I suspect singlemode would be better than multimode for this part (no/less mode dispersion); and I'm able to say at this point that singlemode is not really all that difficult, having just bootstrapped a singlemode network in under 6 months from no fiber experience and with very little budget. While the conventional wisdom is "short runs == multimode", it's quite possible for the conventional wisdom to be outdated.
-- Cats, coffee, chocolate...vices to live by
It's more about matching the emitter to the fibre. If you want to use VCSELs to get low cost and high speed, you're pretty much stuck with MMF because the VCSELs are multimode. There exist single mode ones now, but they aren't available in packaged arrays with connectors, I don't think. I also don't know how fast they can be modulated.
The nasty mode jumps during VCSEL turn-on are a significant source of jitter in some applications.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
Multimode dispersion is in the range of 1-2 ns per kilometer, not a problem in my application. A cheap VCSEL will put a lot of light into a fat multimode fiber.
John
We're getting RMS jitter around 10 ps, using a cheapish 850 nm VCSEL, a silicon PIN diode, and multimode fiber.
The old cleaved CD lasers were wonderful. Some of the VCSELS are good, some are terrible, for pulsed work. We make our suppliers promise not to change their parts without telling us first.
John
Twinax should be better (the shield currents are much less); I'm thinking that repeaters are cheaper than fiber transmitters, and three repeaters would make your longest run (150 feet) pretty manageable. I think. It's not clear from Belden's data for 9463 cable what happens at this frequency range, but a test might be worthwhile.
[snip] [snip]
I'm adverse to clipping since it'll generate reflections.
There are ways to "pre-emphasize" that are not intuitive (not the matched filter kind of thinking... rather Thompson-style off-the-wall thinking :-). Unfortunately I can't divulge (it's proprietary), but I've done over a 1000' of twisted pair, 10Mb/s Manchester, with very good shape restoration. ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
Remember: Once you go over the hill, you pick up speed
Ever used the mil STP media?
It is "Shielded Twisted Pair". That is a single twisted pair inside of a shield, inside a teflon jacket. All at less than an 8th inch diameter for signal sizes. They also make shielded power type cables, which are three spc runs inside the shield. The single pair signal wire is what you want.
Military cables use it for every differential pair on a cable run and IIRC, it performs better than 5e or 6 Ethernet wiring.
Then again, General Cable has some offerings that look better than many I have seen. Their data sheets give you an idea of what parameters to look for though. Some nice, educational info there.
You can get 50 micron fiber that can do 10Gb now too, so you should be able to use xcvr modules on each end that will allow you to resolve/bypass any of the issues that have been thus far discussed.
Just take a look at a high end network switch and see how they do it. You can bet that their fiber interconnects are low latency, zero error links.
Oh... that's right... you seem to think that you are always inventing a new mouse trap and the final solution couldn't possibly already be in the channel.
Do you really think that the path lengths (considered to be equal to start with) would change that much one from another?!
That's what I suggested yesterday, more or less. As you say, the problem has already been solved elsewhere...
Regards,
Chris
Put a "carrier" on the line and see how fast a spike carries down it a given length.
What is that called? Delay Skew? Measured in nanoseconds per given length. Commonly ns/100 meters. Velocity of Propagation?
Their Genspeed 10,000 cable looks pretty good.
I have to send dc-coupled, baseband pulses of arbitrary width and duty cycle, with
Given the cable attenuation numbers we're seeing, the 2 GHz would be way down in the noise. And quantizing the pulse edges to a 2 GHz carrier would grossly blow the jitter budget.
John
Interesting. All of those cables have attenuation in the range of
34-44 dB per 100m at 350 MHz, and the premium cables aren't necessarily the best. Our runs will be more like 50m, so our attenuation will be 17-22 dB 350 MHz is roughly the bandwidth we'd need, and 17 dB loss will wreck a pulse big-time. I guess typical network hardware has to do big-time adaptive equalization to get gbps data through cables like this.John
You can simulate the DC coupling.
I was probably doing pipes before you even owned your first computer.
Shannon is your friend.
It is mass produced, and the numbers are cat 6A compliance level.
Did you look at their RGB TP cable? IIRC it was faster.
Reflections aren't much of a worry with a series-terminated driver.
I know several ways of doing preemphasis. All of them take significant amounts of tweaking (generally automatic) when the cable length is both long and widely varying. My point was that if the duty cycles are low, you can ignore the ISI problem and use the early part of the curve, which is much much steeper.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs (atsign) electrooptical (period) net http://electrooptical.net
It would have been quicker to just say "no."
John
If you only need 50 m runs, for instance LMR-400 coaxial cable would have only about 10 dB attenuation at 2.4 GHz. Thicker, much lower loss variants exists, but of course the bending radius become a limiting factor and at some point a thick cable can fall into the undesirable waveguide mode, when the wavelength is small compared to cable diameter.
When using a narrow band PLL (say 0.1 % bandwidth) the PLL cycle to cycle jitter is small. The question is now how to detect the change in the data pattern and at the _next_ PLL cycle transition generate the output.
Assuming simple amplitude modulation of the carrier low power level in the idle state to maintain PLL lock and a higher level for your variable length pulse. The question is how fast can you ramp up the amplitude and how fast you can detect this step. Assuming this can be done in 1 cycle at 2 GHz, this would give 0.5 ns timing granularity with low (PLL) jitter.
Amplitude modulating a carrier with a square wave will produce side bands which also must pass the medium.
As a rule of thumb, the coaxial cable losses (expressed in decibels) is proportional to the square root of relative frequency. Assuming 6 dB total loss at 2 GHz, the loss at 6 GHz would be just above 10 dB, thus the higher sidebands would only be slightly more attenuated than the square wave. However, I am not sure if this rule of thumb can be applied to the LMR cables.
Perhaps with a slightly thicker cable, you could simply use an envelope (diode) detector at the receiving end and still get decent jitter figures.
Using amplitude modulation eliminates the DC component from the cable, thus avoiding any ground loop problems etc.
You could always send a positive pulse down the fibre for a one and a neg pulse for zero, or even a few cycles of carrier, then use a bit of gating and a latch at the other end to save the state. The key thing is low cost network parts of proven reliability to do the job and you won't get the propagation delay variability with temperature that you would get with wire, as the dielectric constant changes...
Regards,
Chris
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.