comparable to the bandwidth of old school hard drives --

afair parallel-ata went up to 16MHz DDR before it went to the 80 conductor cable with the extra grounds
 

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: About ribbon cables at fairly high frequencies.
« Reply #5 on: July 23, 2020, 08:51:31 am »
I made some calculations on my SPICE software on transmission lines, but I'm not sure whether a ribbon cable can be regarded as a transmission line.
It could probably be, but then you treat it as a transmission line - 100Ω source termination, 100Ω sink termination, amplitude at the sink is 50% of the amplitude applied to the source resistor. Not sure if you want to bother.

Also, I intend to insert resistors of some 22 ohms at the TX point.  And the receiver is a MCU that has some 5 or 10 pF on the inputs.

For me, 50cm ( 20 inch ) will be perfectly alright. Actually, it's a I2S interface, but it's insensitive to jitter. The only importance is that there are no bit errors. Possible ringings must have subsided after a maximum of 30ns.
That's a short cable and a fair bit of time for things to settle (in terms of signal propagation velocity). You might get away with simply treating it as an RC network. You will get 100% of source amplitude at the sink.
Ribbon cables may have some 100pF of capacitance per meter. I don't remember if it referred to capacitance between two neighbor wires or a wire and both of its neighbors, but that's still ≤100pF for half-meter in the worse case. Even with 100Ω source resistance (and no sink termination), you are looking at 10ns time constant and <30ns rise time to 90%. Possibly no overshoot and ringing at all because it's close to the characteristic impedance.
Inserting grounds between signals significantly reduces crosstalk so that shouldn't be a problem either.

I'm not sure how to simulate that kind of thing in SPICE properly. If you have a scope, simply build it and test it. If something looks bad, post pics. If you have a capacitance meter, test your cable.
:popcorn:

edit
Another trick is AC-coupled sink termination, which reduces overshoots when the source is under-terminated but doesn't cause the 50% reduction of amplitude at "sufficiently low" signal bitrates.
« Last Edit: July 23, 2020, 09:01:37 am by magic »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: About ribbon cables at fairly high frequencies.
« Reply #6 on: July 23, 2020, 04:46:39 pm »
Ribbon cables may have some 100pF of capacitance per meter. I don't remember if it referred to capacitance between two neighbor wires or a wire and both of its neighbors, but that's still ≤100pF for half-meter in the worse case.

Easy: take the impedance parameters of free space.
Zo = 377R
Lo = 1.26 uH/m
Co = 8.84 pF/m

Take the impedance Z and velocity factor c/c_0 of the transmission line.  In this case, we know the velocity shift is entirely due to dielectric constant (e_r > 1, mu_r = 1).

Divide Zo by c/c_0, then divide again by Z.  This ratio is how much extra capacitance the transmission line has due to its geometry and dielectric.  Multiply by Co to get the capacitance per length.

For inductance, do the same thing, but upside down: take Z / (Zo * c/c_0), then divide Lo by this ratio to get the inductance per length.

That is, when Z < Zo, it's because C > Co and L < Lo.  The ratio these differ by, is simply the impedance ratio.

And the ratio of L to C is the impedance squared, Z = sqrt(L/C).

And the velocity factor is related to the product of them, sqrt(L C).  Namely, c/c_0 = 1 / (2 pi c_0 sqrt(L C)).

More generally, the index of refraction is c_0/c, and goes as sqrt(e_r mu_r).  Works just the same for one-dimensional transmission line modes, as it does for general propagating waves. :D

(Or, I think I did that right, reasoning it out just in my head.)

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline schratterulrich

  • Regular Contributor
  • *
  • Posts: 50
  • Country: at
    • Elektronik & Layout
Re: About ribbon cables at fairly high frequencies.
« Reply #7 on: July 23, 2020, 05:19:21 pm »
You can get the book "High-Speed Digital Design - A Handbook of Black Magic" from Howard Johnson.
There is a whole chapter on Ribbon Cables - Rise time, frequency response, crosstalk,...

Maybe you can find something on his website, too:
http://www.sigcon.com/Pubs/news/3_10.htm

I have got good simulation to measurement correlation by simulating flat flexible cables with transmission lines in IBIS simulations. But it should work in SPICE, too.

If you know the impedance of the cable the easiest way should be to insert a series termination after the IC pin to match the output impedance to the PCB trace and a second series termination in front of the pin header to match the PCB trace impedance to the cable impedance. Simulations will show, that this works. ( assuming Zout(IC) < Z(pcb) < Z(ribbon) )
 

Offline jmelson

  • Super Contributor
  • ***
  • Posts: 2852
  • Country: us
Re: About ribbon cables at fairly high frequencies.
« Reply #8 on: July 23, 2020, 06:15:30 pm »
In my particular case, I need to transfer bitrates at a maximum of 12Mhz. Can someone perhaps estimate the maximum length of the cable?  I'm using it with a ground wire inserted between the signal wires.

Also, I intend to insert resistors of some 22 ohms at the TX point.  And the receiver is a MCU that has some 5 or 10 pF on the inputs.

For me, 50cm ( 20 inch ) will be perfectly alright. Actually, it's a I2S interface, but it's insensitive to jitter. The only importance is that there are no bit errors. Possible ringings must have subsided after a maximum of 30ns.
You need to either series terminate at the input end with the characteristic impedance of the cable minus the output resistance of the driver,
or terminate the destination end with the characteristic impedance.

Series termination may cause slow edge transition time due to the considerable cable capacitance.

The best way to do this is with differential drive (26C31 and 26C32 type drivers and receivers) and twisted-pair ribbon cable.  Or, for LVDS, use
one of the many LVDS driver/receiver sets.  The 26C31 can do 12 MHz easily, the LVDS drivers can do 100 MHz pretty well.  50 cm is a piece of cake for differential twisted pairs, we routinely do 30 m with minimal degradation.  The characteristic impedance will be around 110 Ohms for these cables.

Now, for single ended, things get messier, but 12 MHz and 50 cm will still be quite easy.  Return currents on the ground wires may cause a little crosstalk.

Jon
 

Offline dietert1

  • Super Contributor
  • ***
  • Posts: 2473
  • Country: br
    • CADT Homepage
Re: About ribbon cables at fairly high frequencies.
« Reply #9 on: July 23, 2020, 06:18:26 pm »
12 MHz is nothing for such a regular and high quality transmission structure. If you add some termination on the receiver side (e,g, 100 Ohm with 100 pF in series to ground), that cable can be more than 10 meters. With all those grounds, radiation and losses will be low. Sorry i don't have a formula, but it depends on many factors like drive impedance and receiver input capacitance. If the drive impedance is less than 50 Ohms, you may want to insert 33 Ohm resistors into the driver side, like they used to do for memory drivers. Sometimes it is enough to only "treat" the clock signal and all other signals can be left as they are.
Make a test connection and a scope measurement and everybody will be happy to comment your results.

Regards, Dieter
 

Offline Berni

  • Super Contributor
  • ***
  • Posts: 5050
  • Country: si
Re: About ribbon cables at fairly high frequencies.
« Reply #10 on: July 23, 2020, 08:11:47 pm »
Ribbon cables can go really fast with the right pinout and plenty of grounds.

For single ended lines make the wires next to it ground. For differential put the pairs together and surround those with ground.

I ran plenty fast signals trough bog standard ribbon cable. Used it for MIPI display data at about 1Gbit over about 15cm. Had PCIe 2.0 (5 Gbit) running over about 10cm of it just fine. SATA III will also work just fine over it. Its all about having the right pinout. If you have a SPI bus with all the signals next to each other in the pinout then you can have problems even at 10 MHz, put the grounds in and its >100MHz no problem.

But as others have said once the cable gets long enough it should be treated like a transmission line. It needs to be driven by the correct impedance using series resistors and it needs something on the other end to terminate it. Do that and im pretty certain your I2S bus can run over 50m of ribbon cable and still work.
 

Offline madires

  • Super Contributor
  • ***
  • Posts: 8276
  • Country: de
  • A qualified hobbyist ;)
Re: About ribbon cables at fairly high frequencies.
« Reply #11 on: July 23, 2020, 08:22:14 pm »
Old parallel SCSI supports up to 320MHz over ribbon cable.

Edit: 80 MHz DDR
« Last Edit: July 23, 2020, 09:04:21 pm by madires »
 

Offline helius

  • Super Contributor
  • ***
  • Posts: 3688
  • Country: us
Re: About ribbon cables at fairly high frequencies.
« Reply #12 on: July 23, 2020, 08:32:12 pm »
Old parallel SCSI supports up to 320MHz over ribbon cable.
No, SCSI-320 is 320 MB/s, for a 16-bit bus.
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17429
  • Country: us
  • DavidH
Re: About ribbon cables at fairly high frequencies.
« Reply #13 on: July 23, 2020, 11:22:39 pm »
Tektronix had no problems using ribbon cables to transfer 100 MHz bandwidth signals inside oscilloscopes and under similar conditions, 100 MHz digital signals would work just as well.  An example on the receiver side is shown below.  The poor soldering is a quick repair job and unrelated to what needed to be diagnosed and repaired; thick gold plating makes for difficult soldering.

Update: When I say 100 MHz bandwidth, I should have included that that is with a minimum of aberrations in the 3.5 nanosecond step response.
« Last Edit: July 26, 2020, 02:29:44 am by David Hess »
 
The following users thanked this post: KE5FX

Offline schratterulrich

  • Regular Contributor
  • *
  • Posts: 50
  • Country: at
    • Elektronik & Layout
Re: About ribbon cables at fairly high frequencies.
« Reply #14 on: July 24, 2020, 06:10:22 pm »
a picture is worth a thousand words so I have made a quick and dirty measurement of a pcb with an attached ribbon cable.

1031884-0

TDR without cable:



TDR with cable:



I can see
1) launch from semirigid to pcb
2) section of pcb trace (coplanar waveguide on bottom) 74 Ohm
3) section of pcb trace (microstrip on top 0,15mm width FR-4 1,55mm)
4) thru hole with pin header, stub of pin, stub of ribbon cable,...
5) ribbon cable with quiet perfect 100 Ohm (G-S-G)
6) some capacitance at the end of ribbon cable
 
The following users thanked this post: Berni, thm_w, helius

Offline magic

  • Super Contributor
  • ***
  • Posts: 7454
  • Country: pl
Re: About ribbon cables at fairly high frequencies.
« Reply #15 on: July 25, 2020, 07:11:18 am »
This doesn't look bad. I remember testing other non-RF cables like audio and they weren't flat but showed a rising slope, not sure what that would be.

It occurred to me that with perfect source termination the capacitance of the cable doesn't even come into equation because it all gets charged in one round trip time. With 10% overtermination it's still looking very well and easily meets the 30ns settling time requirement.
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 22436
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: About ribbon cables at fairly high frequencies.
« Reply #16 on: July 25, 2020, 08:18:34 am »
I've seen audio cables where the ground is just strands orbiting (helically) around the core, not braided at all.  So, essentially zero (magnetic) shielding value... and probably a helical waveguide sort of behavior as far as impedance, velocity and so on.  I'm not sure whether that would go up or down with frequency, but it sounds like it will be dispersive.

Also PVC insulation is shite, so it's lossy dispersion at that. :P

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 

Offline David Hess

  • Super Contributor
  • ***
  • Posts: 17429
  • Country: us
  • DavidH
Re: About ribbon cables at fairly high frequencies.
« Reply #17 on: July 26, 2020, 02:39:43 am »
It occurred to me that with perfect source termination the capacitance of the cable doesn't even come into equation because it all gets charged in one round trip time. With 10% overtermination it's still looking very well and easily meets the 30ns settling time requirement.

Exactly, it is the same as any other constant impedance transmission line.

In the example I showed from a Tektronix 2230 oscilloscope, the outer pins are ground and the inner pins are a differential transmission line.  The receiver end is single ended parallel terminated into 50 ohms (two terminations) and *no* source termination was used which was pretty common in Tektronix oscilloscopes.  At much higher frequencies they used double termination.
« Last Edit: July 26, 2020, 02:42:45 am by David Hess »
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf