|
Quote: Originally Posted by XJet |
|
|
|
|
|
While it's true that the use of a reliable-transport layer has the potential to improve performance in such situations, it's not necessarily going to work that way in practice.
The XBeePro modules use a CSMA approach to transmitting data-packets so if the interference is strong enough, it will cause the transmitter *not* to transmit, even if it hasn't received an ACK packet before the timeout occurs. With a high (enough) general noise floor the XBee modules will just sit and wait for the noise to drop before they even try to transmit. This means that if there's a strong source of wideband noise visible to the transmitter but not the receiver, the XPS is at a decided disadvantage over something like FASST which will keep transmitting regardless. |
|
|
|
|
That does not match the results of my testing. I have two pretty good indicators
that tell me when XPS Tx is transmitting, and when it's retransmitting.
The first is the voltage display on my Tx. It drops a tenth or two when it
starts retransmitting packets.
The second is a side effect of what I'm using to generate noise, which is my
500mW 2.4Ghz video system. With the video receiver, I can actually *see* 2.4Ghz
RC systems transmitting on overlapping frequencies (Spektrum is the worst).
Watching it, while XPS is on, I see a little bit of visible noise on the video image
from XPS Tx/Rx when they're transmitting normally, and *a lot* of noise when
the XPS Tx starts retransmitting. A spectrum analyzer may not be able to see this,
because the noise generators most people are using often just swamp the XPS
signal on the spectral display. My video signal doesn't have that problem, because
it is both the noise for XPS, and shows noise *from* XPS.
Anyway, what I've found is that at no time, does the XPS Tx stop transmitting,
and re-transmitting when the link goes bad, and is ultimately totally swamped
by my video system. We're talking putting the half watt transmitter inches away
from either XPS Tx or Rx, sometimes both close together, and sometimes
separated by a great distance (so it only swamps the Rx, or only swamps the
Tx) and I can see it re-transmitting furiously any time the link is broken for any
reason. The Tx never gives up. Unfortunately if you mix failsafe and a power supply
problem together, the Rx does sometimes give up, but that's a different problem.
|
Quote: |
|
|
|
|
The other issue is that an XPS system can use a lot more than the usual 10% of its timeline whenever it fails to get an ACK from the receiver. XPS claim that up to 10 users can share the same channel (a theoretical maximum, as anyone working with CSMA networks will tell you, things can start to degrade well before the "theoretical maximum" is reached) which means that if you have (say) four XPS systems working on the same channel and the noise floor rises, those four transmitters alone can completely monopolize the time-line because of all the retries that are forced. In fact, with four transmitters each issuing three retries, the timeline would be well and truly saturated -- meaning that even if packets did get through, the latency would rise significantly. The apparent inability to hop means that none of those four transmitters would be able to switch to a cleaner channel either. |
|
|
|
|
That's true although it'd be rare to find 4 XPS on the same channel, as the Rx
does try to find an open channel when it boots. You'd have to have pretty much
saturated the rest of the 2.4Ghz band to force that many XPS systems together
onto the same frequency, in which case there may not be anywhere safe for it
to hop to, even if it tried.
|
Quote: |
|
|
|
|
This might well be true -- although it must be remembered that with a 10% duty-cycle, the XPS data packets (in fact the data packets of all 2.4Ghz systems) are a pretty small target so the effect of short-burst impulse noise tends to be minimal (or at least significantly less than on 100%-duty-cycle systems like PPM/PCM) anyway. |
|
|
|
|
Again, I was thinking something along the lines of fully swamped 50% of
the time. 10ms worth of noise on and 10ms off would cut the other
systems down to 50% of their packets getting through, while XPS would
retransmit through the holes.
Ultimately we don't know what the noise looks like though. I think XPS will
work better in some environments and worse in others.
ian