In a previous episode of Hackaday, [Rich Olson] came up with a new no-etch circuit board fabrication method. And now, he’s put it to the test: building an nRF52 Bluetooth reference design, complete with video, embedded below.
The quick overview of [Rich]’s method: print out the circuit with a laser printer, bake a silver-containing glue onto the surface, repeat a few times to get thick traces, glue the paper to a substrate, and use low-temperature solder to put parts together. A potential drawback is the non-negligible resistance for the traces, but a lot of the time that doesn’t matter and the nRF52 reference design proves it.
The one problem here may be the trace antenna. [Rich] reports that it sends out a weaker-than-expected signal. Any RF design folks want to speculate wildly about the cause?
We are split on the issue of DIY PCB fab here at Hackaday Headquarters. Some of us are able to get results out of simple toner transfer and etching that push the limitations of a 1200 DPI laser printer, with practice and calibration. Others of us simply ship the work out, which certainly makes economic sense these days. Don’t get us started on CNC routing, photo-resist etching, and whatever else. Now along comes [Rich] with his method that we want to try out! It’s a miracle we have any time to write.
That trace antenna is sized for the 63mil of FR-4 behind it on a normal PCB. Your paper and substrate probably has a much lower permittivity than the ~4.5 of FR-4 which means that the antenna is too short for 2.4GHz on this board. Probably would be better off using off-the-shelf ceramic chip antennae until someone with some RF-test gear can get a trace antenna sized for this substrate.
There’s probably some distributed capacitance due to the porosity of the trace material.
…? Capacitance is formed when you’ve got a metal nearby a ground, and the capacitance is formed by storing charge on the outer layer of the metal and ground. Popping tiny holes in the trace volume probably isn’t going to change things that much – there was no charge storing going on there anyway. You don’t get capacitance *internal* to the trace because the metal’s all at the same potential.
Now, the Q of the antenna will definitely be lowered due to the lower conductivity. That’s obvious. But the problem with the antenna is almost certainly the substrate thickness as AMS mentioned.
The trace around the perimeter running a track’s width from the business end of the antenna isn’t helping either.
Indeed, antennas tend to be very sensitive to stray conductors running nearby. I expect that trace is picking up a large portion of energy the antenna is transmitting.
I like your style! The fact that you could make a BLE work at all is great. Maybe a thicker trace for the antenna would increase the Q by reducing the resistance. Can you just hand paint the antenna with the silver paint and recheck?
Thanks Cray! finally posting here after some of the flames have dwindled a bit…
I think gains by adding additional silver paste will be incremental – probably not solve the issue.
I did tack on a 31mm-ish wire (2.4ghz quarter-wave-length) – and managed to boost the signal by about 18db! So problem is definitely the antenna itself…
It’s a bit tricky – but it is possible to coat the silver paste traces with low-temp solder. Think this might be best approach to doing a trace antenna.
solder-coating the silver traces could also be useful any place lower resistance is needed…
realistically a chip antenna is better option…
It seems that there is no GND plane, the antenna needs a ground plane to make up the other half of the antenna. also there is the trace going around the antenna, this will de-tune the antenna greatly.
But normally the antenna does not want a ground plane beneath itself. It wants it under the rest of the circuit as a good and stable ground reference.
Quite likely the reduced performance of the antenna is due to something called the “skin effect” which is the property of RF current only flowing on the very skin of a conductor. The author states the conductors have a high resistance and he counters that by making them thicker. All well and good for DC and lower frequency AC/RF, but little to no help for 2.4ghz. Instead of the energy being radiated it is heating the antenna via I-R loss.
http://www.allaboutcircuits.com/tools/skin-depth-calculator/
Yeah, if they simply copied the dimensions for a copper trace antenna, that would have been designed for a much lower impedance conductor. For this design method perhaps switch to a surface-mount ceramic antenna component.
Yeap, besides the increased skin effect loss as others have pointed out the dielectric constant of the substrate is a wild card
A mirror image of the electric field is projected on to the ground plane of the inverted F antenna. When you don’t have that plane, you are not implementing the antenna correctly and also the impedances is wrong.
It’s not supposed to have a plane under the antenna.
Take a look at the TI reference design here, which has the geometry of the antenna. But it does need a good low-impedance ground connection, which it doesn’t really have here.
Just a clarification here (also below) the plane that’s actually missing here is just a ground pour beside the antenna, not a plane underneath the antenna. You still might be able to do that here.
The width and thickness of the cardboard has to be adjusted for the differences in dielectric.
Did I actually said about ground plane behind the inverted F? The ground plane should be under the circuits – and there *should* be via connected to the inverted F.
https://en.wikipedia.org/wiki/Inverted-F_antenna#/media/File:PIFA_antennae.png
The darker grey is the ground plane.
Many many issues here…. not just one
This design has no ground plane at all! It has just some tiny ground traces – look at the two capacitors near the crystal. There is also not a single via to a ground plane – because there is none. That is no RF design. It’s more something soft and brown, but not chocolate :-( But it’s amazing that he gets more than 10cm of range with this.
The lower conductivity increases the skin depth. But the low conductivity increases losses also.
“Maker” doesn’t “Make” – Again! Options:
1. Hope people posting here can help you (reasonable chance).
2. Get a REAL education in Electrical Engineering.
Sorry to be so harsh – but that’s the way it is IMO.
Probably need to be at the graduate level before you’ll see microwave design and some finite element simulation and that’s a specialized course and not part of core.
There are excellent app. notes for anyone else that bother to google, so there are no excuses for not reading up on things. Seems like “makers” are too busy to make video than to do the reading up part.
Yeah, before you touch an antenna of any sort you should have a degree in how to build one from scratch.
Say, Mr. EE, how many classes did you take in operating system design, OO and functional programming, protocol design, flash memory managment, audio DAC and codec implemtation, http server and database design, and computer graphics? What’s that, you say? Alright, please step away from the computer, it is clear you are not learned enough to design one from scratch so we can’t have you browsing the internet.
Now that the sarcasm is out of the way: I would offer help, but the problem of antennae is outside my field. Were it a code problem, a μC issue, math or chemisty, I might be able to help. So keep making, makers, and don’t let the trolls but you.
That’s not a proper implementation of the inverted F antenna: See https://en.wikipedia.org/wiki/Inverted-F_antenna
> The advantage of doing this is that the input impedance of the antenna is dependent on the distance of the feed point from the grounded end.
– The ground plane is also part of the antenna which is missing from this. The impedance is totally wrong because of that.
– The dielectric constant will be different, so track width/media thickness etc needs to be adjusted
– The silver paint has high resistance, so low Q.
This is worse than Chinese knock-off – simply copying the form without understanding the material differences. Making antenna isn’t easy and making antenna with the wrong materials is asking for trouble.
How exactly do you think people learn?
By doing research before experimentation.
I have no formal RF training or design experience. But with many hours of reading up on microstrip routing, RF routing and component selection guidelines, and reading through the TI inverted-F antenna appnote many times over before starting layout, I made a 2.4GHz gadget that worked over expected distances on the first board spin.
Failure is a good teacher, but expecting to learn by failure alone is stupid.
To be clear here the “plane” portion isn’t a ground plane *under* the antenna. It’s actually a ground plane *beside* the antenna – there’s nothing under the antenna at all. Still the same thing, viewed from the side, but easier to implement because it’s all in one layer. This could’ve been done on this board pretty easily.
A ground plane under the circuit is necessary.
When you modify the design, you’ll need to run RF antenna simulation.
@Capt. McAllister… Some of us actually spend time reading up the app. notes on antenna design. There are excellent app. notes from TI that explains things. It is so easy to access all that on the internet, so there is no excuse of being ignorant. No excuses – you make youtube video etc means you have access to internet.
The dielectric constant of the substrate is quite pointless in this “design” – because there is no ground plane anyway. At first I also thought that the high resistance could be the culprit of bad performance, but that’s secondary.
Not only “making antenna” but “making RF circuit”.
Having read through and enjoyed these comments on antenna (under)performance I think it’s clear: new need a recurring “Why Doesn’t This Thing Work” column. Anyone have ideas for the “victim” hardware for the first couple of these? Send it to our tips line please!
If it’s not running with an ESP. It’s not a test for me! (j/k)
I so love me some novel “look, ma, no etching!” PCB manufacturing methods that are _at least_ as convoluted, laborious, multi-step and involving any number of _other_ chemicals to often produce a massively inferior “PCB”. FFS, I don’t give a crap about the etching operation itself, it’s all the fiddling I’m objecting to! I guess “slap a PCB blank onto a 3020 CNC and pick it back up when it’s done” is still the way to go, for a few more decades at the very least, unless you feel like paying several king’s ransoms for 10 postage-stamp sized boards (out of which you have no use for the other 9) and/or wait a few months to have it manufactured in the Land of the Cheap…
There are tons of quick turn manufacturers in the Land of the Cheap that turn around orders as fast as any of the local houses and with decent quality. Elecrow has been good to me in the past, taking only 2 weeks to go from order to doorstep.
Did he solder the component on the left side antenna lead properly? Can’t tell for sure from a photo, but that doesn’t look fully soldered to me.
Hi RoboMonkey. Did some testing – and I can confirm the soldering up to the antenna trace is good. Problem is definitely the poor conductivity of the antenna itself.
This kind of thing always worked, but is marred by the high cost and slow inferior process compared to other ways.
yes it always worked i’ve been doing this for years.
First. Very cool concept!!
A couple of items. The nRF52832’s matching network shown on the schematic in the video is to match the ANT (RF OUT) pin’s impedance (which is not 50 ohms) and converts it to a 50 ohm feed point. It also functions as a low pass / harmonic filter. A additional network is needed to match the antenna. Normally this a PI network when a monopole or inverted-F antenna is used. Most of the time not all three components are needed. So one can start with two unpopulated shunts and one series with a 0 ohm resistor. The actual number of components and values may be determined at the time of antenna tuning and testing.
Also there is no ground pour on the top layer or on the bottom. This is needed to be the “mirror” for the antenna. This is very important for a inverted-F antenna. It could be (probably is) the reason for the observed performance.
This grounding may be hard to do as holes vias would need to be added to stitch the bottom layer to the top. Not impossible as just another “board” could be printed that mates up to the current one and holes and solder added. Note that there should not be any ground under the antenna proper.
Also get rid of that surrounding trace that is holding that antenna hostage. :-) I would not be surprised if the load cap values need to be changed. Way to tell is to look at the Center Freq at 2440MHz. But it works and sometimes you don’t want long range anyway.