Raspberry Pi Pirate Radio -- Episode 355

This looks like a fun project to do with my son. I wasn't able to open this 'exclusive to members' content until I became a member. When I did I was expecting some more detailed documentation on this project, but only saw a parts list which, while helpful, wasn't necessary. Is there any 'deeper look' documentation? Particularly regarding the low & high pass filter circuitry? I'll admit I may be overlooking the link, but any help I would appreciate.

Kind regards!

Hi Nathan!

This was one of the very early projects on element14 Presents, before we really got into our groove, so to speak, and I didn't get a chance to do much follow-up regarding band pass filters. However, since you've joined element14, you do have access to one of the largest communities of electronics engineers and enthusiasts to ask questions of!

Feel free to drop questions in these comments, or start a post in the RF community!

These links might be helpful to get you started:

RF (Radio Frequency)

The specified item was not found.

Software Defined Radio Lessons with GNU Radio

Hi Nathan,

You're right, the values dictate the frequencies, and the schematic provides the overall shape. The values in either of the filters (the one in Frank's post, or mine) provide a reasonable bandwidth for this particular schematic. If you wanted much sharper cut-off, the circuit would need to change (it would need more components) and then it is harder to build because each part needs to be reasonably accurate. As it is, a few pF difference will make an impact to the filter, so the components listed are slightly more accurate than normal, and are also suitable for high frequency work.

Hi Frank,

I'm really dominating the Q&A on this topic; starting to get embarrassed, but I need to clarify something - to be sure my understand is on the right track. Using your example from rftools, signals will only be attenuated starting below 90 and above 110? Meaning, as in my demo, if I transmit a signal at 102.9 and there is a foreign signal at 103.9, they can still have an affect each other? Or is it that everything within 90 & 110 is attenuated from each other? I'm beginning to doubt what I've learned and when I do I freeze from moving forward :-/

HI Nathan,

No worries, I am pleased to try and help...  Filters are usually specified at the -3dB point and a Chebyshev filter is considered to have a sharp response.  So, everything below 90 MHz in the design I posted is greatly attenuated as is everything above 110 MHz.  The band in between passes most of the signal although Chebyshev have more ripple and are not quite as flat as some of the other filter types.  This ripple is more easily seen in the design that Shabaz posted although mine would have it as well.  For this particular crude application I think we can ignore that however.

As Shabaz has noted, there is variance in the components we are using and so it is necessary to have some kind of a plateau either side of the design point (102.9 in your case).  I had a look in my text which says Chebyshev are somewhat more sensitive to this but I think the design by Shabaz is sufficiently broad to cover that.  By the way, if I ever write something that seems to conflict with Shabaz, always go with what Shabaz says :-).  You could explore this by varying component value by their stated tolerance in the design tool if interested in the influence.

Now, how much should we worry about other stuff sneaking into what is being transmitted?  And it is important to remember this is transmission, not reception.  I haven't really investigated the radio here but it is bit banging the output if I remember correctly.  That means the Raspberry Pi is generating rectangular output over time varying between 0V and 3.3V.  It gets rounded off but it is still a really ugly signal.  If we look at it in the frequency domain then there will be lots of harmonics at other frequencies.  This is of course undesired in transmission because it causes interference across the spectrum.  But how bad is it?  First of all it is really low powered so it isn't going to transmit too far anyway.  Nonetheless, it is interesting to try and attenuate those harmonics and get rid of undesired energy away from the design point.

To give an idea of what happens, the FFT below is from a post I made a while back on entry level digital storage oscilloscopes with FFT capability.  It is a 49 kHz signal from a clock with a strong 148 kHz third harmonic.  The spikes that can be seen would be equivalent to the undesired transmission we want to get rid of.  Note that the 3rd harmonic and the primary signal are quite a ways apart (49 kHz Vs. 148 kHz).

Your signal won't look like this, just giving you an idea of what is going on.  If it was a perfect sine wave there would be no harmonics.  Your signal will have harmonics spaced some distance from the desired 102.9 MHz just like my FFT.  The idea with the filter is to make sure the radio transmission is inside the band pass and the harmonics or at least most of the undesired energy are outside of it.  Make sense?

Frank

Yes this is great Frank, a lot of good points here! One might accuse me of overthinking this project, but I have an OCD nature and you understand me with "it is interesting to try and attenuate those harmonics and get rid of undesired energy away from the design point". This is really about exploring what I can and cannot do. I obviously got the project working I just want to dabble and experiment with the filter side of things; clean it up if you will. Thanks again and until my next 'point of confusion'...

nk85  Forgive me if someone else has already mentioned this and I missed it but... you're using wirewound resistors (aka inductors).  You would have gotten better filtering if you would have used carbon film.  This little adaption cleaned up my pirate radio output.

As fmilburn states, the Pi is bit banging the signal out so I opted for a simple lowpass filter to attenuate the harmonics.

Thank you for bringing this back up.  I've been thinking about digging back into trying to make this work at Amateur Radio frequencies.  Wow.  I hope I didn't try this a year ago with the lowpass filter in...

Good point.  And the breadboard added unknown capacitance as well. 

Hi Nathan,

Frank's right, he picked a better filter design actually, and both should be good enough for your scenario.

You had the right approach, exploring website calculators, it's just that the particular input and calculator produced something that in practice was more unrealistic.

But this is the normal way, everyone uses the website calculators or software to design filters, and exploring how things change as values are modified, or as more stages are added (for steeper roll-off). It's a cheap way of virtually experimenting without needing to buy components, but it will be only as good as the values you feed it. For instance, if you known your components have (say) 5% tolerance, then you'd have to run your formulas or simulation with the different expected values, to at least see if the effect is significant or not. Prior to online calculators, people used tables from books.

There are free simulators like LTspice for totally custom component configurations (as opposed to calculators which will only offer defined topologies), which you could use too, but for filter design and other small building-blocks it can be better to use online calculators, since they are pre-progammed with established topologies.

When you come to want to test it with test equipment and draw the frequency response, the NanoVNA is probably a good way (quite low cost too), I've not used it, but there's several blog posts by others experimenting with it and it looks ideal for this.

Hi there, I'm back at it for a bit. Some fun with paint lol - it's been ages. Can you tell if if I'm somewhat on point here? I've attached an untampered image if you want to use it to correct my mistakes. Thanks so much!

Hi Nathan,

Those proto hats are nice as they allow the breadboard to be transferred directly to a more permanent solution.  I suggest putting it on a breadboard first, make sure it works, and then solder it like it was breadboarded.  There are some issues with it at the moment:

• Rs is the supply resistance and won't be on your board
• Rl is the load resist and is the speaker - it is not on the board either
• Remember that the vertical strips on each side of the breadboard gutter are connected.  If wired as shown, it will be necessary to cut the trace which is unnecessary if wired differently.
• Use the lead wires on through hole parts to minimize the soldering and no traces need be added to the back
• The ground and +5V/+3V are already connected to the power rails
• Run the wire to the input from the inside rail - as shown it is connected to the header that plugs into the Pi

I'll try to mark up the board and post it in a bit.

Have a look at this....

I can't remember which pin is the one bit-banging on the Pi so check that.  There are lots of ways to lay this out but note that this will minimize soldering by using the leads on the components and traces on the protoboard to make the connections.  If you decide to string it out the same way as in the schematic that is OK but I like to put the wire on top of the board, and not bridge solder or put wire on the bottom as it is easier to see the circuit and I think looks better.  But that is just personal preference.

You might want to flip it 180 degrees as well depending on where the speaker wires go which would also minimize the jumper length from the Pi pin to the filter input.

Have a look at this....

I can't remember which pin is the one bit-banging on the Pi so check that.  There are lots of ways to lay this out but note that this will minimize soldering by using the leads on the components and traces on the protoboard to make the connections.  If you decide to string it out the same way as in the schematic that is OK but I like to put the wire on top of the board, and not bridge solder or put wire on the bottom as it is easier to see the circuit and I think looks better.  But that is just personal preference.

You might want to flip it 180 degrees as well depending on where the speaker wires go which would also minimize the jumper length from the Pi pin to the filter input.

Wow, I really appreciate you taking the time to highlight those points and for marking up a sketch. By 'speaker' do you mean 'antenna'? Some of the components I was needing were only available in SMD, so to over come that I was going to use an adapter, of sorts, as in the image below. Meaning that I couldn't stretch it like you set it up. I'll either have to modify my bandpass components based on availability with 'through hole' or keep the schematic as I built it and make adaptations based on what you did.

Doh!  Yes, I have been working on an audio amplifier :-)  Embarrassing sign of old age...

Which parts are the ones you can only get in SMD?  The reason I ask is that sometimes it is easy enough to solder them between 0.1" pads.

I might as well put it out there. I'm hesitant because I don't want to hear that it won't work lol. But the truth will 'set me free' - or 'not spend money where I don't need to. Tell it to me straight!

The LO201R68 inductor has a 0201 footprint - that is tiny and will be extremely difficult to hand solder.  I recommend at least 0805 if you are trying to hand solder SMD parts for the first time although 0603 aren't too bad.  The Bourns inductor has an uncommon footprint.  I see you have tightened up the band.  I suggest axial through hole where possible.  You can get most caps in SMD 0805 (your Kemet cap is 0805). 

Hi Nathan,

Great to see you're really getting into filter design. As Frank says, the larger parts (0805, or, worst-case 0603) are needed. I don't think I've ever soldered 0201 (you'll also see why if you order one of those : ).

Regarding the filter design, this looks valid, but in practice the issue you'll have is that all parts have a tolerance (maybe 2.5% or 5% etc), and if you choose to design such a narrow filter (1 MHz bandwidth at 100MHz) then there's a chance that the implemented filter could easily be off by a MHz or more. There's actually no need to implement such a narrow filter, because the output from the Pi won't (or won't have much) have anything apart from the harmonics, which are exact multiples. So, to guarantee that the filter works for your purpose, it is better to go for a more wider bandwidth filter.

On another thread, Mike mentioned that you may not even need a bandpass filter, and a lowpass may be sufficient incidentally.

Question on Harmonics and Filters

There's no harm with a bandpass (provided it is of a wider bandwidth as mentioned), but according to that thread, you will likely be fine with just a lowpass filter.

I would much rather use through hole components to be sure. But as I tightened up the band, it was asking for parts that didn't come available in through hole type. Reason being, for the tighter band, was to avoid other in-use frequencies near by (above and below the 102.9). Thats been a big part of it I guess, trying not to interfere with neighbouring in-use frequencies. Also the online catalogs (like Newark, DigiKey-Key & Mouser) overwhelm me; so many choices and so many factors to each choice. I'm definitely to rookie to be left to successfully choose the parts, so I've been delegating to the tech. support from each company to satisfy the schematic. That turned into SMD parts pretty quick. I can't always get it right the first time, and I don't suspect to, but this extracurricular interest might not be for a tight budget guy. If I could only get the part selection nailed down I shouldn't have to spend and re-spend because I didn't do my home work. Knowing the reason for the tighter band, what would your shopping list look like? I'd be interested in what Shabaz puts together too!

My shopping list would be for Coilcraft surface-mount inductors, mainly because they make ones with a tight tolerance (2.5 or 5%). For capacitors it would be NP0 types which I think are only surface-mount. The reason for those is that they have a tight tolerance too. Such a narrow filter needs accuracy in the design.

Here's a sketch of what I believe is coming out of the Pi :

The triangles represent output (it won't be triangular shaped), and the orange triangle is the desired output, however the Pi will output all three triangles in the diagram (and many more, theoretically off toward infinity). This diagram assumed the Pi software is set to broadcast at 102 MHz. The content at 306 MHz, and 510 MHz and so on, needs removing.

The width of the triangle is 0.3 MHz (it won't be precisely that, but it is of that order, and again theoretically it goes off to infinity, but at vanishingly low output).

According to this sketch, the Pi will output nothing between about 102.15MHz and 305.85 MHz, so a filter that has a response such as the orange curve would work well. Steep filters cost more (need more components), but in this case with the orange curve there is almost 200 MHz of space for the curve to drop off as low as possible, so a steep edge is not needed.

Another option would be the green filter curve, and again sharp edges are not needed, because the Pi will not output anything undesired up to ~306 MHz.

The danger with the very narrow filter (purple curve) is that due to the component tolerance, it may have shifted from 102 MHz to 101 MHz, filtering off a lot of the desired orange transmission. For this reason, very sharp narrow bandpass filters are almost never made using inductors and capacitors.

I am away from home with only a phone - Shabaz gave a great answer though

This is a great visual Shabaz. You and Frank both have provided so much insight! I've archived this entire thread for future referencing. Can you refer any good academic books (with a beginner in mind) that speak to this field of topic(s)? I'd be lying if I said I understood everything you both posted since we started chatting. The more I pick your brains though the more I am further intrigued. I adopt a bazaar thirst sometimes for having to understand everything that is going on, even if only semi-related to the topic at hand.