#rfbiscuit – Use Case 3: Bandpass Filter

My starting point to make RF Biscuit was actually to make lowpass / bandpass filters to experiment with rpitx – using the Raspberry PI as a signal source. (thanks to F5OEO this way) – My plan was using a FilterDesigner like Elsie – and design a filter that would cut away any unwanted harmonics – the Raspberry PI is outputting a square wave so you can expect the odd harmonics – most prominently the 3rd and the 5th – to appear as unwanted spikes in the spectrum.

Although not easy to spot the completed filter looks like this:


(unfortunatly I didn’t had the DSLR with me this time and I am using an oldschool Sony 3M-Pixel camera, which lying around in the lab)

My plan was to implement a bandpass in a mesh capacitor-coupled topology – using Elsie I came up with the following part values:


btw. Is this now a 5-Order bandpass??

Implementing this Filter I found out that I took too little time reviewing the design of Rev A. – because the inductors ( Murata LQW18ANR39G00D 0603) in series with the capacitors can not be easily fitted on RF Biscuit as you can see in the following picture:


So for Rev B. I would like to avoid constructs like this for this filter topology.

Also the inductors are really sensitive devices – especially when you solder them by hand. I had to replace two of them when I tried to solder them the second time – while handling them with tweezers it is really easy to damage the windings of the inductor.

In the Simulation this Filter should perform as a 85 Mhz to 105 Mhz bandpass:


And the measurement with the VNA shows following result:




#rfbiscuit – Use Case 2: Attenuator

For the second use case I wanted to try design a 30dB DIY-Attenuator. I used this attenuator calculator (german) to calculator the resistors for a PI-configuration at 50 Ohm input and output impedance. I needed to place 53.3 Ohm and 789.8 Ohm resistors – in order to generate the right resistor values – I choose to put for each resistors, two resistors in parallel. The needed values I calculated with the tool pointless-resistance, which automatically searches the values of the E-Series for suitable candidates within a defined tolerance.

for 789.8 Ohms – I used 820 parallel with 22k – which results in 790.5 Ohms

for 53.3 Ohms – I used 56 parallel with 1k2 – which results in 53.5 Ohms


In the Simulation you can see that major share of the power has to be dissipated by the lower resistor close to the input. The 0603 Resistors I am using have a power rating of 100mW which is the limiting factor. 100mW translates to 20dBm which is enough for my experiments. (see Watt-dBm converter) – But a more clever solution would be to use different resistors in physical size, because in our schematic only R2 has to dissipate the majority of excess power. For example a 1218 Resistor would be sufficient enough for 1Watt / 30dBm. (See resistor sizes and power dissipation)

After the math it is time for solder – and the result looks like this:


And to measure its performance lets connect it again to the VNA:


Measurement results in the Range 1-500 Mhz:


We are pretty close to our 50 Ohm impedance. The flatness of the attenuation ranges from -30.01 dB to -28.35 dB. The parasitic capacitance already start to show up a little.

Measurement results in the Range 1-1000 Mhz:

50Ohm_RFBiscuit_30dbAttenuator_1000MhzThe parasitic capacitance are getting stronger and flatness is -30.64 to -28.53dB

I conclude that the attenuator is good enough for my “lowfi” experiments. But lets compare it to a commercial available attenuator (specified up to 3Ghz).

commercial -20dB 3Ghz Attenuator (1-500Mhz):


The impedance is just a spot (the vertical lines that are visible are only the markers). Flatness is: -19.97 dB to -18.52dB

commercial -20dB 3Ghz Attenuator (1-1000Mhz):


The impedance is still spot on – although you can see spurs which are related to the measurement. The spot is a bit broader because the accuracy of the VNA degrades a little above 500Mhz The flatness is: -19.11 dB to 21.78 dB

#rfbiscuit – Use Case 1: Dummyload

In this first example I would like to design a 50Ohm-Dummyload capable to dissipate 10 Watts – the resistor I have chosen for this task is a Bourns PWR263S-35-50R0F – which is specified for 35 Watts (with proper cooling). This resistor is a thick-film resistor, because we want to keep parasitics inductances low and has a tolerance of 1%.


To measure the performance I attached the DIY-Dummyload to a VNWA3E Networkanalyzer


In the frequency range from 1-500Mhz the DIY-Dummyload shows up as a dot in the Smith Diagramm. Exactly what we want.


above 500Mhz the parasitic capacitance of the RF Biscuit starts to show – but depending on your application such a result might be considered still as good-enough.


Let’s compare this result to a commercial 50Ohm-Termination


The following Smith Diagramm shows the 50Ohm-Termination in the Frequencyrange of 1-1000Mhz:


In this Diagramm you can see that the dot is a bit broader – which can be explained as the VNWA3E is less accurate above 500Mhz . You can also see spurs which are releated to the measurement device and which are not part of the characteristics of the 50Ohm Termination.

I will update this part with an evaluation of the thermalperformance of the DIY-Dummyload.

#rfbiscuit – Welcome to RF Biscuit

Hello, World!

This blog is dedicated to research the possibilities of low-cost prototyping of RF gear using the RF Biscuit platform. RF Biscuit is a OSHW project aiming to provide a small, simple, low-cost but yet versatile PCB that can be used to make filters, attenuators, dummyloads and even amplifiers (using gainblocks) the DIY-way. It current features include SMA PCB-edgeconnectors, a prototyping signalpath and the option to use a RF-shielding can.  The creation of the RF Biscuit follows the principle of “release early, release often” and heavily relies on feedback from the community to evolve into an even better protyping platform.

Let me invite you to use and experiment with RF Biscuit – if you are interested to join the circle of Beta-Testers feel free to contact me:


You feedback is highly appriciated 🙂

Thanks, Georg

PS: PCB-Design is done in KiCad and the GitHub-Repo is here.