Uniden P25 Simlucast - Possible Solution to Ponder

[AK9R]

deleted

 

[Alliance01TX]

Thanks

Thanks Gents

And do want to thanks the folks that took the time and effort to provide good Technical information, ponder possible solutions and look forward too....

These help both educate and provoke thought - not conflict - in a meaningful manner.

Pondering the question was the intent and appears several solutions (not one) are perhaps on the horizon and that will improve the hobby...

I have never doubted the experts that are far afield of my capabilities - I have taken issue with a small few whom despite their superior knowlege have less than average managerial and people skills in allowing for free-flow of ideas and concepts - and go on to personal attackts levels for whatever reason...

Thanks all and will ask the thread to be closed and please do continue the "what if" as at some point it may help with the Next Gen Products.

Bill

 

[JamesO]

Some of the more recent comments are doing a better job getting outside of the box and tossing some more interesting ideas and concepts into the mix.

There may not be a real solution for any or all of this, but again, there may be some small steps that some can implement that offer some improvements or stability.

Who knows the solution may be a hybrid receiver design with a zero IF conversion stage somewhere in the signal path?

With the cost of the SDR dongles being so cheap and small, maybe these can be merged into the next generation of receivers.

The other issue is the scan time and reconfiguration and settling time of the receiver. But maybe there will be 2 receivers sections in the next generation of radios as well?

But to all the innovators, experimenters, hackers and crackers, keep and open mind, think outside the box and if you have anything to offer, even conceptually, why not speak up.

 

[slicerwizard]

Yep P25 Waiting Time is still there and alive. Just amazing.

Thanks for the confirmation. It looks like Uniden is just reusing the same old code (and hardware). The silly thing is that the firmware already knows how to look for the P25 sync pattern - it has to in order to decode P25 control channel data; so on a voice channel, if the sync pattern is seen, just mute the raw audio immediately. Problem solved - no more motorboating.

It actually gets worse then that... read my post about their new squelch table solution.. mind boggling. They're trying to fix this by holding the squelch open consistently.

Well, I think I would also ignore squelch on digital channels. If you're getting good decodes, you have a signal; otherwise, it's time to move on.

I certainly agree with the filtering aspect - CQPSK creates a rapid frequency shift at every symbol edge, which a typical C4FM filter will destroy. And the looser filters required for simulcast modes would degrade C4FM decoding, so there is no one size fits all filter.

Some users are clamoring for DMR and/or NXDN decoding in these new scanners There seems to be a perception that a firmware update can solve their problem and that Uniden is just holding back an "extreme upgrade" that will decode additional modes. Based on what you guys are saying, if there is a basic deficiency in the discriminator design of these new scanners, can firmware get us these additional modes? Is the current design a dead end WRT decoding anything but P25?

The modulation modes used by NXDN and DMR/TRBO can be handled by a standard FM demodulator. P25 uses the IMBE vocoder, while the others use AMBE+2; if the DSP chip that Uniden is using doesn't handle AMBE+2 (and it likely doesn't), then you have a big problem.

With the cost of the SDR dongles being so cheap and small, maybe these can be merged into the next generation of receivers.

Oh hell no. We don't want the horrible RF specs of a cheap dongle. Uniden front ends are bad enough already.

The other issue is the scan time and reconfiguration and settling time of the receiver.

Since when? Where did this idea come from? Why is it an issue for LSM and CQPSK, but not C4FM?

But maybe there will be 2 receivers sections in the next generation of radios as well?

Unlikely and generally not required. The micro is busy enough just dealing with one. And how would two receivers fix a "scan time and reconfiguration and settling time" issue?

 

[slicerwizard]

The modulation modes used by NXDN and DMR/TRBO can be handled by a standard FM demodulator. P25 uses the IMBE vocoder, while the others use AMBE+2; if the DSP chip that Uniden is using doesn't handle AMBE+2 (and it likely doesn't), then you have a big problem.

An anonymous coward has pointed out that the newer Unidens support P25 Phase II, which uses the AMBE+2 vocoder, so there is no technical reason for not supporting NXDN and DMR. RAS does introduce a wrinkle though, as Motorola messes with message CRCs, so the ability to verify message integrity is compromised, but that was never Uniden's strong suit anyway.

 

[redburgundy]

Perhaps slightly off-topic, but in my experience, setting the P25 Threshold value manually to 11 for the 436 gives improved performance for my local trunked non-P25 Motorola system (Montgomery County MD).
Does that mean the P25 Threshold setting affects the IMBE/AMBE vocoding?
If so, in what way?

 

[slicerwizard]

Perhaps slightly off-topic, but in my experience, setting the P25 Threshold value manually to 11 for the 436 gives improved performance for my local trunked non-P25 Motorola system (Montgomery County MD).

Setting the threshold manually to the optimal value eliminates the degraded decoding that occurs while the radio is hunting for the best threshold levels.

Does that mean the P25 Threshold setting affects the IMBE/AMBE vocoding?
If so, in what way?

It affects it by feeding the vocoder with voice frames that contain fewer bit errors. When the bit error rate climbs above 10%, the decoding falls off a cliff as frames become completely unrecoverable. Radios handle this by repeating the previous 20 ms voice frame, which leads to the diiiiiigitaaaal sounnnnnnnds that you hear. When it gets bad enough, the radio just mutes the audio altogether.

 

[redburgundy]

Setting the threshold manually to the optimal value eliminates the degraded decoding that occurs while the radio is hunting for the best threshold levels.

The P25 Threshold value corresponds to a set of three numbers.
For the 436, the middle number seems to be always 1.65.
What do those numbers mean?

 

[slicerwizard]

The P25 Threshold value corresponds to a set of three numbers.
For the 436, the middle number seems to be always 1.65.
What do those numbers mean?

I would expect that they represent the three audio voltage levels that set the boundaries (decision points) for converting a signal into one of the four possible digital symbol values (dibit values)

http://home.ica.net/~phoenix/wap/Misc/Proper HDU Decision Point Tracking.PNG

In the image, the green bits show a P25 signal after FM demodulation and some noise filtering; each green chunk represents one symbol (2 data bits). The blue bars represent the average voltage of each green chunk. The 3 white lines are the decision points that determine which of the four possible symbol values each of the blue bars represent. If a blue bar extends past the top white line, the symbol is a +3; if a bar ends between the top and middle bars, it's a +1; bars that end below the middle line are -1; and bars that extend below the bottom line are -3.

The middle white line represents the center of the signal; that would be the equivalent of the 1.65 value. It's probably fixed because the designers know what voltage represents the center of a properly tuned signal. If the scanner is a bit off frequency, that fixed 1.65 value might cause problems. With Unidens, you can check if the scanner is off frequency by using the Window Voltage command that's built into the serial control protocol. You could also try deliberately tuning a bit off frequency, say 0.5 or 1 kHz high or low and see if the 1.65 value moves up or down during digital voice calls to compensate for the tuning error.

The other two numbers represent where the lower and upper white lines are placed. You'll probably find that they're generally both about the same distance away from the 1.65 center value; a real world example would be 1.43 1.65 1.86; 1.43 is 0.22 below 1.65 and 1.86 is 0.21 above.

That P25 waveform starts with 14 +1 symbols and the next symbol is a +3; if the white lines were very close together, some of those 14 +1 symbols would extend above the upper white line and would be erroneously decoded as +3 symbols. If the lines were too far apart, that +3 symbol might not reach past the top line and would be treated as a +1 symbol instead. Getting the symbol values wrong is what causes degraded digital audio.

Because this signal came from a discriminator tap plus sound card interface, the audio wanders up and down a fair bit and the white lines (decision points) have to be constantly adjusted to represent the correct boundaries between the four symbol values. Inside a scanner, this wavering shouldn't be as bad. The Threshold Value that you pick sets the initial values for these three decision points, i.e. close together or more spread apart, and then the scanner fine tunes them while decoding a digital call, so picking the right value means that the scanner has to do minimal tweaking to get dialed in.

 

[jcardani]

Hi Slicerwizard,

Great writeup on the threshold info!

Not to get too off topic, but while it's on my mind, can you please summarize the logic of sound card sampling 1) discriminator for C4FM up until di-bit recovery and 2) the radio's 450 KHz IF for I/Q CQPSK up until di-bit recovery. The sound card for the latter will either have a 12KHz I and 12KHz Q signal input (for hardware I/Q) or a 24KHz input (if I/Q will be done via software). I would like to get started with experimenting with some sort of solution but would like to understand the logic and how to code this better and you have a great grasp of these concepts. I use Microsoft Visual C++ 2010.
I will eventually port to DSP or dsPIC though.

thanks much!

 

[slicerwizard]

Those are pretty slow sampling rates. P25p2 is 6000 symbols per second, so you'd have two or four I/Q pairs per symbol? For P25p1, you'd have 2.5 or 5 I/Q pairs per symbol. I think that would lead to a number of issues, like aliasing and poor symbol centering for starters.

 

[jcardani]

Hi Slicerwizard,

I came up with those numbers from KA1RBI's downconverter that converts the 455 KHz IF to 24KHz and it's sampled at 96KHz by the sound card.

So what would be the best solution ?

thanks

 

[slicerwizard]

Ok, so those aren't sampling rates - they're what the 455 kHz IF signal is downconverted to. You should be taking the raw I/Q pairs and applying a noise filter. You then need to detect and quantify the changes in the I/Q dot's rotational speed as the CQPSK transmitters apply phase shifts for each symbol. The last step is the easy bit - taking those phase shifts and synthesizing a C4FM signal.

Personally, I'd initially play with this on a PC with a DVB-T dongle. It's easy to record some CQPSK I/Q streams and then work with them.

 

[SCPD]

It might be possible for Uniden to retrofit their existing design without resorting to a full blown low-IF to IQ stage. Since the FM discriminator strips out the amplitude, a second A/D converter could be used to sample the amplitude - side by side with the FM deviation. The radio already handles AM demodulation so this shouldn't be a stretch. The CQPSK signal is strongest at zero deviation - and weaker at the symbol decision points. Those peaks in amplitude could be used to drive their clock recovery instead of the peaks from the FM discriminator. The big difference is the new timing reference marks the boundary between symbols instead of the symbol center.

This would only be relevant to CQPSK signals and useless for C4FM.

Max (if he's still reading) can speak up as to whether this might work.

http://forums.radioreference.com/di.../283705-lsm-why-scanners-suck-real-story.html

If the radio's microcontroller already has separate A/D inputs for AM and FM - this could *theoretically* be a firmware change (but don't hold your breath).

 

[KA1RBI]

It might be possible for Uniden to retrofit their existing design without resorting to a full blown low-IF to IQ stage. Since the FM discriminator strips out the amplitude, a second A/D converter could be used to sample the amplitude - side by side with the FM deviation. The radio already handles AM demodulation so this shouldn't be a stretch. The CQPSK signal is strongest at zero deviation - and weaker at the symbol decision points. Those peaks in amplitude could be used to drive their clock recovery instead of the peaks from the FM discriminator. The big difference is the new timing reference marks the boundary between symbols instead of the symbol center.

This would only be relevant to CQPSK signals and useless for C4FM.

Max (if he's still reading) can speak up as to whether this might work.

http://forums.radioreference.com/di.../283705-lsm-why-scanners-suck-real-story.html

If the radio's microcontroller already has separate A/D inputs for AM and FM - this could *theoretically* be a firmware change (but don't hold your breath).

Another related post is this one:
http://forums.radioreference.com/so...ardware-scanners-receivers-2.html#post2110408

The "AM" tone wouldn't really add anything that we couldn't already derive from the "FM" (disctap) signal. Don't know how the eponymous unitrunker performs symbol recovery now when it's reading samples off of the disctap, but presumably there must be a symbol timing / recovery loop that software-locks to the incoming waveform and that acquires the proper phase offset within it in order to grab symbols. From this it would be relatively trivial to phase lock a software-generated sinusoid (if we wanted to).... This might even be (for practical purposes) superior to the tone that we would obtain from the "AM" demod, since it would be free of amplitude noise....

The real problem is that the FM demod of the complex constellation doesn't give you a representation of the instantaneous phase - it gives you a voltage corresponding to the frequency, (which is the first derivative of the phase w.r.t. time). Accordingly if you wanted to know the phase you could attempt to integrate the disctap voltage. This may be why some specs tell you that the proper way to demod CQPSK is integrate-and-dump of the output of the FM detector (disctap).

It might be interesting to try an implementation of integrate and dump of the disctap of CQPSK and see if the result would improve the B.E.R. For example referring to Fig. 4 of my "LSM" page you would set an accumulator to zero starting at time zero and then
for (i=0; i < 20; i++) accumulator += sample
at the end of the loop you would set make a record of the accumulator and then set it to zero again, then
for (i=20; i < 40; i++) accumulator += sample
then take another reading of the accumulator (2nd symbol) then rinse, repeat... The result for the first five symbols would be fairly large positive values (watch out you don't overflow the accumulator). Similarly for the sixth symbol you'd expect relatively large negative values, For +/- 45-degree symbols you'd expect the integrated voltages (accumulator values) to be somewhere around 1/3 of the voltages for the +/- 135-degree symbols... I would fully expect the results of this experiment to be sub-optimal if compared to a proper zero-IF demod of the complex QPSK constellation...

Max

 

[JASII]

Uniden P25 Simlucast - Possible Solution To Ponder

For those listening to 800 mHz simulcast systems, has anybody tried using an antenna with no gain to see if that actually improves reception? I have been thinking of trying one of the Comet Miracle Baby antennas to see if the theory of "less is more" would actually improve reception on 800 mHz simulcast systems.

 

[davenlr]

For those listening to 800 mHz simulcast systems, has anybody tried using an antenna with no gain to see if that actually improves reception? I have been thinking of trying one of the Comet Miracle Baby antennas to see if the theory of "less is more" would actually improve reception on 800 mHz simulcast systems.

I put a ST2 antenna up on my roof, ran new coax included, and did a really nice and neat installation, and every one of my P25 sites (simulcast and not) were choppy, audio cut out, scanner would stop on a talk group but no audio would ever come through.

Switched antennas to another ST2, that is mounted much lower, on a ground mount, and all the sites instantly started coming in perfectly.

Apparently, if you live in the city, you want an antenna as you suggest, no gain, and even possibly using the highest loss, cheapest coax you can buy for it

If I wasnt also using the same scanner for some departments that havent made the switch yet, Id just leave the freaking rod antenna on the back of it instead of using a real outdoor antenna.

 

[whsbuss]

I put a ST2 antenna up on my roof, ran new coax included, and did a really nice and neat installation, and every one of my P25 sites (simulcast and not) were choppy, audio cut out, scanner would stop on a talk group but no audio would ever come through.

Switched antennas to another ST2, that is mounted much lower, on a ground mount, and all the sites instantly started coming in perfectly.

Apparently, if you live in the city, you want an antenna as you suggest, no gain, and even possibly using the highest loss, cheapest coax you can buy for it

If I wasnt also using the same scanner for some departments that havent made the switch yet, Id just leave the freaking rod antenna on the back of it instead of using a real outdoor antenna.

Strange how this goes against everything we know in radio. Always trying to get the most/best signal doesn't work on digital systems.

 

[tumegpc]

I put a ST2 antenna up on my roof, ran new coax included, and did a really nice and neat installation, and every one of my P25 sites (simulcast and not) were choppy, audio cut out, scanner would stop on a talk group but no audio would ever come through.

Switched antennas to another ST2, that is mounted much lower, on a ground mount, and all the sites instantly started coming in perfectly.

Apparently, if you live in the city, you want an antenna as you suggest, no gain, and even possibly using the highest loss, cheapest coax you can buy for it

If I wasnt also using the same scanner for some departments that havent made the switch yet, Id just leave the freaking rod antenna on the back of it instead of using a real outdoor antenna.

I did the same thing with the same results. I mounted a Wilson Yagi about 25ft high with a rotor and got terrible P25 Chop.I have another yagi just sitting inside above my desk and performs so much better with less chop.

 

[SCPD]

my point is WHY should I or any owner of a brand new product advertised as supporting P25 have to modify, accessorize or reverse engineer something to get it to do what should be it's basic functionality?

CQPSK-LSM has been part of the P25 standard since the late 1990s. Uniden should know after not ONE, TWO but now THREE generations of their digital scanner product that the traditional discriminator tap design DOESN'T WORK for reliable reception of CQPSK-LSM modulated signals.

They should know by now that no amount of DSP tweaking can overcome bad data being fed into the vocoder. They should know that other manufacturers incorporate I/Q to get good, clean data to feed to the DVSI chip. If not, then it is time to fire their engineering department and hire some who are competent and experienced in designing modern digital RF devices.

A consumer should not be FORCED to make complex changes to an off the shelf product to get basic, solid performance of an advertised feature. If this were a car advertised as getting XXX MPG and it only gets YYY, then you would complain (and rightfully so) and probably want to seek legal action against that automaker. Oh wait, looks like that's already happening:

MPG Fraud - McCuneWright, LLP - Attorneys at Law

Reliable, consistent reception of P25 is why we are buying these scanners. The fact is Uniden knows, or should have known, that their product has a hardware design flaw, and their remedy should be to correct it at the hardware level.

Their history says they will unlikely do so because re-engineering a product would cost a fortune, not to mention a recall campaign would cut too far into their profit. Look at how well they are handling the "headphone gate". Pathetic. Just pump out more poor performing mediocre product at premium prices and people will buy it. That is America. No one cares about quality anymore.

The OP is 100% right on!! Why should we be quality control for this?. I just love all the posts on how to attempt to fix an issue that should not be ours to fix. I am not going to stand on my head holding an 800MHz yagi, sticking a paper clip into the antenna jack, or play with a 100 menu settings to try and get a decent result. Enough of this is new technology, new scanners, etc. The same design in the scanners has not been changed only new bells and whistles added. Just give me a scanner that works, not one that will make me breakfast.