rrTV-PHOTO   New HD TV
HOME   rrTV-PHOTO   GALLERIES   MY GALLERY   HELP-FAQ
myHOME PM pmRR MEMBERS 347 ONLINE 38 EVENTS SEARCH REGISTER  START HERE
 
12 pages [ <<    <     9      10     ( 11 )     12     NEXT    >> ]18353 viewsPOST REPLY
A Main Hobbies . Boca Bearings . Modefo's RC Helicopters

.
.
Radio - Servo - Gyro - Gov - Batt > Homebrew PCM Receiver: QPSK/RF Design
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, Folks!

Sorry for the delay - I've been completely lost in filter design! Given the performance targets that we're hoping to attain, I'm having to re-learn everything, including the true nature of basic capacitors and inductors. I've known all along about 'Q', which is the figure of merit that quantitatively describes how "pure" capacitors or inductors are (that is, how much resistance and inductance a capacitor has, or how much resistance and capacitance an inductor has; S-parameters are a more accurate method of conveying this). However, I'd never really realized just how "bad" parts really were, especially considering the desire in this application for extremely tight filter "skirts" (that is, the ratio between the frequencies that the filter passes, and the frequencies that it absorbs). It's hard to build these kinds of filters! As a consequence, I've also been out searching all over the place for folks that offer better components.

One great find was something that I hadn't seen before. Vishay offers "silicon capacitors" that deliver dramatically better accuracy and performance than any of the alternatives that I've seen, all the way down to a super-tiny 0201 package (that's just 1mm by 2mm)! Silicon ("chip") inductors have been around for years, but their Qs are quite poor at the R/C frequencies. The Vishay caps are an entirely different story - they have extremely low resistance and inductance - so their Qs are far higher than anything else I've seen. Great stuff! If I could just find inductors that are as nice, I'd be in business, but I suspect that they don't exist. So, the battle of the filter design will continue.

Now that I'm deep into this, it's delayed the start of the VCO test board for a while. However, the 72-73 MHz bandpass filter, the baseband audio/low RF filter, and the software SRRC filter will essentially be the primary determinants of the receiver's performance, so it's worth a bit of time right now to focus on it. I'll get back to the synthesizer as soon as I can reach a decent breaking point.

One thing's for sure: I'm gaining an even deeper appreciation for those of you who design RF for a living! Well done, guys!!!

Cheers!
MarkF
10-15-2004 Over year old.
HOMEPAGE  
 
 
MarkFSenior Heliman - Location: Palo Alto, CA -
Howdy!

For folks who aren't into filter design, I thought it would be worth showing you how huge the effects of Q can be. Here's the response of an "ideal" 6-pole, passive Chebyshev I bandpass filter for the 72-73 MHz band (an ideal filter assumes that the components are theoretically perfect):



Now, watch what happens when we try to build this filter out of real components (I've chosen chip-type inductors in this case just to illuminate the problem a bit more):



Not only does the filter response curve change dramatically, note that the signal level has been attenuated by nearly 90 dB - in other words, the signal's gone! Now, before folks jump all over me about trying to create a completely irrational filter, I'll emphasize that I chose this example specifically to show how bad the problem can get. The filters that I'm pursuing are quite different than this reductio ad absurdum example. Nonetheless, that's precisely why RF has traditionally been thought of as such an art. As Yoda might say, "Very peculiar, these devices are!"

Have Fun!
MarkF
10-15-2004 Over year old.
HOMEPAGE  
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

A Fascinating Article!

Hi, Gang!

Speaking of how complex RF design can be, there's always enough balls in the air that it's pretty easy to make stupid mistakes. There's a really interesting article in the October, 2004 issue of the IEEE's Spectrum magazine. Called "Huygens Hero", it details how one engineer fought heroically to discover, isolate, and correct a crucial flaw in the Cassini spacecraft that would have resulted in the complete loss of data from the Huygens satellite, totally trashing the Titan mission.

The radio link between the Huygens probe and the Cassini spacecraft is based on 8K bps BPSK (Binary Phase-Shift Keying) - essentially equivalent to sending just 1/2 of a QPSK signal. The problem was that the company in Italy that created the BPSK receiver onboard Cassini had completely missed the fact that the frequency change induced by doppler shift would affect the symbol timing recovery logic, and NASA missed it since they skipped a full link test for cost-cutting reasons. Even from my perspective, those were both astonishingly dumb mistakes, but they did happen. Worse yet, the firmware couldn't be updated after launch. Tough problem! I won't spoil the article by mentioning the solution, but it was a clever alternative.

If you have access to Spectrum, it's a very worthwhile read!

Enjoy!
MarkF
10-15-2004 Over year old.
HOMEPAGE  
 
 
MarkFSenior Heliman - Location: Palo Alto, CA -
Hi, Gang!

I finally came up with what might be a decent filter design after more than a week of seemingly endless variations (I've probably been through more than a thousand combinations [you know... monkeys at typewriters, and all that]). Its a fourth-order passive Chebyshev I resonator filter with end coupling. Not counting the 50 ohm source termination resistor, its insertion loss is about 8.4 dB - with it, it's down an additional 6 dB. I realize that this is kind of high, but this is also a high-attenuation filter. Here's the frequency response:



As you can see, the response is down ~54 dB at a 10% frequency offset. To get this far, I worked really hard to get the right component ranges so that I could take advantage of Vishay's silicon capacitors. I've extrapolated Q figures as best I could from the Vishay and CoilCraft data sheets. The four "funny" inductor values are
CoilCraft's 10mm shielded tunable inductors.



This seems like it might be a good start, but for one or two concerns. First, I'm a little worried about the input/output impedance of the filter. As you can see in this zoomed-in view from 72 MHz to 73 MHz, the real impedance varies from ~33-51 ohms across the band, and I don't understand complex impedance enough yet to know if what's shown here is OK or not. Obviously, I'd prefer it to be flat, but anything I do to improve the impedance trashes the frequency response. Is this normal, or maybe even acceptable anyway?



I'll have to be very careful with the layout on a filter like this, and even so, it'll obviously be somewhat worse than shown here. Still, as far as I can tell, it's free from self-resonant frequency problems, has only 44 nS pk-pk group delay variation across the 72-73 MHz band (which is insignificant on any individual R/C channel), and I think it might actually be buildable, if I can get away with that impedance change.

I'd greatly appreciate your feedback on whether you think that this might work!

Thanks a bunch!
MarkF
10-16-2004 Over year old.
HOMEPAGE  
 
 
Phil Cole
Veteran
Location: Redwood City CA

Mark,

You do plan to put an isolation amp in front, don'
t you? It it's source impedance is low enough, the load of the filter shouldn't mess things up too much.

To get the source imedance low, you would introduce a bit of negative voltage feedback.
10-17-2004 Over year old.
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, Phil!

As you had recommended, absolutely! [Phil had explained to me a while back that the antenna's impedance was essentially random due to the uniqueness of each heli's wiring, and recommended the amp to provide a constant impedance source]. In addition to that, I've been wanting to add ~15 dB of real gain prior to the QuickSwitches, anyway, to accomodate their switching noise, so we'll target an input gain of perhaps 25 dB.

The Sirenza amp that's used in the latest rev of the SDR-1000 is a possible candidate, and it's got plenty of headroom to add the negative feedback that you suggest - that's a great idea! As you'd also recommended before, one of these days, I'm going to have to pick up (and learn!) a copy of Spice.

I really appreciate your advice! Creating that filter was a bear - I even had to change most of the components by more than 5X from the nominal values to yield that filter. I really wasn't looking forward to having to do that all over again! [Update: Fortunately, this filter architecture seems quite scalable, rather than having a "hair-trigger" sensitivity; I was able to port it to the 6M band in about 15 minutes. That's a good indication that it really is buildable (or that I've managed to tweak things enough to confuse Filter Solutions' integrated simulator, but the evidence that I've seen so far suggests otherwise)].

By the way, I just realized one very significant benefit of moving as much filtering as possible to the front-end. Whatever filtering exists before the QuickSwitches only needs to be built once, while it'll be necessary to build two copies of everything that comes after the QuickSwitches: one for I, and one for Q... Doh!

I'm also rapidly coming to the conclusion that the 2.6 mS of delay for the baseband SRRC filter may be hard-coded into the problem. There is a possible way of attacking the problem that involves decimating the signal by a factor of two, which would reduce the sample rate to 48 KHz prior to the SRRC filter. To take advantage of this, though, we'd need a different filter design program which understands that a complex signal isn't limited to one-half of the sampling rate, but can instead handle a bandwidth that's equal to the sampling rate. Unfortunately, none of the programs that I've seen can do this. I'll keep my eyes open, just in case one of them can handle the problem. Meanwhile, I think I'm stuck with the 2.6 mS delay for the SRRC filter. Bummer!

Thanks again, Phil!!!

Best Wishes!
MarkF
10-17-2004 Over year old.
HOMEPAGE  
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, Gang!

OK - enough's enough. After creating half a dozen symbols, packages, and devices in the Eagle schematic capture/PCB design package, I began to get quite frustrated. While I've used Eagle for a couple of small projects in the past, I really don't like this package. Every time I sit down in it, I feel like I'm wasting a ton of time on operations that should be far easier. So, I gave up and went hunting.

After checking out four different CAD packages, I've found my new home. If you're looking for a much better CAD tool, please check out WinQCad for what looks to be a far better program. It provides a much more straightforward design flow, is very fast, and contains an excellent autorouter that blows away anything else I've seen except the top of the line Specctra autorouter. In comparison, the autorouter in Eagle is complete junk. The amazing thing is the price for the entire design suite: $150 for 500 pins, $300 for 1,000 pins, or $600 for unlimited pins - all of which support any number of PCB layers. The unlimited version is cheaper than the limited "mid-range" version of Eagle, but more capable than the high-end version of Eagle. As one rather major example, Eagle supports only normal flat schematics, while WinQCad supports full heirarchical schematics, with fast heirarchy navigation.

One very interesting feature is that the autorouter is actually configured to be Specctra compatible, so that you could even buy WinQCad just for its autorouter, while continuing to use your existing schematic/PCB design program. Not me, though. This is a well thought out design suite that's powerful, easy to use, and just plain more efficient. On top of that, the included libraries are far better organized than Eagle, but WinQCad can also use Orcad and Protel libraries, too!

I'm very impressed with WinQCad. Compared to the other junk that I've seen, this is a better, more polished, less buggy, and more refined alternative. It's impressed me enough that I've just bought it, and it's what I'll be using for this receiver. I realize that the selection of a CAD package for home use is a personal decision, but before buying another CAD package, please take the time to give WinQCad a shot!!!

Cheers!
MarkF
10-20-2004 Over year old.
HOMEPAGE  
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, Greg!

May I ask you a question? I blew through Ansoft Designer SV pretty quickly, preferring the much simpler to use Filter Solutions. However, I decided to give Ansoft another try, and it truly is in a different league. It's a much richer, more complex application, that not surprisingly has a steeper learning curve. What caught my eye in particular was that it has a broader array of filter types, including the specific resonator architecture I'd created in Filter Solutions. However, its filter designs don't require that 50 ohm series termination resistor. It would certainly be nice to gain back that 6 dB loss!

I don't have any problem getting Ansoft to spit out the base filter design, and the loss characteristics do look better. However, what's the best way of "tweaking" a filter design to use real-valued components in Ansoft? With Filter Solutions, it was really nice: you changed a component, and it would automatically calculate the matching L or C value and update it - very nice! With Ansoft, though, changing one part just trashes the frequency response, then you have to change the matching L/C manually.

Since you've used Designer, I was wondering if you could suggest the easiest way to tweak filters?

Thank you very much for any guidance that you're willing to pass on!

Cheers!
MarkF
10-22-2004 Over year old.
HOMEPAGE  
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, Gang!

Now that was a pleasant surprise! I've been wondering how things were going at Premier Devices on the VCO, expecting an answer to the feasibility of the 6 meter (200 MHz) part any time now. Instead of getting an answer, the samples just showed up today from China! Through this process, I've also learned that Premier has much better support than I'd thought. My misunderstanding was due to Google!

It turns out that if you google for "premier VCO", you'll wind up going in to their site through a means that shows only the Shanghai support link. Going in through the main http://www.premierdevices.com site shows that they in fact have international support, including here in the states. I had a chance to talk with them, and I'm impressed!

So, what showed up is a completely new part, with its own part number: the "V-0210-05"! Together with the V-0305-05 part for the 72 and 75 MHz bands, we're covered. Even better, though I offered to pay NRE and buy the samples, everything so far's been free from Premier! Kudos to Premier Devices for terrific service, indeed!

Have Fun!
MarkF
10-23-2004 Over year old.
HOMEPAGE  
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, Greg!

By the way, I just realized that your concerns about vibration will become more important. The variable inductors that I used turned out not to be surface mount parts after all (the Q for SMT parts is just too low at these frequencies). Consequently, we'll have to do a careful test to see how vibration effects change the receiver's response, just as you predicted.

Barring any other feedback, we'll proceed with the initial filter that I came up with in Filter Solutions, while continuing to search for a better solution in Ansoft as a background activity. Hopefully, the variable input impedance of the Filter Solutions filter won't hurt performance too much, thanks to Phil's recommendation on lowering the output impedance of the buffer amplifier / RF preamp. The real truth will be found once we actually build a prototype...

Cheers!
MarkF
10-25-2004 Over year old.
HOMEPAGE  
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, Gang!

Sorry for the long delay - I've been wrapped up in a mobile development conference, and quite a few very long days at work. As a result, I'm afraid that I haven't made any significant progress since my last update, but I have discovered something interesting!

Ever since TMoore started dropping hints about the new Futaba 14 MZ next generation R/C system, I've been extremely curious to see whether Futaba might have finally begun to push the RF performance envelope. Well, despite the significant improvement in the $2,200 14MZ, it would initially appear that they are still being quite conservative. As the link above notes, the frame rate in the 14 channel 14MZ is 17.5 mS - more than twice as slow as the frame rate that we're targeting here for a 16 channel system!

Now, I will note that the 14MZ offers 2048 point PCM resolution, instead of 1024 point resolution. If we wanted to offer 2048 point resolution in this setup, our frame format would have to change. The easiest way to do this would be to add one additional bit to each word, but that then screws up the differential channels - we'd instead need to add two bits for each of the packed differential control words, which would then yield PCM4096. Let's say we went that route. Adding two bits to each of the 9 control words would increase the frame length from 94 to 112 bits. With the two additional bits in the control word, we'd also pick up two more switch channels, for a total of 10 switch channels + 8 analog channels = 18 total channels of PCM4096. In this case, the frame rate would be ~9.4 mS, which is still 46% faster than the 14 MZ. So, it would again appear that Futaba's not pushing the envelope at all.

There is another possibility. To give Futaba the full benefit of the doubt, let's imagine that they've completely eliminated differential encoding in the 14MZ. To create this type of system, the transmitter would just send 14 full analog channels every frame. At the bit rate used here, we'd wind up with 17x11 bits, which would yield a frame rate of 15.6 mS - in other words, pretty close to the 14MZ's 17.5 mS frame rate. If Futaba has actually taken this approach, then my hat's off to them, for that would indicate that they really have been fairly aggressive. If instead they're still using a differential frame format that's similar to the 9Z, then they're being just as staunchly conservative as they've been all along.

What's the truth? While history would suggest that they're being conservative, there's currently no way to tell. Once it ships, I'll have to see if I can get my hands on one to find out what's really going on with the "revolutionary" 14MZ.

Cheers!
MarkF
10-30-2004 Over year old.
HOMEPAGE  
 
 
Bezjak
Heliman
Location: Boca Raton, FL

Mark,

I've been travlelling the last 10 days and haven't checked into your thread.

With regard to tweaking filters, you are already doing what I do. Change a part to a standard value and see what the reponse looks like. Then change the next part....

For all of the theory behind filter design, I have not seen a package that takes component Q into account from the start.

This process is really what I use Designer for the most -- Seeing how filters will perform in the real world using practical components.

Greg
10-31-2004 Over year old.
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

HI, Greg!

Welcome back! I hope that it was a relaxing/enjoyable vacation, or a profitable/successful business trip!

Thank you very much for the feedback - I kept banging around in Ansoft trying to find the "magic button", and it apparently doesn't exist! I must admit that Filter Solutions' approach is awfully convenient in this regard.

This weekend, I had the chance to spend a bit more time in WinQCad, and it's proving to be a good choice. Like all CAD packages, it has its distinct quirks that take adjustment, but it'll definitely do what I need it to do. Good stuff! Thus far, I've mainly just been playing around to learn the package better; next steps will be to start creating the VCO test PCB layout. As you said, it'll be fun to finally start building something!

On other topics, even if the Futaba 14MZ has dropped differential decoding, we're going to stick with it here - it'll be neat to deliver nearly twice the frame rate of the latest, greatest "super system"! I'll have to decide whether we'll go with PCM4096; that would mean a worst-case delay between servo channel outputs of just 250 nS! In the interest of trying to stay on the leading edge, though, I'll probably give this a shot. The funny thing is that I'll probably never be a good enough flyer to notice the difference from either the faster frame rate or the increased positioning accuracy, but I'm really looking forward to dropping this into the hands of an expert, and finding out what they think of it! Think Alan or Danny Szabo here...

Thanks again for the help!

Cheers!
MarkF
11-01-2004 Over year old.
HOMEPAGE  
 
 
TurboStew
Senior Heliman
Location: Ft Collins, CO

MarkF,

I am not sure of your commercial intentions with this, but being a DSP engineer, I believe we need an open source programmable TX out there. I was thinking of designing a retrofit board for some cheap, readily available TX (hitec?). Your proposed reciever (and transmitter module) would be an excellent addition to such a cause. Any thoughts?

8.5 pound 90 birds Rock!
11-03-2004 Over year old.
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, TurboStew!

As for my commercial intentions on this project, I have none! In fact, I've already published the source code to the first-generation FSK version of this receiver here on RunRyder.com. It isn't a "true" open source license, but I think the terms are pretty reasonable. I basically said that you can only use it if you won't sue me, and that if you do something commercial with it, that you send me something of value that's proportional to the value that's received, purely on the honor system. For example, I suggested that if you're a major R/C manufacturer that bases your entire product line on my code, you might want to send me a nice new digital oscilloscope?

I will definitely be freely publishing the software and hardware for anything I develop on this project, likely under the same terms. While I didn't set out to build a transmitter, it became necessary that I build one as soon as I decided to develop the QPSK version of the receiver. Consequently, I'll need to develop a transmitter that's also optimized for extremely low latency, and that will transmit QPSK.

Your background as a DSP programmer is very encouraging to me; the majority of programmers simply don't understand how much work is necessary to achieve truly low latencies. For me, latency is paramount, so the code I published for the FSK receiver uses some quite sophisticated techniques, up to and including a real-time on-the-fly machine language compiler, to achieve the target latencies. DSP folks are a different breed, who are much more familiar with this type of challenge. The bottom line is that I'd love to have you join the project!

Why don't you say a bit more about your goals for the transmitter? The best way to know if we're marching in the same direction is by starting to describe where we hope to wind up! Thanks a bunch for your interest in building an R/C system that will really kick ass!!!

Cheers!
MarkF

P.S. While it isn't a big deal, I've been assuming that I'd be developing a replacement transmitter board for the Airtronics RD8000, since it has very nice sticks. I've got no problem with Hitec at all; if that's a good solid base, let's make it happen!
11-03-2004 Over year old.
HOMEPAGE  
 
 
TurboStew
Senior Heliman
Location: Ft Collins, CO

Yea, I know what you mean, real-time doesn't mean "as fast as you can enter the data and watch it on the screen". My background is in closed-loop control systems, the present being brushless DC servos, 20K fixed-point servo loops are my gig. Anyway, after tossing around the idea for a few years now, one of the things I want in a radio is programmability by assigning equations to variables. I posted this the other day:

http://www.runryder.com/helicopter/t138988p1/

Of course, this is only part of the many pieces that would be cool. The cool thing about open source is we may have several software versions out there with interfaces from simple to complex. I, being a geekineer, want utmost in flexibility with little emphasis on simplicity.
.

I am a 3d pilot attempting more on-the-deck stuff. I am starting to be bugged by my ECCPM interactions. I want less latency! I want linearization (not just static, but dynamic, using servo speed)! I want force feedback! Anything I can help you with at the present?

8.5 pound 90 birds Rock!
11-03-2004 Over year old.
 
 
MarkF
Senior Heliman
Location: Palo Alto, CA

Hi, TurboStew!

Excellent - I like your concepts, and think that they'd be a good match to my own goals. I will be leaning slightly towards a friendlier transmitter interface, since I'd like the system to be perceived by the R/C manufacturers as a legitimate commercial candidate. However, having a dual-mode "expert" UI would be very cool! I completely agree that the advanced capabilities should be the top priority, and I'm more than willing to accept the challenge of trying to wrap a friendly interface around them later on.

As far as direct help goes now, if you're a hardware designer, I could sure use the assistance! As you can probably tell from this thread, firmware's easy for me, but I'm currently buried in the process of learning RF hardware design. If you're like me, and lean towards firmware, there are still a couple of areas where you could get started pretty quickly. The first thing you might want to check out would be Analog Devices' terrific little BF-531 Blackfin DSP. I'm quite impressed with the combination of fast performance, low cost, and low power consumption that it offers. If you agree, I think that it'll be a great core for both the transmitter and the receiver. Check it out! When you do, keep in mind that even though Analog Devices calls it a "16 bit DSP", the chip has extensive 32-bit math capability that'll come in quite handy for the sophisticated mixing that you've got in mind. After you've checked out the Blackfin, please let me know what you think!

The other thing that might help is to take a look at the source code for the initial FSK version of the receiver. While it was created for the ARM7 CPU, it'll give you an idea of my design approach, code style, and how I've rather heavily commented it so that others could pick up the code for their own use. Do note that the QPSK version of the receiver will be quite different from the FSK code, since creating a high-performance QPSK receiver is far more complex than receiving simple binary-sampled FSK - not to mention that the entire event compiler infrastructure will have to be ported to the Blackfin. Still, I think this will be a good general introduction to the project. Once again, I'm very interested in your feedback after you check this out!

Meanwhile, I'm slowly trying to slog my way through creating the first element of the receiver's hardware design, a local oscillator test board. I can't wait until I can get back to coding again!!!

Thanks again for your interest in this project!

Cheers!
MarkF
11-03-2004 Over year old.
HOMEPAGE  
 
 
KCT
Veteran
Location: Ontario, Canada

Hi Mark,

any news or progress on your cool TX/RX project ?

Cheers

Kay


That's not flying, that's crashing with style...

TeamBob00, WildCat, Helitron
08-23-2005 Over year old.
HOMEPAGE  
 
 
w.pasman
Elite Veteran
Location: Netherlands

Yes I'm also curious! Is your project still alive? Please update us
08-24-2005 Over year old.
HOMEPAGE  
 
 
RC-CAM
Senior Heliman
Location: USA

Me too.
08-26-2005 Over year old.
HOMEPAGE  
 
 
12 pages [ <<    <     9      10     ( 11 )     12     NEXT    >> ]18353 viewsPOST REPLY
XHELI.COM . Autography FlightPower . Advantage Hobby

.
.
Radio - Servo - Gyro - Gov - Batt > Homebrew PCM Receiver: QPSK/RF Design
  UPDATE SCREEN   PRINT TOPIC Advertisers 

Subscribe to This Topic

Monday, September 8 - 7:40 am - Copyright © 2000 - 2008 runryder.com | email | link to rr | runryder needs cookie