First RF project

[tggzzz]

Have you contacted any local amateur radio (ham) groups?

They will typically have two types of members: ones that like to use radios and ones that like to build radios.

That's a good idea. I think I, too, would enjoy building a radio, but I would also enjoy not falling into the most common first-timer pitfalls (board layout, component choice, etc, etc).

My strategy is to build some simple circuits, e.g. filters, and the test equipment to test them. I'm not too interested in building a complete radio from individual components; my interests lie in using components and modules to do something "new" and "interesting". (Where the definition of new and interesting is entirely egocentric )

Currently I'm experimenting with an SDR dongle to see how it can be used as a spectrum analyser, scalar network analyser, and time-domain reflectometer. One day I'll get back to making my 2GS/s 4GHz scope.

[djacobow]

Currently I'm experimenting with an SDR dongle to see how it can be used as a spectrum analyser, scalar network analyser, and time-domain reflectometer. One day I'll get back to making my 2GS/s 4GHz scope.

Which one are you playing with?

It is gradually dawning on me (because I can be a bit slow) that a good part of the problem with the RTL-SDR is not the lack of front-end tracking filter, but that the IF filter after the mixer is just not good enough. That could be because the dongle was designed from 2 Msps or whatever for DVB-T and, though you can run the ADC slower, I'm not certain the IF filter narrows automatically or or anything like that.

However, I am also in possession of a FunCube Dongle Pro +. It is designed for 250ksps and I suspect has the IF filter to match. I haven't flown with it to collect signals. I have found fumbling with a laptop and dongle to be a bit of a challenge while flying a plane, particularly in the local airspace. I need to work out that minor problem.

I am also interested in another radio, the Airspy, which seems to have made a point of having good IF filtering. However, it is a high-bandwidth device (10MHz) which actually makes my life harder.

Good luck with the scope. Projects...

[T3sl4co1l]

The paper does not mention any gain stage prior to the mixer. Is that common and/or reasonable?

Sort of.  A passive mixer always loses 6dB by its operation, which means your SNR is 6dB worse, straightaway.  Your dynamic range is also reduced by as much, plus whatever the maximum signal level the mixer itself can run at.

With a preamp, SNR rises, so that by magnifying the input (and adding only a little noise in the process), the mixer's 6dB loss looks like only, maybe, 1dB, so that after the amplifier's say 2dB noise figure, the SNR is now 3dB off ideal, a 3dB improvement.  Dynamic range has been reduced even further, however, because not only does the amplifier experience IMD and compression, it also drives the mixer that much harder.  If the RF amp is part of the AGC circuit, this can be somewhat mitigated, but only with respect to the station being received, not over the total band (i.e., a station at 110MHz and 0dBm might overdrive the front end, while trying to receive 112.5MHz at -60dBm).

The same SNR-dynamic range concerns apply to ADCs (the ENOB being a very obvious limit to both!).  If the sample rate is high, much filtering and averaging can be applied to recover more ENOB (effectively, filtering the quantization noise improves the noise floor), so it need not be exactly 72dB.  It's all about the processing.

If you do wish to explore a hetrodyne converter (it's not really superhet unless the RF band is also being tuned as a tracking filter), I would suggest something like an SA602 with a varactor tuned LO (driven from a DAC, assuming you have one handy) and a frequency counter (so you can keep adjusting the DAC output to keep the frequency offset correct).  This should target a 10.7MHz IF, provided by a ceramic filter -- a standard component in any FM radio.  The SDR can run at whatever sample rate is desired, but if you have the capability and processing, you might as well keep it at maximum, since quantization noise is independent of sample rate, which means you can do more filtering to recover more SNR.

Tim

[tggzzz]

Currently I'm experimenting with an SDR dongle to see how it can be used as a spectrum analyser, scalar network analyser, and time-domain reflectometer. One day I'll get back to making my 2GS/s 4GHz scope.

Which one are you playing with?

One of the generic R820T2+R2832U devices. The front-end (partial) datasheet can be downloaded from http://rtlsdrblog.rtlsdrblog.netdna-cdn.com/wp-content/uploads/2013/04/R820T_datasheet-Non_R-20111130_unlocked.pdf

It is gradually dawning on me (because I can be a bit slow) that a good part of the problem with the RTL-SDR is not the lack of front-end tracking filter, but that the IF filter after the mixer is just not good enough. That could be because the dongle was designed from 2 Msps or whatever for DVB-T and, though you can run the ADC slower, I'm not certain the IF filter narrows automatically or or anything like that.

That would be far too sophisticated! But the "SDR philosophy" is heavily geared towards minimal RF components and do all the interesting/clever filtering and demodulation using DSP after digitisation. They only back away from that where necessary.

[rfeecs]

Sort of.  A passive mixer always loses 6dB by its operation

Just had to comment on this.  I have seen this statement often and it is not true.  A passive mixer can have as low as 3dB conversion loss theoretically.  The ADE-6+ has about 4.5dB conversion loss at 100MHz for example:
http://www.minicircuits.com/pdfs/ADE-6+.pdf

If you do wish to explore a hetrodyne converter (it's not really superhet unless the RF band is also being tuned as a tracking filter)

Not sure what you mean be that.  Superheterodyne just means supersonic heterodyne, meaning mixing radio frequencies instead of audio frequencies.

[rfeecs]

It is gradually dawning on me (because I can be a bit slow) that a good part of the problem with the RTL-SDR is not the lack of front-end tracking filter, but that the IF filter after the mixer is just not good enough. That could be because the dongle was designed from 2 Msps or whatever for DVB-T and, though you can run the ADC slower, I'm not certain the IF filter narrows automatically or or anything like that.

I have zero experience with the RTL-SDR and I'm a bit confused that you are not able to select 50kHz channels.  I see people using them to receive FM broadcast channels which are 200kHz apart, so why can't it receive AM channels that are 50kHz apart?

Apparently many of these units have an R820T receiver front end, which is a superheterodyne receiver incorporating an image reject mixer with 65dB image rejection.  So you don't have to worry much about image frequencies.

I would assume the RTL device properly filters the IF signal before sampling, so where is the problem coming in?

Is it because there are very weak and very strong channels present at the same time?

[djacobow]

It is gradually dawning on me (because I can be a bit slow) that a good part of the problem with the RTL-SDR is not the lack of front-end tracking filter, but that the IF filter after the mixer is just not good enough. That could be because the dongle was designed from 2 Msps or whatever for DVB-T and, though you can run the ADC slower, I'm not certain the IF filter narrows automatically or or anything like that.

That would be far too sophisticated! But the "SDR philosophy" is heavily geared towards minimal RF components and do all the interesting/clever filtering and demodulation using DSP after digitisation. They only back away from that where necessary.

I'm such a n00b here, I don't quite get it, but it seems to me that if your IF filter cannot block out stuff above the nyquist limit, then you will get images in your digital data, and all the DSP in world can't fix that.

If I understand the datasheet, this particular dongle was designed for 6 MHz channels, and so has an IF appropriate for that, running the sampling at 6-8 MHz, too.  As such, that works for TV, but running the sampling slower is an automatic invitation for trouble. Also, the adjacent channel spec in the datasheet is 47dB which seems not great, and that's assuming the 6 MHz channels. I know a lot of hobbyists are running the sampling slower because they want to use Raspberry Pi or USB1.1 and want it to be easier for the computer to keep up, but it seems like in the general case this shouldn't work very well at all.

But now I am wonder, is it worth a quick-and-dirty attempt to hack one of these dongles to hack in low pass filter on the IF between the R820T2 and the 2832U, like at 15 or 20kHz, then operating the 820T2 in Zero-IF, so that the signal sits right at baseband. Now I get my narrow band without the reflections from other channels. Would this work? It seems like building a filter that would block the IF all the way down to audio frequencies would be easy.

Could that work at all?

[djacobow]

I have zero experience with the RTL-SDR and I'm a bit confused that you are not able to select 50kHz channels.  I see people using them to receive FM broadcast channels which are 200kHz apart, so why can't it receive AM channels that are 50kHz apart?

...
I would assume the RTL device properly filters the IF signal before sampling, so where is the problem coming in?

Is it because there are very weak and very strong channels present at the same time?

Well, I don't know. I would not rule out user error. But I think one constraint I maybe failed to mention is that running the unit at its full sampling rate of 6MHz puts a load on the computer and USB that introduces other problems. And because the signal I was interested in is not nearly that wide, I thought I'd want to run slower. But I think everyone has made it very clear to me now that if the IF is designed for 6 MHz, you really have to run the sampling that way.

Now the issue of strong and weak station could be another issue. I can imagine that upping the gain to get the weak station you want could cause the strong station to overdrive the input LNA and the mixer and introduce all kinds of distortion. Could that cause images even if the IF filter is correctly matched to the sampling rate? It seems if that is the case, then front-end filtering is the solution. But why don't scanners and normal receivers have this problem?

I'm starting to feel guilty about asking so many questions. Really, if there is reading that is the right place to start, I'm open to it. I feel like I have read plenty of basic DSP and several RF intro things, but I have't found the practical "Art of RF Electronics" I'm imagining is out there.

[T3sl4co1l]

Sort of.  A passive mixer always loses 6dB by its operation

Just had to comment on this.  I have seen this statement often and it is not true.  A passive mixer can have as low as 3dB conversion loss theoretically.  The ADE-6+ has about 4.5dB conversion loss at 100MHz for example:
http://www.minicircuits.com/pdfs/ADE-6+.pdf

I thought that felt wrong... 3dB should be it.  A single balanced mixer for example drops half the signal (it's alternately being connected to the output, or not, by the LO drive), so loses half the voltage; half of which, in turn, is at DC (baseband + harmonics) (the average reads half, or -6dB) and the other at LO + harmonics.  A double balanced mixer alternates between on and on (+/-), so only loses 3dB, which is equivalent to splitting fo into fo + LO and fo - LO.

Although, 6dB is at least a typical ballpark (your 4.5dB example not being terribly far off, and truly terrible mixers like single diodes being even worse still).

If you do wish to explore a hetrodyne converter (it's not really superhet unless the RF band is also being tuned as a tracking filter)

Not sure what you mean be that.  Superheterodyne just means supersonic heterodyne, meaning mixing radio frequencies instead of audio frequencies.

It was always my intuition that "super" was because it's a hetrodyne set (i.e., converts RF to IF via LO + mixer), but "super-" because of the addition of tuned-RF tracking (e.g., a standard AA5 set).

Wikipedia says Armstrong called it "super" to begin with, so I guess that is indeed the historically correct term, and there's no such thing as a "hetrodyne" set!

Tim

[rfeecs]

Well, I don't know. I would not rule out user error. But I think one constraint I maybe failed to mention is that running the unit at its full sampling rate of 6MHz puts a load on the computer and USB that introduces other problems. And because the signal I was interested in is not nearly that wide, I thought I'd want to run slower. But I think everyone has made it very clear to me now that if the IF is designed for 6 MHz, you really have to run the sampling that way.

Now the issue of strong and weak station could be another issue. I can imagine that upping the gain to get the weak station you want could cause the strong station to overdrive the input LNA and the mixer and introduce all kinds of distortion. Could that cause images even if the IF filter is correctly matched to the sampling rate? It seems if that is the case, then front-end filtering is the solution. But why don't scanners and normal receivers have this problem?

I'm starting to feel guilty about asking so many questions. Really, if there is reading that is the right place to start, I'm open to it. I feel like I have read plenty of basic DSP and several RF intro things, but I have't found the practical "Art of RF Electronics" I'm imagining is out there.

You may need to find someone that's experienced with the RTL-SDR.  But it looks like there is an additional programmable anti-aliasing filter inside the RTL chip.  It's not clear to me.  But I don't know how it could work as an SDR without one.

Yes, if you overdrive the front end it could screw everything up.  You will get intermodulation distortion causing channels to mix together.  Large signals in two channels could cause IM products in other channels and swamp out smaller signals for example.

[vk6zgo]

Sort of.  A passive mixer always loses 6dB by its operation

Just had to comment on this.  I have seen this statement often and it is not true.  A passive mixer can have as low as 3dB conversion loss theoretically.  The ADE-6+ has about 4.5dB conversion loss at 100MHz for example:
http://www.minicircuits.com/pdfs/ADE-6+.pdf

I thought that felt wrong... 3dB should be it.  A single balanced mixer for example drops half the signal (it's alternately being connected to the output, or not, by the LO drive), so loses half the voltage; half of which, in turn, is at DC (baseband + harmonics) (the average reads half, or -6dB) and the other at LO + harmonics.  A double balanced mixer alternates between on and on (+/-), so only loses 3dB, which is equivalent to splitting fo into fo + LO and fo - LO.

Although, 6dB is at least a typical ballpark (your 4.5dB example not being terribly far off, and truly terrible mixers like single diodes being even worse still).

If you do wish to explore a hetrodyne converter (it's not really superhet unless the RF band is also being tuned as a tracking filter)

Not sure what you mean be that.  Superheterodyne just means supersonic heterodyne, meaning mixing radio frequencies instead of audio frequencies.

It was always my intuition that "super" was because it's a hetrodyne set (i.e., converts RF to IF via LO + mixer), but "super-" because of the addition of tuned-RF tracking (e.g., a standard AA5 set).

Wikipedia says Armstrong called it "super" to begin with, so I guess that is indeed the historically correct term, and there's no such thing as a "hetrodyne" set!

Tim

Well,a Direct Conversion radio or a Regenerative Receiver could probably be called "heterodyne sets".

Both were designs which existed,& were used at the time.
A DC set or a "Regen" in the oscillating mode produce an audible beat note or heterodyne at the Rx output.

Frequencies above the Audible range were commonly called "Supersonic" frequencies,with little differentiation on whether they were referring to an Acoustic or Electronic signal.

A Superheterodyne was so called because the resultant heterodyne was above the range of Audible frequencies.

The other use of "Supersonic" was to refer to the speed of an object,such as an aircraft,projectile,etc.

Since that time "Supersonic" has come to  only have the latter meaning,being replaced in the other sense by "Ultrasonic".

[djacobow]

So, I flew the other day with the Funcube Dongle Pro+ with a simple rubber ducky and a 108-118 MHz bandpass filter in front. The FCDPP has a 192kHz sample rate and appropriate filtering after the mixer to go with.

Results were much better than I ever got with the RTL-SDR. With some fiddling I was able to get usable signals from most VORs that I should have been able to receiver. Of course, the dongle was quite deaf compared to the ship's radios, but they are working from properly tuned antennas on the OUTSIDE rather than a ducky inside.

However, still not smooth sailing. With some fiddling of AGC and gain at various points, I was able to get a nice signal, but there didn't seem to be a setting that got good results in all conditions of received power. That's not good for what should be an autonomous system.

Overall, I'm encouraged, but this project just gets harder and harder. It is one thing to fiddle with SDR# or even gnuradio-companion and casually decode signals, but making a system work is seeming much harder.