P25 simulcast multipath interference

[KC0CSE]

lost

You guy's are leaving some of us in the dust....I'm lost....is it "fix able"?....

PRO 106
PRO 197...........LIVING WITH SIMULCAST.....(With a paper clip antenna)

 

[jsncrso]

Yes, with a yagi antenna pointed at one site with attenuation on (to block signals from all other antennas). I scan a very difficult system (so much so that out county is installing a multicast backup...we are mostly water) with 100% decode rate on a BCD996xt. In fact I've many times picked up a transmission more clearly than a $4,000 county radio sitting beside me. If I use the stock antenna I get about a 30% decode rate on average. Attenuation of the other sites is key.

 

[rak313]

You guy's are leaving some of us in the dust....I'm lost....is it "fix able"?....

PRO 106
PRO 197...........LIVING WITH SIMULCAST.....(With a paper clip antenna)

I am not an expert in P25. And I have limited experience with with it (just the software defined radio project with the USB dongle - and the countless papers I have read in the past 2 weeks). But from what I have seen -I believe one should be able to make a good P25 scanner at a low price point.

Is it "fixable" as in adding firmware to the existing Uniden and GRE scanners - I have no idea - because I do not know what hardware is in those scanners.

 

[jsncrso]

I am not an expert in P25. And I have limited experience with with it (just the software defined radio project with the USB dongle - and the countless papers I have read in the past 2 weeks). But from what I have seen -I believe one should be able to make a good P25 scanner at a low price point.

Is it "fixable" as in adding firmware to the existing Uniden and GRE scanners - I have no idea - because I do not know what hardware is in those scanners.

If a $4,000+ radio, and the fact my county is spending a couple million bucks on a backup multicast system to address P25 simulcast issues that the Motorola techs have mitigated but not completely solved, I hiiiiiighly doubt a "cheap scanner" can be made to address those issues. If a radio is receiving multiple conflicting carrier waves the same time, there is no way without significant software-side processing that the radio's DSP can figure out what wave is which bit. The absence of an alternate clock signal between site and receiver makes this a real issue. This is a physical layer issue with electromagnetic waves, which unfortunately precludes any higher digital layers.

But technically yes, a cheap scanner can easily scan P25 Simulcast, I currently have a working example. You just have to have the correct antenna setup like you do with any other system or band you want to scan.

 

[rak313]

If a $4,000+ radio, and the fact my county is spending a couple million bucks on a backup multicast system to address P25 simulcast issues that the Motorola techs have mitigated but not completely solved, I hiiiiiighly doubt a "cheap scanner" can be made to address those issues. If a radio is receiving multiple conflicting carrier waves the same time, there is no way without significant software-side processing that the radio's DSP can figure out what wave is which bit. The absence of an alternate clock signal between site and receiver makes this a real issue. This is a physical layer issue with electromagnetic waves, which unfortunately precludes any higher digital layers.

But technically yes, a cheap scanner can easily scan P25 Simulcast, I currently have a working example. You just have to have the correct antenna setup like you do with any other system or band you want to scan.

If the motorola techs cant make it work for the professional field radios - then that system is not designed correctly. I dont think much of simulcast - but its what is being used alot. If most simulcast systems had problems with the pro radios - then simulcast would have a bad reputation - and maybe it does - but I haven't heard much of that from the users.

As I mentioned at the start of this thread - one can design an equalization filter like they do for multipath on your cell phone. Removing the echos that result when a signal bounces off multiple paths before it gets to the antenna, is exactly what an equalization filter does. Receiving 2 or more copies (from the multiple transmitters) of the same signal - delayed a bit from each other - seems to me - very similar to multipath. (It was pointed out that the delay spread - between multiple transmitters - is designed to be small compared to the 208 microsec symbol time and would make an equalization filter unnecessary).

The question that started this thread - what makes the $4k radio work - and the consumer scanner not work. You are saying that for the system in your county - neither work.

However, your solution it pretty simple, and about the only one that can be done with the current crop of scanners.

 

[beachmark]

If the motorola techs cant make it work for the professional field radios - then that system is not designed correctly. I dont think much of simulcast - but its what is being used alot. If most simulcast systems had problems with the pro radios - then simulcast would have a bad reputation - and maybe it does - but I haven't heard much of that from the users.

As I mentioned at the start of this thread - one can design an equalization filter like they do for multipath on your cell phone. Removing the echos that result when a signal bounces off multiple paths before it gets to the antenna, is exactly what an equalization filter does. Receiving 2 or more copies (from the multiple transmitters) of the same signal - delayed a bit from each other - seems to me - very similar to multipath. (It was pointed out that the delay spread - between multiple transmitters - is designed to be small compared to the 208 microsec symbol time and would make an equalization filter unnecessary).

The question that started this thread - what makes the $4k radio work - and the consumer scanner not work. You are saying that for the system in your county - neither work.

However, your solution it pretty simple, and about the only one that can be done with the current crop of scanners.

Actually Rick, he is in a unique place where the simulcast overlap may be pretty much unfixable by mortal man. It is Dare County NC, with encompasses most of the Outer Banks of NC. Look it up on the map: it is 60-70 miles long and has the Atlantic on one side and the Pamlico Sound on the other, except in the swampy part to the west where nobody much lives. The radio signals get out on the saline water with a high conductivity value and just plain go everywhere from every simulcast site. So the normal overall methods of timing sync'ing the sites just fall apart as there is no limit to where the signals end up.

My only wonder is why this was not seen right up front for Dare Co; it is the most obvious place in the world to have this issue, and there has been plenty of experiecnce in 800 and 1900 cellular in the past 2 decades in water-bound areas like Tampa that screams "YES, this will be a problem!!" More sites and directinal antennas with downtilts would help some, but it sounds like they are essentially going non-simulcast in some areas to solve the issue more cheaply.

Actually an equalizer like in GSM, which in the lab allowed up to 3 symbols of time overlap, and which in practical field experience allowed 2+ symbols of overlap with the system still operating OK, should be a big help in that situation. I am not sure if it would fix it all, but it is prime case for such an animal.

And, OBTW, I did find an oblique reference to equalizers for P25 in a Tait document, but I have not found any info to back up if they really are being used in P25 recvieers. I'll keep digging. So I can't say if that is a factoralready at work in the Dare Co situation or not.

(PS I'll comment later on the Matlab stuff you presented....it's fascinating and very kind of you to post all of that!)

 

[beachmark]

You guy's are leaving some of us in the dust....

Naw, it's just showing off....well, rak1313 is; I am just faking it! LOL I hope you don't mind the threadjacking; this is all very interesting stuff to some degree.....

 

[kv5e]

Simulcast RF Design

Simulcast RF design is much different than single site designs:

1. Sites must be lower HAAT, power, and have downtilt antennas to limit radiation on the horizon
2. At 9600 bps symbol rate, the delay spread should not exceed 1/4 symbol width
3. Many more sites are needed than non-simulcast

The problem areas from delay spread are where near equal power RF is received from a near and more distant site where the delay spread is greater than 1/4 symbol width. At 9600 4-level this works out to ~ 52 microseconds where delay spread effects push decoding over the cliff.

The propagation delay from sites more than ~9 miles away are the main culprits of delay spread. This can be from direct Rician type signalling or even more Rayleigh type multipath distorted energy.

In urban environments at frequencies above ~700 megahertz, urban clutter introduces additional challenges with shadowing which can produce more near equal power areas where the delay spread challenges appear.

I know that many systems are constrained by leasing/costs and the decision is made to utilize higher sites and fewer of them. This kind of design for simulcast is asking for increased delay spread problems.

Additionally, equal power between C4FM simulcast signals can introduce AM nulls so some form of controlled beating needs to be calculated and frequency offsets added to base station carriers which do not alias with the codeword decoding of the 4FSK signalling. This can be done with simulated annealing and an algorithm to calculate the offset values.

The demodulators in the subscriber radios are optimized to minimize delay spread problems, but even expensive radios cannot overcome poor simulcast design when it is attempted on a budget.

Simulcast design is a black art, and not all engineers are good at it.

Craig

 

[jsncrso]

Kv5e pretty much summed up what I wanted to say. And Beachmark is correct. The water, inversion layers, and being in direct LOS of multiple antennas can play havoc on the system here. Fortunately, many of the kinks have been worked out now a year and a half after our system went online, though they are still pursuing the multicast backup option the last I heard (In other words, it's reliable now, but a year ago, it was horrible).

Also rak313, equalizers might help,but you have to remember that a signal reflecting off of a building has a much lower gain and much less of a delay spread than a signal directly from another tower, so they aren't going to be nearly as effective.

I will say aside from the simulcast issues, scanning a simulcast system is better than scanning a multicast system, as you hear traffic from all over the county in one system, versus the time consuming process of scanning each multicast site as a separate system.

 

[beachmark]

I am not an expert in P25. And I have limited experience with with it (just the software defined radio project with the USB dongle - and the countless papers I have read in the past 2 weeks). But from what I have seen -I believe one should be able to make a good P25 scanner at a low price point.

Is it "fixable" as in adding firmware to the existing Uniden and GRE scanners - I have no idea - because I do not know what hardware is in those scanners.

I think you start with a big advantage.....if your equipment has a true IQ demodulator. I am pretty convinced that the scanners have a detector that only is a form of FM detector. This would have worked OK to decode the P25 Phase 1 CCH that used C4FM... until LMS came along with the QPSK modulation on the CCH.

P25 Phase II uses DQPSK on the downlink for both CCH and TDMA voice channels; any standard FM dectector is going to do some 'damage' to the amplitide part of the DQPSK waveforms that will result in significant phase errors.

 

[beachmark]

I can understand you doubting this - so did I when I first came across it. But your ascertain that you are just getting the same info over and over - is not correct. As long as there is (unbiased) noise in the system of a few A/D bits, you will get new information each sample.

And you are right - you need to make sure that the LSB's of the A/D are into the noise. Otherwise the noise contributed by the LSB of the A/D will not be splattered uniformly across all frequencies - but can act like a non-linear device, and create spurious signals.

Attached is an example i made using Matlab. This is not a very practical example - but it illustrates how one can get information from oversampling.

I created 2 complex signals sampled at 1,048,576 Hz - sig1 is at amplitude 120, and frequency 4096 Hz.
sig2 is at amp 0.2 and frequency of 1024 Hz. I also created a gaussian noise source with RMS amplitude of 4.

I add these 3 items (sig) and truncate (throwaway the fraction of the number) with the resulting signal representing an 8 bit signal (with values between +/- 127 ). At this point - one would think that the sig2 would be lost - as its less than 1 LSB.

Note: that 120 is 55 dB larger that 0.2.

I took 65 k samples and process them in an FFT, The results were scaled such that an amplitude of 128 would be 0 dB. Notice in the top figure you see both signals at the correct relative amplitude, and way above the noise.

The 2nd figure shows the matlab window where the 1st 10 samples (both real and imaginary components) of the signal (sig) are printed in the window to illustrate that the signal is made of only integers -128 to + 127, meaning 8 bits. The 3rd figure is the matlab source to create this example.

Now it is unlikely that one can get (65k) 48 dB of gain in any real 8 bit A/D before some spurious signal showed up. An A/D will spec SPDR - spurious free dynamic range - and that tells you how much gain you can expect to get.

As an aside - most 24 bit (18-19 bit effective) audio A/D converters are 1 bit devices sampling in the 1-5 MHz range, followed by complicated noise shaping feedback and digital filters. They achieve 18 bit performance by not only oversampling, but by making sure the spectrum of the noise introduced by sampling only 1 bit, is out of the desired band. There are seismic versions good to 22 bits (with a very low bandwidth),

First, let me thank you for being so kind to post all of this. It is great so see so much info and work being shared.

OK, I am starting to believe. I know the dynamic range some new generation DSP repeater products out there, and this starting to explain why they get such good dynamic range.....much more than one would calculate from just 6 dB per A/D bit without going to true 18-20 bits of A/D. There must be something in the noise being in there that somehow randomly and independently codes the signals so they can be separated and get something akin to compression gain out of the process. (Sorry...I understand PN coded processing gain...I have to relate this to that for now!)

Weel, looks like I have some homework to do.....but that is a good thing. Thanks for opening my brain!

And let me know when you want to go into biz.....I see a new branch of our family RF biz....best scanners in the galaxy....

 

[PiccoIntegra]

Have any of you guys come across the 455 KHz IF Downconverter page yet? The guy that put that circuit together is one of the OP25 project developers, Max(KA1RBI). I'm surprised he hasn't chimed in on this discussion. I linked his call sign to his posts here on RR. He has talked about that circuit here before. In the circuit notes he mentions using an IC rather then the toroidal mixers. I would really like to see something like this put together, but I don't have the knowledge to do it myself. I keep visiting that page once a week to see if there has been any updates. :lol:

As far as the digital scanner schematics are concerned, I've only seen one set of schematics out in the wild. It was posted here in the forums, but the link is now 404. It was for the RS Pro-96. You may be able to order current Radio Shack service manuals from them. You won't find any for the Uniden models. But they all are designed on the same opamp based slicers, at the discriminator output.

 

[jsncrso]

Schematics are only a small part of the equation, you have to remember software and programming make up the bulk of what happens at a scanner, especially the new DMA scanners. You can build circuits and electronics all day long, but in the end you need good software in order for them to make them work the best, it's one reason why GRE and Uniden puts out new versions of scanner firmware. Without good software and programming, a scanner is nothing more than a brick.

 

[PiccoIntegra]

There isn't any amount of "good software and programming" that can compensate for poor hardware design. Once the signal is mangled, which is exactly what happens at the discriminator level on LSM based systems, you cannot reliably recover the information from the bit stream.

 

[beachmark]

There isn't any amount of "good software and programming" that can compensate for poor hardware design. Once the signal is mangled, which is exactly what happens at the discriminator level on LSM based systems, you cannot reliably recover the information from the bit stream.

Bingo.....'though I would not call a discriminator or quadrature detector a poor hardware desing...it is just the wrong hardware for any of the DQPSK modulations (which includes LSM control channels on the donwlink). At least that is the best I can conclude at this point for some scanners...

 

[rak313]

First, let me thank you for being so kind to post all of this. It is great so see so much info and work being shared.
....

And let me know when you want to go into biz.....I see a new branch of our family RF biz....best scanners in the galaxy....

you're welcome. I have learned a good deal from your posts as well. Thank you too!

As for the "new business" - I am a recently retired sonar engineer. I did have some experience in mobile radio working for GE 30 years ago (when AMPS was not out yet). Things have really changed.

I think a small device - with very minimal display and buttons - with a USB on 1 end (used to program for your system - and as a mechanical holder when in use) - antenna port - and small speaker - would make a good scanner.

Or even better - an iphone or android AP that was a p25 scanner. Not sure if its technically possible though - probably not.


BTW I just came across this - a link to the notes of a modern course in RF CMOS design

http://www.ece.ucsb.edu/yuegroup/Teaching/ECE219Winter2013/ECE219 Win13 course syllabus v1.2.pdf

The notes are here:
https://www.dropbox.com/sh/d6fw6zvftceu7wf/YmcfVuwKpq/Lecture slides and notes

The beauty of this is its a modern course from a college professor, So it should be up to date - and reliable.

 

[beachmark]

Have any of you guys come across the 455 KHz IF Downconverter page yet? The guy that put that circuit together is one of the OP25 project developers, Max(KA1RBI). I'm surprised he hasn't chimed in on this discussion. I linked his call sign to his posts here on RR. He has talked about that circuit here before. In the circuit notes he mentions using an IC rather then the toroidal mixers. I would really like to see something like this put together, but I don't have the knowledge to do it myself. I keep visiting that page once a week to see if there has been any updates. :lol:

As far as the digital scanner schematics are concerned, I've only seen one set of schematics out in the wild. It was posted here in the forums, but the link is now 404. It was for the RS Pro-96. You may be able to order current Radio Shack service manuals from them. You won't find any for the Uniden models. But they all are designed on the same opamp based slicers, at the discriminator output.

Thanks for the links ot the downconverter page...that is the right kind of stuff. There is no limiter to wreck the amplitude components of the QPSK signals as would be found in a standard FM discriminator/detector. The notes on the revised version are not complete but it is certain to work fine based on what is there.

So that still leaves the question of what detection hardware is in the scanners....???

Again, as I think has been said, the C4FM modulation used in the P25 Phase I control channels should be reasonably detectable with a standard FM detector. It is when LSM and the QPSK modulations of P25 Phase II on both control and voice channels came along that this starts being a serious problem.

Another dumb question: Are folks using these scanners to listen to the talkback mobile/portable transmissions back to the site as well as the talkout transmissions from the site? I ask since the P25 Phase II modulations are different in the 2 links (QPSK talkout versus CPM talkback). 1) The post-demodulation decoding of the actual bits received in the talkback link is quite different, and 2) the multipath performance looks to be inferior in the talkback link.

If you care to see the symbol coding complexity in the P25 Phase II talkback modulation look here:
http://www.aeroflex.com/ats/product...Testing P25 Phase 2 TDMA Application Note.pdf

Basically, you have to properly demodulate 2 or more symbols in sequence, to be able to figure out the bits tramsitted in the first symbol. I.e., if you have a sequence of symbols ABCD..... , you have to demodulate symbols A and B to decide what bits were in symbol A, then demodulate symbol C and look at it along with symbol B (and maybe symbol D too) to decide what bits were in symbol B, and so on. So, in the talkback link, there is no simple 1 to 1 correspondence of a phase symbol to a specifc pair of bits.

So if listening to the talkback on a scanner, the scanner DSP processing is going to have to do this well also. That is why I asked if folks having scanner problems are listening to the talkback.

 

[beachmark]

As for the "new business" - I am a recently retired sonar engineer. I did have some experience in mobile radio working for GE 30 years ago (when AMPS was not out yet). Things have really changed.

So did you work in Lynchburg VA??

 

[rak313]

So did you work in Lynchburg VA??

No, I was in Syracuse,NY working on a personal communication venture project that never got allocation of frequencies from the FCC - so it was cancelled. But we did have contact with the guys in Lynchburg - and I visited there a few times.

BTW, their 1st cellular radio was built like a tank, cost and weighed as much.

 

[xmo]

beachmark: "...It is when LSM and the QPSK modulations of P25 Phase II on both control and voice channels came along that this starts being a serious problem. ..."
-------------------------------------------------------------------------------------------------

Actually, LSM isn't really a Phase II thing nor is QPSK.

Phase two modulation is described in TIA TSB-102.BBAA

"Two-Slot TDMA Overview"

Basically, it says:

"The 12.5 kHz physical layer operates at a gross bit rate of 12 kb/s.

Harmonized Continuous Phase Modulation (H-CPM) is defined for the inbound link so that subscribers may use constant envelope modulation.

Harmonized Differential Quadrature Phase Shift Keying (H-DQPSK) is defined for the outbound link so that
simulcast delay spread is improved as compared to Compatible 4 Level Frequency Modulation (C4FM) modulation used in Phase 1.