Prism vs Hermes vs Ruby vs..

[Anymouse]

Up until now, I thought that the Hermes and Ruby chipsets were terrible cards compared to the 200mw Senao. Looking at the numbers, doesn't the Senao outperform them on both receive sensitivity and output power?

Although I think it is good that I found this forum, because we are about to launch a rural WISP (after doing consulting work for several other WISPs), relying on long-distance links, which apparantly the Prism chipsets do not do.

Specifically why is that? I've done work with other non-802 technologies that have issues due to their own protocol implementations, but isn't the 802.11b protocol universally the same?

I hadn't even heard about StarOS until today - I think we will look into it for all our future base stations, but is the long-distance capabilities limited to the Hermes cards, and do they only work with the special StarOS firmware updates?

Or is there off-the-shelf cards that can handle high distances?
Maybe ones that have nice specs like the Prism 2.5's do?

[Anymouse]

Oh, and another thing I am curious about - is the "long distance" issue only when the card is in AP mode, or if it is a client as well?

If so, then could one use a Ruby/Hermes card for the AP, and a Senao for the client? Link budgeting with -82dbm sensitivity and 15db output power at both ends is rather frightening for our area (very little interference, and very flat terrian, we aren't planning on having picocells) - although a senao at the client end sounds like it would eliminate many hidden node issues..

[tony]

Welcome to the forums!

Of all the cards we've developed drivers for, the Hermes / Ruby cards are amongst the most 'mature' (firmware-wise) when it comes to WISP use.

The card's short range limitation is a problem with most 802.11 cards, and is due to a timing issue. (ie. the card only waits x number of microseconds for an ACK from the remote end, and then retransmits if needed). This 'x' value unfortunately cannot be changed on most cards, including the Intersil. The Hermes / Ruby cards have the support to change this value, as do some of the newer 802.11a / g cards. (to what extent is to be seen). This limitation exists in both AP and Client devices.

Of all the cards, we recommend a Hermes / ruby card, with possibly an RF Linx PCB AMP if needed.

To make use of the long-distance feature of the Hermes / Ruby cards, you must use one of the following products:
@) StarOS (Any Edition) - remains 802.11[b] protocol compatible
@) Proxim COR / ROR outdoor routers - proprietary protocol
@) Karlnet (possibly) - proprietary protocol

Thanks!

[usairwave]

Other venders can make long distance shots as well on a PtP basis. Cisco bridges for one and go up to 200km (of course you will not get that far on flat ground and I doubt they can go that far anyhow with any real throughput....) You can multipoint the cisco bridges but if you are going that far, keep it for backbone, no clients.

Rule of thumb in my book:

- flat ground
- 100'-150' towers
- 15-20 miles is the max with grids and amps.

Just my 2 cents.

[Anymouse]

Hmm, although the output power and sensitivity of all the Hermes/Ruby cards aren't that great. Plus the customer premise equipment for StarOS is a tad pricier, than say - our current solution of a Linksys WET11 over PoE in an enclosure next to the antenna (pigtail is the only antenna cable).

The StarOS implementation of 802.11b is completely standard? So, would it be possible to use a StarOS base station with a Hermes/Ruby card - and then two packages for our customers, based on distance - either the WET11/Prism (nice SNR for closer range, and much lower price) - or a Net4511-20 and Hermes card with a grid for the longer ranges?

I guess my question is whether the StarOS modified distance protocol will work for multipoint - will clients closer to the base station (on PRISM cards) have a signifantly worse quality of service due to the higher "x" value?

Thanks for answering all these questions, it might be saving us a lot of time and money working with the Prism cards..

[tony]

Yes, we remain fully 802.11 compatible, so you can tune your AP for the furthest client you will serve, even if they are not StarOS-based. For optimum performance, your distance users should use a product that allows for distance adjusting such as StarOS or Cisco. (Cisco also remains 802.11b compliant from what I understand).

Thanks!

[Skaught]

What is an RF Lynx PCB AMP?

[tony]

What is an RF Lynx PCB AMP?

Here it is:
http://www.rflinx.com/2.4GHz%20Bi-Directional%20PCB.htm

[usairwave]

Another rule of thumb for cisco:

Only use cisco with cisco.

We have seen several issues (mainly with scripts of websites not loading) if you have cisco aps with non-cisco clients. I know other people have had success, we haven't.

For us, clients past 8 miles are really tough to get on without a decent tower. The standard limit on 802.11 from what I understand is 12 miles without a timing solution.

On flat ground, I don't think you will get that many clients past 12 miles.

[go.fast]

What is an RF Lynx PCB AMP?

Here it is:
http://www.rflinx.com/2.4GHz%20Bi-Directional%20PCB.htm

Do you recomend the RF PCB Lynx amp over the YDI amp or do you think they are both just as good?
I like the price....
George

[Anymouse]

Just an update - we have been running for several months using Prism 2.5-based base stations and customer premise equipment, both entirely PoE-based and off-the-shelf (Tranzeo TR-3300 / TR-CPE). I highly recommend both, especially the TR-CPE - now I gag whenever I see a PC Card or a USB-based solution : P

Anyways, we have a very well-designed infrastructure, and our coverage is as we expected for the nearby clients - but we ran into the problem we expected. 62km away, we can get a reasonable signal (-80), however the timing issues cause about 75% of the packets to be dropped.

Just a clarification, that StarOS does not, in fact, support tweaking the distance timing on cards with any reasonable output power (ie. 200mw?). We have several clients who would like service, but are very, very, far away from our base station - and yet receive signal.

If we implemented a StarOS base station with the highest hermes/ruby based card we can find - as well as an RFLynx bidi amp, would our clients still be able to use the prism2.5 based solution? Or would they also require StarOS equipment?

[sure-data]

You can just set the AP, but then only one half of the connection is working properly. Both the Client and AP need to be "set" for the distance.

Closer-in clients without distance settings don't suffer enough that you can really see it.

If you want to stick with prisim 2.5 clients, they need to be within (about) 12 miles of the StarOs AP.

We "repeat" with 2 cards and StarOS, one as client, other as AP, and "cluster" our users around the nearest AP.

If you have customers 60 miles out and they want service, the cost to install them a StarOS client is the least of their worries.

[Anymouse]

60 kilometres, not miles. Huge difference.

And no, repeaters are not what I am looking for - obviously I can build repeaters between us and the client, but the cost to build and maintain versus the one or two people in the area it would service - is not worth it. I am looking for a true long-range point-to-multipoint solution.

If that means running two base stations colocated (and therefore tying up two of the three precious frequencies) - one that is Prism 2.5-based, for short-range clients to combat partial LOS issues - and one that is StarOS-based on an amplifier for long range clients, that should work.

I will look into the costs of using StarOS (including assembling hardware components into a CPE) as a long-range solution, however.

[Anymouse]

I just want to be sure about this - if we replace one of our Prism 2.5 base stations (Tranzeo TR-3300) with a StarOS-based system that uses a hermes-based card with a bidirectional amplifier (to match the output power and receieve sensitivity of our previous solution) - then set the StarOS to, say, 65 km (40 miles).. clients that are only 8 km (5 miles) away will not take a significant performance hit?

Our biggest problem with that base station at the moment is not due to signal strength - it is due to timing issues with the Prism 2.5 cards. StarOS looks as if it can fix this.

[tony]

Your clients at 8km should not have any performance issues relating to setting the AP to 40 miles.

Thanks!

[dkii]

I dunno about 40 miles, but I have a 25 mile link that is getting > 450KB/s, with no noticable impact on closer units, although most of the associations on it are between 12 and 16 miles.

[Anymouse]

So, only hermes-based cards can be tweaked for distance, right? What is the highest-power card that can be tweaked?

Overall, what we are really looking for isn't a new type of AP - but rather a simple value for timing somewhere in the firmware of the Prism cards, if we had a "modified Prism firmware", that would solve all our problems.

Is the reason that you cannot do a similar tweak on Prism-based cards due to the fact that Intersil wants to charge an arm and a leg for developer info?

It really just boils down to changing a timeout value somewhere deep in the firmware, right?

[lonnie]

Intersil does not have the same MAC. The feature set is similar but different. We tried real hard but no luck.

The Orinoco are rated at 35 mW and that is the only cards that we know of that can be tweaked. That means that if you want to go the distance you require an amplifier. Just consider it a cost of doing business.

RFLinx has OEM amplifiers for around $150 and they'll put out 500 mW.

[tony]

We worked with the Intersil firmware developers to try and find a way to expose a method for tweaking ACK timing on the Prism 2.x cards, however we have found that the firmware in general is designed in such a way that it was not possible to do in it's current state.

The ACK timeout value is not something that resides as a simple number, but rather as an algorithm that has quite a few variables that change depending on the mode of operations, link speed, etc.

[Anymouse]

Ahh, okay - thanks for the info, I was really curious why nobody had a Prism 2.x-based long-range solution out there.

We will likely end up using an amp, and although there are several drawbacks (amplification of noise, and inherant drop in bandwidth due to switching) - it seems to be the only option out there for long range (20-60km) multipoint within the 802.11b standard.

[tony]

With our current shot with two standard Avaya silver cards, we are still able to achieve over 400KBytes/sec at 52 Miles.

We are always interested in hearing other people's results.

Thanks!

[Anymouse]

With an amplifier? What type of amplifier?

I thought that amplifiers would dramatically reduce throughput, due to the fact that they have a rise and fall time to the switching between TX and RX.

[lonnie]

We are using YDI and RFLinx amplifiers. Amplifiers do reduce throughput due to their switching times, but it is not all that dramatic and certainly makes the shot possible and very usable.

[dkii]

That 25 mile shot is with YDI indoor amps at both ends. I haven't seen a noticable difference in throughput with/without amps (signal allowing)