If you are someone whose interests lie in the field of RF, you won’t need telling about the endless field of new possibilities opened up by the advent of affordable software defined radio technology. If you are a designer or constructor it might be tempting to believe that these radios could reduce some of the problems facing an RF design engineer. After all, that tricky signal processing work has been moved into code, so the RF engineer’s only remaining job should be to fill the not-so-huge gap between antenna and ADC or DAC.
In some cases this is true. If you are designing an SDR front end for a relatively narrow band of frequencies, perhaps a single frequency allocation such as an amateur band, the challenges are largely the same as those you’d find in the front end of a traditional radio. The simplest SDRs are thus well within the abilities of a home constructor, for example converting a below-100kHz-wide segment of radio spectrum to the below-100kHz baseband audio bandwidth of a decent quality computer sound card which serves as both ADC and DAC. You will only need to design one set of not-very-wide filters, and the integrated circuits you’ll use will not be particularly exotic.
But what happens if the SDR you are designing is not a simple narrow-band device? [Chris Testa, KD2BMH] delivered a talk at this year’s Dayton Hamvention looking at some of the mistakes he made and pitfalls he encountered over the last few years of work on his 50MHz to 1GHz-bandwidth Whitebox handheld SDR project. It’s not a FoTW in the traditional sense in that it is not a single ignominious fail, instead it is a candid and fascinating examination of so many of the wrong turnings a would-be RF engineer can make.
The video of his talk can be found below the break, courtesy of Ham Radio Now. [Chris]’s talk is part of a longer presentation after [Bruce Perens, K6BP] who some of you may recognise from his activities when he’s not talking about digital voice and SDRs. We’re jumping in at about the 34 minute mark to catch [Chris], but [Bruce]’s talk is almost worth an article in itself..
If SDRs are a foreign language to you, perhaps you should read our Hackaday Dictionary piece explaining the technology, and if you are interested in RF design you should also take in HackRF creator [Michael Ossmann]’s talk on the subject.
Thanks [Jacek] for the tip.
11 thoughts on “Fail Of The Week: The Pitfalls Of Designing A Wideband Radio”
RF has a distinct tendency to humble the overly confident. Which is part of the fun. I think “there be dragons in this cave” would be a better description than fail.
I was once an adventurer like you…then I took an arrow in the knee.
Talking about spurious signals in receivers reminds me of a satellite beacon receiver design I worked on about 25 years ago. The system had a 30GHz input, and the final IF processing was done at 455kHz. The part I helped to design was the 70MHz 2nd IF to 455kHz 3rd IF tracking downconverter. During the proof-of-concept prototyping phase (with unshielded boards all bolted to a big sheet of Al and RG178 running everywhere), it was noticed that it would lock onto and track a strong spurious signal that was a bit mystifying.
After some days of testing all manner of combinations of oscillators in the RF chain, we found that it was actually coming from the microcontroller.
The 8051-deriviative controller had an 11.0592MHz crystal (it needed that exact frequency for the baud rate). I investigated the problem by connecting a frequency synthesiser to the 8051 XTAL pins. When I stepped the XTAL frequency by 1kHz, the spurious signal moved 6.33kHz. This confused me for a while, as I first thought this was actually exactly 6kHz plus some measurement error. (So much for confirmation bias.)
The 8051s of that vintage divided their clock by three. The nineteenth harmonic of one third of 11.0592MHz fell inband at 70MHz (at exactly 70.0414MHz).
The problem went away once the various modules were assembled into their shielded enclosures.
Hmmm. Memory fails me. There must have been yet another IF stage between 70MHz and 455kHz, perhaps at a few MHz. This would have been needed to achieve the image rejection and spurious goals, given the wide tracking range.
I had a cup of tea and it came back to me. The extra IF was at 10.7MHz. This frequency was chosen to allow the use of common IF filters made for FM radios. Similarly, 455kHz was chosen for the final IF to allow the use of common, cheap IF filters made for AM radios.
10.7MHz was also low enough in frequency to allow the use of CMOS switches in the 10.7MHz to 455kHz quadrature downconverter.
The first LO was a VCXO. The second LO was a DDS (which was actually a pretty bold choice for something of that vintage).
When hunting for spurious signals in a receiver design, there are multiple debug approaches one can use:
1. “The Empiricist”
Perturb the frequency of each of the oscillators in turn, and see what that does to the frequency of the spurious signal.
2. “The Analyst”
Make a big table of all the interactions of all the oscillators. Consider the mixing products of each pair of oscillators (e.g. F1 + F2 and F1 – F2). Also consider the mixing products of all the harmonics of each of the pairs of oscillators (e.g. nF1 + mF2 and nF1 – mF2, for all integers n and m).
My design used a DDS, so I would also have to consider images of harmonics and harmonics of images.
For anything but a very small number of oscillators, this turns out to be a huge task.
I tried doing this for my design, but came unstuck because I hadn’t considered the divide by 3 stage inside the 8051.
Nice problem solving. Thats my nightmare…say its a complicated rf soc and due to all these peripherals so close to one another some funky noise gets generated for some weird edge case. Ive got a problem where mcu isnt pulling a line for amount of time im programing and its not long enough to kick on the separate rf chip…im not sure how to tackle that now (if its software or hardware) but have to go at it hard next week.. :/
But this isn’t really a surprise, since many people don’t know good receiver design.
The Racal receiver with the Wadley loop, I think the RA-17, it had a preselector, but also just a lowpass filter. Since it unconverted, image rejection wasn’t really an issue. But I wonder if people really just used the low pass filter in reality.
The mixer would have to be good, and since the Wadley loop adds an extra mixer for the same of eliminating all those crystals, the mixers would have to be especially good.
People didn’t really start talking about balanced mixers till 1963 with that article in QST by Squires. That caused a flurry of interest in balanced mixers, but except for the Squires-Sanders SS-1R, balanced mixers only became common place in non-professional receivers when solid state took over sometime in the seventies.
A lot of construction articles for receivers never addressed good signal handling.
When direct conversion receivers started becoming popular in amateur radio starting in 1968, there were endless articles bout better mixers and the like, but they seemed incremental improvement. Then suddenly someone had a tiny QRO rig that used a balanced diode mixer, but it was properly terminated. The topic had been covered, but in terms of VHF diode mixers, but terminating direct conversion mixers really seemed to help.
“Software radios” aren’t magic, at the least they are only as good as the AD converter.
Every new advance has often been accompanied by “we can throw out this…”. Surely good front end filters can be useful in a digital age, the SDRs are more than just doing away with such filters.
I use old analouge radio front ends, driving RTL SDR sticks, filtering makes a BIG BIG improvement
There is nothing “Magic” about SDR design in the end. The Digital part of SDR can add wanted performance in the end-result, but there is always a trade-off in the requirements of the analog portion of the design. As anyone who is a “properly” educated Electrical Engineer (and/or a specifically-focused Physicist graduate) knows: Information Theory is directly tied to the laws of Thermodynamics, especially the Second Law.
Simply Put: “You Can’t Get Something for Nothing”
As a Classically-Trained EE, the problem I see today is that Universities are churning out so-called “Engineers” as fast as possible (4-years or less) without the Fundamentals in a broad range of disciplines needed to give them the basic knowledge needed to cope with the evolution of technology, understand it, and adapt. The new EE’s are more like “Technicians” (or even “Makers”) – in my opinion only.
Since you opened the can of worms…
As an EE, what I see coming out of colleges today are young EE’s that don’t understand hardware when its placed in front of them. They get a lot of computer simulation time, but since budgets are being cut and labs are not being supplied with equipment and parts, Pspice is all they seem to understand about the workings of a circuit, and that does not even come close to covering all of the variables that can effect a circuit, especially RF and analog.
This applies directly to RF design, a.k.a. “The Black Art” of the electronics world. RF demands of the designer that he/she understand the subtleties of not only how components work under the hood but how they interact in very unusual ways. Parasitic effects abound and are not included in simulations as some of the most interesting only show up in a physical layout. Experience cannot be substituted for with simulation. And I’m speaking about RF here, the same tends to apply to design in general. Simulation can cover basic circuit design, Murphy covers the real world.
There once was a time when young engineers had experience as hams or hobbyists. This taught them practical circuit operation. I’m hoping that the Maker Movement brings that opportunity back, Engineering as a profession can only benefit from it.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)