3G signal boosters - how good?

[dennis]

[John Rumm]

Which is why I suggest keeping the RF short, and running ethernet most of the way.

(forget titting about with USB modems, because you then have to rely on a software firewall on the host computer as well, and with no NAT to insulate you from the outside world).

Based on experience, even in the same location they are usually noticeable better just as a result of the better aerial. However for very poor signal areas I would go with an external directional aerial, place the router close to it (but inside) and run POE to it if required.

Yup, outside will certainly help. Much depends on circumstances as to what lengths you need to go to.

[John Rumm]

30' is no trouble for ethernet, a single ethernet segment can be 1000 feet long if required.

[John Rumm]

It works well for me, using a blackberry.

Most of the others also have pico cell type offerings. However these solve a different problem of poor mobile voice comms in remote areas, rather than low broadband throughput where the 3G is the only available broadband connection.

[PeterC]

noticed the coincidence of subjects.

[Steve Walker]

> It works well for me, using a blackberry.

[Dave Osborne]

USB is a finely balanced design where the design goals are to propagate a high-speed signal over an inexpensive cable with inexpensive silicon at each end and without impedance control of cable or connectors. This is a tough ask of the engineering design. The limitation of the cable length in USB is derived from a number of factors, including practical maxima of propagation delay, cable skew and attenuation. It turns out that the maximum practical cable length to meet all the design criteria is 5m or so. This is transmitting data at 480Mb/s. Given that USB was designed to connect local peripherals to a PC, a 5m maximum cable length is entirely adequate.

The speed of light is approx. 3.3ns per meter. Electrical signals propagate down a cable at between about 30% and 80% of the speed of light. If we split the difference and say that the signal is travelling at 55% the speed of light, then the propagation delay is approx 6ns per m. The timing budget for USB2@480Mb/s allows for a 30ns cable propagation delay. Now 30ns divided by 6ns/m = 5m. Hence the maximum cable length is 5m.

At a mere 100Mb/s, CAT5 twisted pair has a maximum length of 100m again due to absolute limits on signal timing (i.e. the amount of time it takes a signal to propagate down a cable).

These are not artificial short-sighted limits, they are engineering compromises which had to be made to deliver a robust design at an affordable price with the technology (and in some cases the mathematical techniques for encoding and error correction) available at the time of design.

I quite agree that historically there have been many short sighted decisions made (e.g. with the maximum memory a system can support) but be assured that you could not tweak USB2@480Mb/sec to operate "over far greater distances with just slight changes to specs and little change to cost or performance".

[Theo Markettos]

FWIW I think there's also a power issue. I tried a ZTE MF627 dongle on a cheap 5m passive extension cable, and it wouldn't power up. It worked fine on a 50cm cable. My hypothesis is that the dongle is quite power hungry, and the resistance losses along the cable cause enough voltage drop to stop it working.

So a powered hub at the end might help.

Theo

[fred]

Far too much emphasis on cheap & cheerless IMO, if the bits that are originally thought to be expensive are made in sufficient volume then, as I'm sure you are aware, the cost tumbles. I'd rather pay more for a better spec/interface/product and with economies of scale I really don't think it would be that much more.

[John Rumm]

History however has demonstrated that you were in the minority. Remember that firewire (IEEE 1394) with a 4.5m limit and 400Mb/sec was available at the same time as original USB, which at best could muster 11Mb/sec over similar distances.

That is why we are now all using firewire in preference to USB... oh hang on a mo?

The problem was first firewire used more expensive chipsets (would have added a couple of $ of cost to each product), and initially there was a small license cost to Apple to use it (only a few cents IIRC). Neither of these problems still exist (the license fee has been dropped, and the chips are as cheap as... well chips). However USB rules the roost due to early mass market adoption.

As to meeting its design goals, USB does so remarkably well. Providing long distance high speed comms was never one of them, so it seems pointless to complain its no good at it. One could have included that in spec, but only at the expense of making other trade offs that would have made it less suitable for its intended purpose.

[tony sayer]

As may be, but ISTR that the original poster had a USB dongle and wanted to make that perform better, not go any buy another one..

[Peter Scott]

Yes that sounds promising. But.... I went to my friend to suggest these ideas and she said.... But how do I get it out when I need to use when I am travelling? No I didn't think of it either. Doh! So elevation is a good solution when you want a fixed 3G installation.

The intelligent aerial (sorry antenna) on BBC looks intersting.

Peter

[Andy Burns]

Pulley at the top to hoist the dongle with?

[Frank Erskine]

"Halfway up I met the dongle coming down..."

[Peter Scott]

Are we not suggesting angle grinders any more?