APCO 25 Software decoding?

grinthock · Post by **grinthock** » Thu Sep 30, 2004 10:55 am

From everything i've read, APCO 25 uses basic QPSK it's not encrypted (depending I know).

Hasn't anyone written a software program for the PC that decodes this? Can we find someone who can/has ?

-- Grint

ExKa|iBuR · Post by **ExKa|iBuR** » Thu Sep 30, 2004 12:55 pm

Well, the problem is licensing issues. IMBE, which is the format that P25 audio uses, is owned by DVSI, and is only licensed by Motorola, Icom, Kenwood, and the other P25 vendors.

It would be just the same as if somebody were to develop a software program that would convert the ISDN that Nortel uses on it's Meridian PBX's into useable analog audio. Probably not the best example, but you get the idea.

-M[/code]

mancow · Post by **mancow** » Thu Sep 30, 2004 5:05 pm

Theoretically all you would need is a 4 level data decoder to interface to the computer running the software right?

Am I way off base?

mancow

Astro Spectra · Post by **Astro Spectra** » Thu Sep 30, 2004 5:07 pm

Actually APCO 25 phase one is just 4FSK. I’m not aware of any real world use of the 6.25 kHz bandwidth compatible QPSK stuff, but I stand to be corrected.

Actually software decoders are quite possible. Obviously Uniden and AOR have implementations that run on their own hardware. Vanu, Inc has demonstrated digital APCO Project 25 waveform decoding an iPAQ-based handheld software radio. See:

http://www.vanu.com/publications/ipaq_hardware_sdrf.pdf

There is at least one undergraduate team that has demonstrated an APCO 25 decoder on a Pentium PC but I can't find the link, Google it yourself.

The most popular way is to buy the cheapest VSLEP or CAI Astro Spectra you can find, junk the front end and mix the 10.7 or 21.4 MHz IF output of your scanner up to the 106.9 MHz Spectra IF using a MiniCircuits mixer like the venerable SBL-1. This uses the Spectra to do the decoding and is easier than tying up a PC. Watch out for the DC volts on the IF input of the Spectra, you’ll need a 1nF blocking cap! Obviously if you want both VSLEP and CAI you’ll need two Spectras and some of the scanner nuts I’ve seen have one of each running off a splitter so they can scan both.

As for the Meridian, there are a bunch of third party products out there that are compatible with the DMS100/SL100 systems.

Nothing impossible but you may have to negotiate an intellectual property (IP) licence from Nortel, Motorola, or DVSI, if you intend to sell a product that uses their IP. That's only fair, they paid big bucks to develop the ideas in the first place!

ASTROMODAT · Post by **ASTROMODAT** » Thu Sep 30, 2004 8:30 pm

AS, are you sure that Uniden and Radio Shack have not paid DVSI to allow them to implement their IMBE decoder?

DVSI is happy to supply IMBE chips, but there is a royalty fee that must be paid to DVSI, on a per chip basis. DVSI also has DSP-based IMBE decoder agreements with huge vendors, such as Motorola, whereby they are charged to use their code, on a per unit basis. They simply load DVSI's protected and licensed code into their DSP's. Obviously, they still pay royalties to DVSI, but in this case it's for the code only, as opposed to the single purpose/dedicated IMBE chips (e.g., a hardware based approach). Either way, DVSI gets paid! It's their IP, and they guard it like a starving junk yard dog (as you and I would, too, if we owned the code!)

Given how beautifully the PRO-96 works and sounds (and it sounds as good as our XTS5000's), I'd be somewhat doubtful that they reverse engineered the DVSI chips, and implemented an IMBE decoder in software.

My point is that if little old Uniden could implement a successful IMBE decoder, sure as heck Motorola would have done it years ago! After 15 years, Motorola is still paying DVSI, so I can't see where they would not have gone it alone if it was possible/legal.

Does anyone know, for sure, that Uniden and/or Tandy are NOT using DVSI chips, and/or they are NOT using their IMBE code (and they are not paying DVSI, in both cases)?

Larry

P.S. BTW, I got a chance to hear the college kids attempt at an IMBE decoder (via the internet), and it sounded like pure s_ _ t! There have been many attempts at IMBE software decoders, and to my knowledge none of them have worked anywhere as near as good as DVSI's chip. I could very well be wrong, but I'm not aware of any successful copycats, to date.

jnglmassiv · Post by **jnglmassiv** » Thu Sep 30, 2004 8:47 pm

grinthock wrote:Hasn't anyone written a software program for the PC that decodes this? Can we find someone who can/has ?

I'm assuming this software would be similar to MDT decoding software. It just appears without a company name in it and just makes vague references to its authors. Underground, I guess.

515 · Post by **515** » Fri Oct 01, 2004 7:29 am

A P25 signal is 4 level FSK, but I don't think a 4-level "data slicer" would be that useful for decoding it...

If you look at a P25 signal from an analog radio's discriminator on a scope, you'll see it doesn't look anything at all like a 4 level Flex/Reflex pager transmission. The P25 waveform is very rounded and the four discrete levels are not obvious like they are on a Flex/Reflex transmission. This is due to the filtering of the transmitted P25 signal to make it fit in the narrowband channel. It's a raised cosine filter if I remember right...

The only way I know of to turn a P25 waveform into raw data is with DSP and digital filtering techniques. I don't have any DSP experience, so I don't really know any more than that. I've spoke with engineers from several P25 radio manufacturers, and asked them if they knew any way to decode a P25 waveform into raw data using only hardware, and every one replied they didn't think it was possible.

Remember this is just the first step--decoding the waveform to raw data. Once you do that, you still have to extract the IMBE voice data, and decode that into an analog signal, which as Larry and others stated is pretty much a software excercise, and is definitely easiest to just choke up the $$$ to DVSI for the chip or code. So really, decoding the IMBE is the easy part. Getting the raw data is more difficult, unless you're a DSP expert.

The question I still have is why does the AOR P25 decoder require a 10.7 MHz IF input? Why couldn't they use a discriminator input? Is there some technical reason, or is it just their preferred way to mate it with AOR recievers that usually have an existing IF OUT jack? Anyone know?

grinthock · Post by **grinthock** » Fri Oct 01, 2004 8:14 am

Not to mention that AOR box is bloddy expensive! I could purchase a PRO 96 or 796 bearcat for that price....

i'd consider purchasing a VHF Astro Spectra, if I could find one at a good price.

or trades

ASTROMODAT · Post by **ASTROMODAT** » Fri Oct 01, 2004 9:51 am

If your aim is to monitor P25, then I'd highly recommend going the PRO-96 route. I have A/B'ed P25 transmissions on a PRO-96 and XTS5000, and believe me, the Radio Shack sounded every bit as good as our brand new XTS5000's. That's $5,200+ versus $499, a full order of magnitude price differential. Even if you buy an eBay hacked Motorola for a grand or so, the RS is still way cheaper, and for monitoring, the RS is a lot easier and more convenient to use. If you want to listen to trunked systems, the RS is a much better way to go. The PRO-96 will display all sorts of channel info, such as alpha tagging, which your commercial radio will not. You also don't have the hassles associated with being affiliated, etc.

Larry

grinthock · Post by **grinthock** » Fri Oct 01, 2004 10:13 am

It's looking more and more like the way to go.... PRO-96

Now to move out some of my old stuff here to raise the fundage.

Then it's going to require a short drive to Niagra.

Was just hoping that there was maybe other options on the table.

I just like that I can drop my MTX-838 onto concrete and not have to worry (Although I never have)

Anyone know when the 2096 is due out?

ASTROMODAT · Post by **ASTROMODAT** » Fri Oct 01, 2004 11:18 am

GRE said January (on the Scanners Unlimited site).

mancow · Post by **mancow** » Fri Oct 01, 2004 11:21 am

I have to agree, the Pro-96 sounds awesome.

It sounds even better than my racal or johnson p25 portables. I've never used a moto p25 so I can't compare it to that but I don't doubt what the others say.

I actually think it's nicer than listening to analog.

mancow

ExKa|iBuR · Post by **ExKa|iBuR** » Sat Oct 02, 2004 6:04 am

Absolutely!

The audio on a Pro-96, with it's AGC and all that is *far* superior to any Motorola radio I've heard. It's definatly worth the $$ - I might even consider defecting from /\/\ and get one myself

-M

Post by **Wowbagger** » Mon Oct 04, 2004 5:58 am

515 wrote:A P25 signal is 4 level FSK, but I don't think a 4-level "data slicer" would be that useful for decoding it...

Yes and no. There are 3 parts to slicing a P25 waveform - determining the symbol time, determining the symbol amplitude, and slicing.

Every 1/4800 of a second, the signal from a properly calibrated P25 modulator will be at either +/- 600Hz or +/-1800 Hz from nominal carrier frequency - that time is called the "symbol time". Any other time, the signal isn't guaranteed to be anywhere special. So the first part of the puzzle is finding symbol time.

Then, you have to deal with the fact that the radio isn't perfect - it could have a carrier frequency error, it could have a gain error on the P25 waveform (e.g. +/-700 +/- 2100 instead of +/-600 +/-1800), it could have buggered filtering. So you then have to look at the symbol time data, and estimate the carrier frequency error and the gain error. Then, you "slice" the data - if it is above +1200, assume +1800. If +1200 to 0, assume 600, and so on.

If you are just decoding you are done with the analog part of things here - if you are a test equipment manufacturer like me you go on to compute how much different the symbols were from the ideal to get the modulation fidelity.

Then you have to do the deinterleaving, forward error correction, and deframing of the signal to get the signaling bits (DUIDs, low speed data, encryption sync data) and the voice bits.

Then you have to apply any decryption algorithms - DES, AES, or triple-DES - to the voice bits.

All that is pretty simple - you could do that with a modern computer, a sound card, and a good scanner - even for Phase II CQPSK ( the "C" is for "compatible" - the parameters of the QPSK modulation were chosen so that at symbol time the signal is identical to C4FM).

Now comes the tricky bit - you need to take the voice bits and expand them into real PCM data - and to do that you need the IMBE vocoder from DVSI. DVSI does not want to release the codec for general PC use (unless you wave LOTS of dead presidents at them). So you either need to buy a DVSI decoder chip, or reverse engineer the DVSI codec - not impossible for an experenced DSP person, but not trivial either.

So it all boils down to the DVSI vocoder - if you can get that then the rest isn't hard - our data slicer code is about 500 lines of pretty heavily commented C code, the deframer/FEC is a couple of thousand lines.