Newer Model Instrumentation Amp

[Paul Keinanen]

The time difference off the ceiling and the direct route is almost the same as
our symbol rate. Things get ugly fast. Without the ceiling things are great.
;-)

So we are talking about 1-3 MHz, is this the actual data rate or the
spread spectrum chip rate ? At 2.45 GHz ?

In a SS system, the spreading chip rate should be several percent of
the absolute frequency to be effective against multipath.

Regardless modulation method, some general precautions against
multipath:

Is it possible to install the base station antenna on the inside of
the dome itself ? This would generate a strong direct (dome, field)
signal compared to reflected (dome, field, dome, field).

If the base station antenna can not be installed in the dome itself
but it has to be installed at ground level or somewhere in the
audience, at least make sure that the radiation pattern above the
horizontal plane is heavily attenuated, reducing power radiated into
the dome (or received from the dome). A metallic shield (or mesh)
above the antenna might be usable, but of course, it creates early
reflections from the shield, which may be (or not) harmful depending
on the modulation method.

It's not a matter of nulls, rather the receivers can't lock onto the signal.

Don't lock at all or sometimes drops out of lock ?

How long does it take to relock ? Is this time longer than an other
antenna with separate receiver typically maintain lock with a moving
receiver ?

There are several ways to implement spatial diversity, e.g. on the
base band level switching between full receivers or simply RF antenna
switching when the received power drops on current antenna.

To speed up relocking one might also notice that since at least one
receiver usually remains locked, this information can be used to limit
search options on the unlocked receiver.

 

[[email protected]]

So we are talking about 1-3 MHz, is this the actual data rate or the
spread spectrum chip rate ? At 2.45 GHz ?

The data rate is just under 1Mbps. We're using ~480Kbps of it, TDMA, IIRC.

In a SS system, the spreading chip rate should be several percent of
the absolute frequency to be effective against multipath.

Regardless modulation method, some general precautions against
multipath:

Is it possible to install the base station antenna on the inside of
the dome itself ? This would generate a strong direct (dome, field)
signal compared to reflected (dome, field, dome, field).

It is (it's certainly not outside the dome, if that's what you mean). It's
been tried in the ceiling, on the field, in the stands... We've tried
directionals, omnis, amplified, padded, just about everything. One thing we
haven't done is separating the transmit and receive antennas (so they can be
padded differently). The solution, so far, is to abandon 2.4G, in favor or
900M, which works reasonably well (except for other system limitations - e.g.
bandwidth).

If the base station antenna can not be installed in the dome itself
but it has to be installed at ground level or somewhere in the
audience, at least make sure that the radiation pattern above the
horizontal plane is heavily attenuated, reducing power radiated into
the dome (or received from the dome). A metallic shield (or mesh)
above the antenna might be usable, but of course, it creates early
reflections from the shield, which may be (or not) harmful depending
on the modulation method.

....been tried. We can get dome time just about anytime we need it.

Don't lock at all or sometimes drops out of lock ?

Drops, *frequently*.

How long does it take to relock ? Is this time longer than an other
antenna with separate receiver typically maintain lock with a moving
receiver ?

Lock and negotiation takes longer than a single hop. I don't understand the
second question.

We are double-sending data, once for each antenna. The radios take care of
that overhead (something we lose at 900M; see above).

There are several ways to implement spatial diversity, e.g. on the
base band level switching between full receivers or simply RF antenna
switching when the received power drops on current antenna.

We do a dumb switch, once on each TDMA slot, currently. Actually, that would
only help on the base. The mobile unit doesn't have any diversity.

To speed up relocking one might also notice that since at least one
receiver usually remains locked, this information can be used to limit
search options on the unlocked receiver.

Don't think that's going to help us, at least directly.

 

[Paul Keinanen]

The data rate is just under 1Mbps. We're using ~480Kbps of it, TDMA, IIRC.

I am still trying to understand the RF-characteristics of the signal.

Is this direct TDMA like GSM mobile phone in which up to 8 handsets
share a common 200 kHz RF channel and each handheld sending the data
as a burst within the allocated time slot. This works well for GSM at
900 MHz and reasonably well at 1800 MHz, so with a significantly
larger bandwidth it should work well on 2G4.

Or are those individual signal time multiplexed into a single baseband
and the modulate a single spread spectrum "carrier".

How wide is the actual emission ? Asking in a different way, how many
(non-overlapping) RF channels can be selected ?

For SS, the spread signal (chip rate) should be one or two orders of
magnitude faster than the actual data.

It is (it's certainly not outside the dome, if that's what you mean). It's
been tried in the ceiling, on the field, in the stands... We've tried
directionals, omnis, amplified, padded, just about everything. One thing we
haven't done is separating the transmit and receive antennas (so they can be
padded differently). The solution, so far, is to abandon 2.4G, in favor or
900M, which works reasonably well (except for other system limitations - e.g.
bandwidth).


...been tried. We can get dome time just about anytime we need it.

A sports dome would be absolutely the last place I would use any ISM
band for any professional communication .

Setting up your system before the event and everything seems to work
OK and there are plenty of SNR etc.

Then the huge public is admitted into the dome, each carrying one or
more ISM devices (Bluetooth, WLAN) etc. While with spread spectrum
emissions, you do not have a discrete frequency channel free/occupied
situation, but adding more and more spread spectrum devices into the
same frequency band will gradually increase the background noise level
and the SNR after despreading drops gradually, until the SNR drops too
low and the communication fails.

In the dome, the RF emitted from each device is bouncing around and
finally absorbed by some soft tissue (the spectators).

Drops, *frequently*.

Sounds very much like NBFM (12.5/25 kHz BW) on 1.3 GHz (23 cm) that I
used a few decades ago.

As the theory predicts, a simple ground reflection will create a comb
filter like spectrum and for a specific narrow band signal channel,
there are deep nulls at very close distance from each other. Moving
just a meter and the signal drops several times. Stopping a car with a
fixed antenna at traffic light would almost always cause stopping at
the multipath null . The only thing that helped, is moving the
antenna a few centimeters (spatial) or changing channel (frequency
diversity).

At highway speeds, the dropouts were so frequent (and short) that it
did not affect the readability of the speech (equivalent to
interleaving and ECC in digital communication).

It still sounds that you are suffering from a narrow band signal with
the RF field strength punctured by multipath nulls distributed close
to each other all over your area of interest. Of course, since some
data is lost, you must use quite a lot of error correction bits.

Since you are only using about half of the available capacity, why not
allocate the rest for ECC bits ? Since the multipath nulls create
burst errors, interleaving should be used to convert burst errors to
random errors that can then be corrected by ECC.

The problem with interleaving is that it adds delay, which is annoying
in two way conversation. As you said that the dropouts are frequent
(assuming several each second), this also means that they are short,
thus reducing the interleave delay needed.

Lock and negotiation takes longer than a single hop.

Those might be your worst problem.

I am still not sure what the emission is like. Do you first lock into
some direct sequence spread spectrum signal, then lock into TDMA frame
and finally lock to the individual signal ?

Or are each handset sending individual spread spectrum sequence during
short time slots in a TDMA way ?

I don't understand the
second question.

In a typical moving multipath situation, the signal is 80-95 % above
threshold and only 5-20 % below threshold. In an ideal receiver, the
recovery would be immediate and at least more than 80 % of the time
the signal would be good even with a single receiver. With diversity
reception with receivers, this time would be much closer to 100 %.

However, if a receiver takes more than 50 % of the good signal time to
relock, even two diversity receivers would not be enough to reach even
close to 100 % service times.

 

[[email protected]]

I am still trying to understand the RF-characteristics of the signal.

Is this direct TDMA like GSM mobile phone in which up to 8 handsets
share a common 200 kHz RF channel and each handheld sending the data
as a burst within the allocated time slot. This works well for GSM at
900 MHz and reasonably well at 1800 MHz, so with a significantly
larger bandwidth it should work well on 2G4.

That's how it works (base gets five slots, each mobile gets one). It works
well except in certain locations, mostly domes.

Or are those individual signal time multiplexed into a single baseband
and the modulate a single spread spectrum "carrier".

No, you had the networking layer pretty much right, above. I don't have the
modulation details in front of me (it's not our radio).

How wide is the actual emission ? Asking in a different way, how many
(non-overlapping) RF channels can be selected ?

FHSS over 43 channels, IIRC.

For SS, the spread signal (chip rate) should be one or two orders of
magnitude faster than the actual data.


A sports dome would be absolutely the last place I would use any ISM
band for any professional communication .

We don't, obviously. It works extremely well in open stadiums (>95% market
share ;-).

Setting up your system before the event and everything seems to work
OK and there are plenty of SNR etc.

Nope, that's the worst-case situation.

Then the huge public is admitted into the dome, each carrying one or
more ISM devices (Bluetooth, WLAN) etc. While with spread spectrum
emissions, you do not have a discrete frequency channel free/occupied
situation, but adding more and more spread spectrum devices into the
same frequency band will gradually increase the background noise level
and the SNR after despreading drops gradually, until the SNR drops too
low and the communication fails.

Masses of people absorb a lot of energy. That actually helps.

In the dome, the RF emitted from each device is bouncing around and
finally absorbed by some soft tissue (the spectators).

Right, but enough gets to the receiver. Spectators actually help without the
dome.

Sounds very much like NBFM (12.5/25 kHz BW) on 1.3 GHz (23 cm) that I
used a few decades ago.

As the theory predicts, a simple ground reflection will create a comb
filter like spectrum and for a specific narrow band signal channel,
there are deep nulls at very close distance from each other. Moving
just a meter and the signal drops several times. Stopping a car with a
fixed antenna at traffic light would almost always cause stopping at
the multipath null . The only thing that helped, is moving the
antenna a few centimeters (spatial) or changing channel (frequency
diversity).

We don't have that option, if it did work. The radios aren't necessarily
moving, though.

At highway speeds, the dropouts were so frequent (and short) that it
did not affect the readability of the speech (equivalent to
interleaving and ECC in digital communication).

It still sounds that you are suffering from a narrow band signal with
the RF field strength punctured by multipath nulls distributed close
to each other all over your area of interest. Of course, since some
data is lost, you must use quite a lot of error correction bits.

Not really. We do a double-send and throw away faulty packets (CRC checked)
but there aren't any correction bits, per se. Everything is G927 compressed
so that hides a lot of uglies.

Since you are only using about half of the available capacity, why not
allocate the rest for ECC bits ? Since the multipath nulls create
burst errors, interleaving should be used to convert burst errors to
random errors that can then be corrected by ECC.

Don't think we have access to the bits. I'll have to look more closely into
that, though. If we had access to them I'm sure we'd be making use of them
for more data.

The problem with interleaving is that it adds delay, which is annoying
in two way conversation. As you said that the dropouts are frequent
(assuming several each second), this also means that they are short,
thus reducing the interleave delay needed.

Each slot is .5ms, ten slots (five transmit, five receive) per hop. The
system latency is pushing the limit for intelligibility (80-100ms). There is
a *lot* of echo-cancellation stuff going on to minimize these issues.

Those might be your worst problem.

Pretty hard to eliminate. The radios have to know who they're talking to
before they can respond appropriately. Then there is the data in flight...

I am still not sure what the emission is like. Do you first lock into
some direct sequence spread spectrum signal, then lock into TDMA frame
and finally lock to the individual signal ?

Or are each handset sending individual spread spectrum sequence during
short time slots in a TDMA way ?

Yes. At least I think that's the way the radios work. They all know ahead of
time what the sequence is. This is negotiated once during system setup while
they're wired together (actually, downloaded from the base).

In a typical moving multipath situation, the signal is 80-95 % above
threshold and only 5-20 % below threshold. In an ideal receiver, the
recovery would be immediate and at least more than 80 % of the time
the signal would be good even with a single receiver. With diversity
reception with receivers, this time would be much closer to 100 %.

However, if a receiver takes more than 50 % of the good signal time to
relock, even two diversity receivers would not be enough to reach even
close to 100 % service times.

I'll have to think about that (SWMBO is yelling to leave).