[R-390] Filters

Sat Feb 16 10:26:07 EST 2013

Hi

When I do them at work, the typical target is an FPGA. There are a variety of approaches. In this case I'd probably do a CIC decimator down to something rational, do the filter(s) as a FIR, and then a CIC interpolator back up to 455 KHz. Run the A/D's and D/A's with a sample frequency around 2 MHz or so. I would go with fairly high order CIC's and a filter or (more likely) cascaded set of filters in the 1 to 2 K taps region. The internal clock on the FPGA would be in the 50 to 200 MHz region so it didn't nuke the radio to bad.  Its probably a pretty solid day or two to work all that stuff out. 

You want the digital filter to be "better than" the existing filters in terms of shape factor, and ultimate attenuation. That takes you out of the cheap / dirty approach to some of this. You want at least 120 db (more like 135). The parts you use will need to be pretty good. The filters themselves would not be simple in terms of horsepower. You actually need *less* DSP horsepower in the gizmo to run it into a PC than to put it back out at 455 KHz. If you have a PC, you do all the real filtering there. You also don't have the CIC interpolator and RF D/A. IF DSP done *right* has it's own bucket of worms, they each need to be taken care of. That adds complexity. 

Lots of details to work out. How big a front end filter ahead of the A/D (probably a ceramic bandpass)? What sort of setup to use as a dithering source (it's an IF, you don't have enough noise to dither adequately) ? The board layout and enclosure would have to be *very* good to get the isolation. Same is true of the power supply. The clock source would need to be quite good to get the ultimate attenuation you are after. Since you don't have massive decimation going on, the A/D would be at least a 16 bit part (18 and 24 bit parts are out there). I'd bet on an 8 layer board, BGA construction on some of the parts, and three passes to get a final layout. Hopefully you can avoid 0.5 mm pitch BGA's.

Tooling the custom enclosure, doing the board runs would need to be funded somehow. By the time you are done, there's probably $20K in the "tooling" bucket. That assumes somebody will do the pick and place for free on the prototypes. Doing a production run probably involves fronting another $20 to 40K. 

More or less you have the same stuff as in a QS1R or something like it, plus the front end filter, plus the RF D/A. Weather it is $1,000 or $2,000 when you were done depends on how many thousand a year you can sell and how many years you plan to do it. Last time I saw scrap IF decks they were below $200. For that matter one can eBay a whole radio for $500 and part it out. 

As pointed out earlier by Charles, for the same money you can just tack a SDR + PC on the radio and have a *lot* more features. Features creep is a very real issue. The SDR + R-390 combo won't quite work right, since you don't have much dithering on the A/D, but it will work. 

The whole point of "it won't quite work right" is the real issue. The R-390(A or not) is a pure analog radio. What ever issues it has are analog issues. Digital radios have their own set of issues, quite independent of anything analog. As soon as you drop DSP into the radio, you bring all the digital "stuff" along with it. You then have both the analog limitations plus the digital ones. If you want to deal with digital issues, just spend the same money on a full digital radio. That way you don't have to deal with the analog issues as well.

Bob

On Feb 16, 2013, at 9:14 AM, John Saxon <johnbsaxon at yahoo.com> wrote:

> I am not as "savvy" as you folks when it comes to the electronics, but I am a "savvy" software guy who has done a lot of low-level, embedded stuff (including micro-coding on array processors for you old-timers out there).  I have a couple of home projects using microcontrollers and I enjoy playing with them.
> 
> Anyhow, you folks have my curiosity up about the software requirements to do a DSP filter suitable for Collins mechanical filter replacement.  Can some of you gurus point me to where the math algorithms are defined?  Gotta be on the web someplace :-)
> 
> FWIW I currently have a 390 and a 390A.
> 
> Thanks & 73,
> John
> K5ENQ
> Pearland, TX
> 
> 
> --- On Fri, 2/15/13, mlmccauley at att.net <mlmccauley at att.net> wrote:
> 
>> From: mlmccauley at att.net <mlmccauley at att.net>
>> Subject: Re: [R-390] Filters
>> To: R-390 at mailman.qth.net
>> Date: Friday, February 15, 2013, 8:23 PM
>> 
>> Absolutely.
>> 
>> I've seen those come and go. The concept is great, but the
>> market is very limited, hence the cost.
>> 
>> A savvy ham with the proper background could homebrew a very
>> nice filter for a relatively cheap price. Soldering the
>> hardware together would be a fairly easy weekend project,
>> assuming you had a surface mount adapter board for the DSP
>> chip. The software would take many weeks to write and even
>> longer to debug.
>> 
>> 
>> 
>> 
>> On 2/15/2013 8:05 PM, Bob Camp wrote:
>>> Hi
>>> 
>>> Thus the variety of outboard dsp filter boxes that have
>> shown up over the years. Once you get them into production,
>> the cost is high enough that there is a very limited
>> audience.
>>> 
>>> Bob
>>> 
>> 
>> ______________________________________________________________
>> R-390 mailing list
>> Home: http://mailman.qth.net/mailman/listinfo/r-390
>> Help: http://mailman.qth.net/mmfaq.htm
>> Post: mailto:R-390 at mailman.qth.net
>> 
>> This list hosted by: http://www.qsl.net
>> Please help support this email list: http://www.qsl.net/donate.html
>> 
> ______________________________________________________________
> R-390 mailing list
> Home: http://mailman.qth.net/mailman/listinfo/r-390
> Help: http://mailman.qth.net/mmfaq.htm
> Post: mailto:R-390 at mailman.qth.net
> 
> This list hosted by: http://www.qsl.net
> Please help support this email list: http://www.qsl.net/donate.html