front-end filter for software radio project

[djacobow]

This may be a beginners question, but maybe not.

Anyway, I've been working on a project using the cheap $20 software radio dongles that are out there. If you are a HAM or just radio dabbler, you've probably run into them. Mine looks like this:

http://www.nooelec.com/store/software-defined-radio/sdr-receivers/tv28tv2.html

If you haven't played with these yet, you must!. They are a lot of fun. They are reasonably good receivers (not fantastic) but are insanely flexible.

Anyway, my project is one that involves tuning in the aviation VOR band 108-118 MHz. Immediately below this band is commercial FM radio with its many-kW transmitters, and directly above it are more airband frequencies used for voice comms. The ground based transmitters used by ATC also have a fair amount of power in this range.

Sadly, one killer weakness of the cheap SDR dongles is that they have very minimal filtering. The result is that I am getting IF overload and images from adjacent frequencies. Basically, my app can't work, which is a pity because I have put many many hours into the software!

Though I wanted this project to be all-software, it looks like I need to make a hardware filter to make these dongles usable for this application. (The app is fully automatic location using all available VORs.)

So, my requirements.

 - filter centered on 113 MHz or so with 3dB bandwidth of 10 MHz. (That's +/- 5 MHz)
 - as sharp cutoff as is reasonable. Many 10's of dB would be great
 - easy to construct (maybe dead-buggable, or with simple PCB)
 - easy to obtain and cheap components

In playing with this problem in LTspice I quickly realized I am in over my head. Just making a multi-stage "L" ladder filter, I can't get anywhere near the results I want without a lot of stages, a lot of guessing, lots of overall attenuation including the passband, and worst the selection of components I'm not sure actually exist.

So.... how to the big boys do this? What works, what doesn't?

I contacted a manufacturer of passive filters for a 4-th order Butter with two female BNCs and I was quoted nearly $700. If it had been $70, I would have bought it, even though that's 3x the price of the dongles.

Best Regards,
Dave J

[Paul Rose]

Look for a good 6 meter ( 50 Mhz ) ham band project like:

   http://www.oe1ifm.at/index.php?option=com_content&view=article&id=76&Itemid=75

And scale it up in frequency ( adjust inductors and capacitors to have the same reactance at the new frequency ).

Hand wound no core (i.e. air core ) should work at this frequency.  I use an inductance meter for stuff like this ( http://www.aade.com/lcmeter.htm ).

Paul

[penfold]

You could cascade a high pass and low pass commercial filter, mini-circuits filter blocks are circa 30USD could do something fairly imaginative using a couple either high pass or low pass to give you a bit more filtering at different frequencies.

You'd certainly struggle with noise if you were to go down the active filter route but perhaps you could do some pre-filtering with a passive filter, then amplify, then active filter then amplify again.  Thats just thinking out loud, would be worth doing some sums on unless the idea gets rubbished any time soon.

Regards
Dan S

[ElectroIrradiator]

What you are asking is not easy. The fundamental problem you have, is that the SDR dongle receivers frequently have horrible large signal tolerances and oscillator phase noise, not as much the lack of a front end filter.

If we ignore the fundamental RF problems for a moment, then asking for 'many tens of dB' below 108MHz, with a filter center frequency of 113 +/- 5MHz, is really hard. At the very least this requires a very good helical resonator filter (Google it), but finding the right one for your application could prove difficult.

Forget about cascading or scaling simple LC filters, they will be orders of magnitude from the performance you require. Additionally, even if you do find a suitable filter, you still face the problem of shielding the dongle itself. For instance if you have a strong, local FM station, it may completely bypass your fine filter, and use the power leads or the PC frame itself as an antenna.

SDR dongles are toys. Fun toys no less, but most are not really cut out for critical RF applications.

[djacobow]

Thanks everyone so far.

I think I have some good leads, but I appreciate ElectroIrradiator's point that this is pretty hard. Yes, my requirements are probably too tight. Actually, they're not even requirements so much as my initial guess at what I'd like.

A gentleman on an SDR group steered me towards a helical cavity filter as you suggest. He built one at a similar frequency range: http://www.f4gkr.org/2012/08/hot-air-gun-artwork-building-a-helical-cavity-band-pass-filter-for-air-band/

I will try that route, but lacking any means to measure the response of my filter, I'm loath to start winding my own coils. I also feel like though I understand the basic principles, lots of physical attributes: box material, spacing of components and size of cavity, etc, influence the performance and I have no way to model any of that.

However, the filter appears to be quite close to what I need, so maybe it's worth it. And so what if I need to also get a Rigol DSA815 with the tracking generator? (Actually, this same person suggested one of the cheap vector network analyzers that can be had on E-bay.)

Anyone have experience with cheap tools for filter measurement?

[djacobow]

Thanks, Pippy. I was wondering what real receivers in this range, do.

I am beginning to feel that this probably can still be done on the cheap (unit cost) but that the cheapie dongle is not a good starting point. But if I'm going to be designing custom hardware, I might as well dump the SDR bit as well as just use a "real" receiver.

Again, the goal is to quickly scan through the VOR frequencies. (They are ~25 kHz wide, 50 kHz spacing,  on 108-118MHz), determine where a carrier is present and the signal matches the VOR format, and add them to a working list. (At any given time there might be 3-5 good candidates within range.) Then, for each signal in the list, sit on that channel long enough to get the morse ID. From there, I can look the stations up in a database and round-robin tuning them updating the azimuth reading. Depending on the relative geometry, I can compute a position to varying degrees of accuracy.

I actually have most of this working, in pieces using recorded waveforms, etc, and with a fast PC it's actually possible to handle many channels at once (though I only have one at a time working) Still, the scanning could be quite fast.

The problem I ran into, as you imply, are images everywhere and all kinds of IM weirdness.

Some of the images can be managed by sampling as fast as possible. (My dongles go to 3.4Msps) and doing filtering inside that band all digitally. But I still need to keep images out of that swath, hence the need for some front end filtering. But at least the filter doesn't need to be 25 kHz wide.

[djacobow]

This bad boy almost fits the bill perfectly, but the surface mount aspect is a turn off:

http://www.minicircuits.com/pdfs/BPF-B113+.pdf

In the BNC terminated filters on minicircuits I could only find separate HPF and LPF devices, and the closest values were 100 MHz HPF and 150 MHz LPF.

[G0HZU]

The Minicircuits filter looks to be a simple lumped BPF. Probably 5th order.

You could make your own version with the same performance for maybe $5 + connector cost if you wound your own inductors. I'd strongly recommend using SMD capacitors on a little PCB.

But you should be able to get about 2.5dB insertion loss and -30dBc by about 100MHz and also at 127MHz with a passband of 108-118MHz. Note: Depending on the Q of the inductors you use you might see the insertion loss degrade to just over 3dB at 108MHz and 118MHz. But it will be a lot bigger than the MiniCircuits filter. Probably 6 inches by 1 inch.

A 7th order filter would have a higher passband loss (maybe 5dB) but you can get 20dB rejection as close as 106MHz and 122MHz. The rejection at 100MHz and 126MHz would be over 50dB.

To get the best performance for your situation you need to find out where the biggest interfering stations are. In this way the filter could be optimised.

[ElectroIrradiator]

This bad boy almost fits the bill perfectly, but the surface mount aspect is a turn off:

http://www.minicircuits.com/pdfs/BPF-B113+.pdf

In the BNC terminated filters on minicircuits I could only find separate HPF and LPF devices, and the closest values were 100 MHz HPF and 150 MHz LPF.

I suspect neither the DIY helical filter you linked, nor the Mini Circuits BPF will have adequate performance to make a serious difference in your case. You really need a filter, which 'drops off a cliff' below 108 MHz, and neither filter seem to do that to a sufficient degree. The MC filter only has 18 dB difference between bandpass center, and 100 MHz. So in a busy metropolitan area you will still get strong interference from FM stations.

It is possible to make significantly better helical resonator filters, but it isn't trivial to do, neither electrically nor mechanically. A good filter for this frequency will frequently have the 'coils' bent/wrestled into submission from 3 mm diameter copper rod, and use 4 or 5 stagger tuned resonators in series, with every internal surface in the filter cavities being silver plated throughout.

The reason the varicap tuned LC filter works for the airband receiver, is that the receiver input is somewhat 'strong', so it can take a fair amount of interference.

I would consider it mildly silly to buy a new SA, if you only need it to tune a single filter. The money could be better spent on scratch building your own, suitable receiver front end, varicap tuning and all. Plus learning the background for designing it. Finding and modding a second hand commercial airband receiver might also be worth considering.

[djacobow]

Thanks, G0HZU and ElectroIrradiator.

I'm going to go ahead and give the MC filter a shot since it's not that much money and ... what the heck. They have an eval board to put it on, and I've asked if they can do so. (I only have a soldering iron and so no easy way to get a large SMD part in place.) We'll see.

I was only kidding about buying the SA / NA. :-)

I will think about learning to design my own receiver. I'm aging digital EE type. No problem designing microprocessors and digital circuits, but my vague undergrad memories of high frequency analog were that these circuits would not work if you even looked at them wrong!

I'd be curious to know if there are parts on eval boards and/or reference designs where the frequency range could be changed with trivial substitution of a handful of passives. Does such a thing exist?