For a radio hobbyist, a sweep generator and spectrum analyzer is a highly desirable tool. There are plenty of projects on the web to attest to that... but the ones I've seen are pretty old-school (double-conversion superhets with massive filters).
Using an Si5351 to generate a sweep from 1-200MHz, a good mixer (say an HFA3101 Gilbert cell, maybe in cascode to reduce IMD as discussed previously), a DC-coupled audio LPF (300Hz, 3KHz) (probably with a 100KHz pre-filter), and an AD8307 log amp, with an Arduino to control it and digitize the result, you could build a very simple direct-conversion spectrum analyzer for $20 in parts.
I know that the LO frequency would convert to DC, so the detector would need to be DC coupled, and that might introduce stability and temp-co problems. But by using even a small sweep, it wouldn't be necessary to preserve absolute DC accuracy - you'd just lose a little frequency resolution instead.
The Si5351 can do good quadrature between 4MHz and
112MHz, so you could use two mixers to get I/Q demodulation, followed by sound-card or Arduino digitization and PC-based SDR. That would get rid of the DC problem, and you could also receive all modes.
Why Si rather than AD? Analog Devices do a number of DDS chips with quadrature outputs. My guess is that anybody else is going to be cheaper than AD but it's a design choice that ought to be at least mentioned.
As you say, by doing I/Q you can preserve phase information.
The filters are all simple lowpasses, and so can be as sharp as you like.
You won't have to deal with the forest of spurs that double-conversion generates if you aren't careful.
On the minus side:
Your overload performance lives and dies by the strength of the mixer, because there are no filters to help. Plus all the high order intermod and cross-mod lands right on top of you, and can't be filtered out.
Your LO is at the same frequency as your RF, so small signals disappear into the LO feedthrough/noise/offset. DDSes have odd low-frequency spurs that will come straight through. You'll also have to filter the LO very well to avoid the quantization causing edge jitter in your hard-switched mixer.
To net it out, I'd say that it's a great idea for light duty use, but spill some sweat on the mixer design and LO filter.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
I once demo'd a Tektronix spectrum analyzer plugin. It had zero image rejection. The manual said that if you turn the frequency dial and a line moves in the wrong direction, ignore it.
--
John Larkin Highland Technology, Inc
picosecond timing precision measurement
jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com
You could just AC couple the output, and get used to the double-hump. Or use an image-reject mixer and get used to the 40db-down ghost. Or use the Weaver method and get used to the tone.
--
Tim Wescott
Wescott Design Services
http://www.wescottdesign.com
I'm looking for work -- see my website!
...if you care what is just above or just below your carrier. A spec-an doesn't really have to though.
That's what I thought - hence the need for a very good mixer. It would be nice if your favorite BPF640 had a Vceo more than
4V for a shade more headroom, but I guess that goes with the territory.
Hmmm, that is a worry. Sounds like a good reason *not* to DC couple, but to embrace the double hump of a 50-300Hz BPF, and handle it in software. That seems possible.
The Si5351 isn't a DDS. It has a 600-900MHz VCO and dividers. It's claimed not to have those close-in spurs.
When you want quadrature, it runs two divider chains from the same VCO (the Si5351C has eight VCOs and divider chains, but I'm looking at the B which has 2). I don't think it's needed for a spec-an though - unless you can use it to eliminate the double-hump.
Are you referring to DDS quantization? Shouldn't be a problem.
Thanks, I might have a crack at it (that's Aussie for "try it" :)
With the most practical option of AC coupling the IF port (or high-pass filtering in software), you'd get a double-hump response to a single tone. With great effort you might get one of the humps suppressed by
60dB, or greatly narrow the notch around DC, and that might make it a useful instrument, but it will always look a bit shabby to someone used to a commercial spectrum analyser. Still, it would be a project worth building.
If trying to DC-couple, then there are a lot of other challenges such as harmonics of the input signal pulling the LO (if it is a PLL), LO signal getting into the input port via the mixer, then reflecting off your DUT, giving you a VSWR-dependent DC signal that is indistinguishable from an input signal, etc. They are mostly solvable in principle with brute force heroic implementation that costs as much as a superhet. One quite nasty problem is 1/f noise in the baseband amplifiers - that is hard to fix or prevent.
Another problem with direct conversion is the response to harmonics of the LO. If you want low noise, stable gain with time and temperature, and linearity with variations in the input amplitude (to avoid IMD between the incoming tones) then you want to drive the LO port of the mixer hard, with a square wave. It will naturally respond to the wanted signal (let's call it f_displayed) but it will also respond to
3*f_displated, 5*f_displayed etc. so, for a single tone input, you will see the wanted tone and all of its subharmonics with diminishing amplitude at the lower subharmonics. Perhaps you could to a first order subtract them out in software, (e.g. if you have a tone at 15MHz, measure it at 15MHz LO, then subtract a third of that out when your sweep is at 5MHz, and subtract out a fifth of the 15MHz measurement when your LO is at 3MHz.) This won't work all that well, so perhaps the best approach is to use it as a narrow-band instrument (covering less than an octave at a time) and put a band-pass filter ahead of it.
I did once consider that for a direct-conversion radio covering say
0-100MHz, you might be able to generate a hard-switched LO that is a
1-bit sigma-delta approximation to a pure sine-wave with practically no content at the odd harmonics of the wanted LO frequency, (but lots of noisy stuff above maybe 500MHz). E.g. get a FPGA with a serdes and make it generate an 8Gbps bitstream that, if low-pass filtered, would give a pure sine wave at the wanted LO frequency e.g. 1MHz. Instead of low-pass filtering the bitstream, instead pipe it straight from the fpga serdes into the LO port of a mixer like the AD8343 or LT5560. Then the mixer should respond to the wanted 1MHz input signal but not to 3MHz, 5MHz,
7MHz, etc. input signals. There would be a response to all of the pseudo-noise hash that the bitstream contains above 500MHz, so you would need a low-pass filter on the RF input, which should be feasible.
Also, consider the option of buying a $10 DVB-T USB dongle and using that instead. There is existing software for it (although I have not used it). In this case the dynamic range is probably quite limited due to the 8-bit converters, but the price is right.
The fine details are in AN619. I haven't digested them all yet, but the non-integral divide ratios are not jitter-free, i.e. edge timing will move to the closest VCO transition, as you'd expect. There's Arduino code for it here, but again, a lot to digest:
The advantage of DDS is continuous 32 or 48 or 64 bit frequency resolution. The disadvantages include the binary radix, the steps in the output, and the requirement for filtering.
I'm looking for a low period jitter programmable clock source, and I guess we'll wind up with an octave-range DDS and a divider chain.
Somebody should make some nice packaged DDS output filters. Someone, TI I think, just came out with a diff-current DAC buffer amp. Two resistors and an AD8130 isn't bad.
--
John Larkin Highland Technology, Inc
lunatic fringe electronics
You do if you want to make accurate level measurements, or measure things like SSB whose line shapes aren't symmetrical.
Pretty much. Of course you can bypass the daylights out of the BFP640 collectors and run them into op-amp TIAs to give you more voltage. It's all DC there.
Another approach is to use a PIN-diode switch to chop the RF, and demodulate afterwards. You have to make sure to buffer it well so that the chopping doesn't change the impedance seen by the mixer RF port.
Interesting part. Hadn't seen it before.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics
160 North State Road #203
Briarcliff Manor NY 10510
hobbs at electrooptical dot net
http://electrooptical.net
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.