DC to RF...starting where?
Analog Life, LLC
Presented at CCCamp 2019
Goals for this talk
- Explain how I entered the world of working with RF
- Explain it without burying you in math
- Give you some resources to make your own way
Who are you?
Ham radio experience?
You may have more experience I do!
I wanted this talk to help shortcut some of the struggles I had (and have) when I was getting started with the field of RF.
Who am I?
- 15 years of electronics design.
- Have worked in a couple industries
- Test and measurement
- Industrial controls
- Component sourcing
- Co-host of The Amp Hour electronics podcast (https://theamphour.com)
- Teaching people electronic design online for the past 5 years as part of Contextual Electronics
Started by working on these
10 PLC = 6 readings per second
Not exactly blazing speed
Higher speed stuff like digital was always there as well, but usually nothing very high speed either (<100 MHz)
So what changed?
RF is everywhere
Every product, that is
RF is everywhere AND in every product
And it's cheap
I assume we have similar goals
You’re interested in making things
You're excited about cheap connected (RF) hardware
And you wonder about how to actually integrate it.
How I got started
I copied the app note
This is actually pretty common
I'd guess at least 80% of most PCB designs are re-implemented app notes
Followed other guidelines in Mike Ossmann's "5 Rules of RF design"
(delivered at Supercon 2015)
My first RF Design
- The modem and the antenna were pre-matched
- I minimized my interaction with it (kept the RF trace short)
- Because I didn't need to take it to production
Can we do this all the time?
It depends on your design goals
If you have a completely uncharacterized antenna and source, you are going to have to work a lot harder.
The complexity of your circuit will also have a large impact on the likelihood of success.
Reasons I wanted to learn more about RF
What happens when things go wrong on the bench?
FCC / CE compliance testing
What happens when things go wrong at the $10K/day test lab?
Being a good RF citizen
Your signals might interfere with someone else's signals and that's not nice.
RF concepts that are tricky
Especially for beginners
We're dealing with things in the frequency domain.
When someone talks about "the spectrum" of a signal
They are asking about the frequency content contained within a signal that exists in the real world.
Components act differently based on frequency
- Block frequencies from "passing through" at low frequencies.
- Allow frequencies to "pass through" at high frequencies
- Where this transition happens depends on the capacitance of the component.
- Allow frequencies to "pass through" at low frequencies.
- Block frequencies from "passing through" at high frequencies
- Where this transition happens depends on the inductance of the component.
You're going to see units like "dB", or "dBm" which an easy way to refer to things that change in value by orders of magnitude
As an added bonus:
Gains add together
It's all about the power
Most RF circuits deal in power, not in just voltage or current
Analyzing RF subsystems is often about minimizing the degredation of the signal through the system
This is referred to as the "link budget"
What does the RF path look like?
Image courtesy of osmocom.org
We need to ensure all other parts of the system introduce minimal degradation and noise
If a link budget is 100 dB, that means the received signal is 1/100000th as much power as the transmitted signal
Impedance matching ensures we don't introduce signal degradation
Maximum power transfer happens when a source and a load are perfectly matched
A counter example:
What happens when there is a source (like an ESP32) and an antenna (like a PCB antenna) that are not perfectly matched?
Unmatched source and load
- It may not work at all
- You won't transmit as far as you thought
- Your system will be less efficient
- Energy is "bounced back" and is usually lost
to heat or radiated emissions (bad)
- Energy is "bounced back" and is usually lost
What is impedance though?
Impedance refers to how a device passes or blocks electromagnetic energy at various frequencies.
Why do we need to do matching at all?
Isn't the antenna delivered to work at 50 ohms?
- The world is imperfect!
Environmental conditions can affect it, including things like the enclosure or thing surrounding the antenna
Antenna manufacturing variations means you might have different specs than stated
"Match a 1000-Ω source to a 100-Ω load at frequency (f) of 50 MHz. You desire a bandwidth (BW) of 6 MHz."
This is actually a measurement tool, which plots various measurements
Image courtesy digikey.com
- S11 - Return loss
- S21 - Insertion loss
- S12 - Power transfer from P1 -> P2
- S22 - Reflected power P2 back toP1
Return loss (S11)
Also known as "reflection coefficient"
- This stands for the "Voltage standing wave ratio"
- Measures how well the antenna is impedance matched to the source that is radiating RF energy.
- It is measuring the matched characteristics, not that it necessarily does.
- VSWR defined by the equation below, including gamma (Γ), which is the S11 parameter shown earlier
Image courtesy antenna-theory.net
What "DC" assumptions fall apart at higher frequencies?
A wire is just a wire
PCB material isn't as important as the components on board
A capacitor is there for charge storage
Current can be isolated by ground cuts
Power planes are always a good idea
"A wire is just a wire"
Actually every wire is an inductor
What does this mean for our circuit?
Past 10 MHz, breadboard signal quality falls off
Image courtesy makerspaces.com
"PCBs construction isn't as important as the components placed on that PCB"
4 Layer Stackup
The PCB stackup impacts every piece of your signal pathways, including the impedance/inductance/capacitance of each stripline.
More complex PCB example:
In the GHz range, the small inductance of a PCB trace can have outsized effects on your signal
- Track width: 0.1 mm (3.94 mil)
- Track length: 50 mm (1.97 in)
- Height above GND: 0.2 mm (7.87 mil)
- Track inductance: 63.2 nH
- Impedance @ 2.4GHz: 953 ohm
Ground plane should always be the layer below where your signal is running
Understanding your stackup is critical for treating your PCB as part of your circuit
"A capacitor is there for charge storage"
Image courtesy of Wikipedia
Image courtesy of iFuture Technology
"Current can be isolated by ground cuts"
Noise reduction techniques in low level analog involves cutting the ground plane to stop noise from "leaking" over
At low frequencies, the cutout means that the signals won't "get around" the cut, effectively shielding the isolated areas
High speed signals will treat this area as a capacitor and effectively "travel through" to the other side.
They often radiate energy while doing so. This is how we make PCB antennas (!)
"Current follows the path of least resistance"
The signal actually cares about the path of least impedance.
At DC, this is the path of least resistance.
In higher speed signals, the impedance takes on more complex terms and the inductance starts to matter more.
The magnetic fields between two signals flowing in opposite directions cancel out
This means the inductance will be lowest directly below the signal path and the signal will preferentially flow black on the ground plane
What about Bluetooth, Cellular, Wifi, LoRa, _______, etc, etc?
All of these communication methods are different versions of the same fundamentals
(and most are really brand names)
- Bluetooth - 2.4 GHz
- WiFi - 2.4 GHz and 5 GHz
- 433 MHz (Global)
- 915 MHz (US)
- 868 MHz (Europe)
- 900, 1800 MHz (Europe, Asia)
- 850, 1900 MHz (US)
3G, LTE, 5G
- Various frequencies (see: http://tiny.cc/cellular_freq)
"A Practical Guide To RF And Mixed Signal Printed Circuit Board Layout" - Brendon Parise and Scott Nance
- "RF Circuit Design" - Christopher Bowick 
- "Planar Microwave Engineering" - Thomas Lee 
- RTL-SDR + GNU Radio
- GSG HackRF
- ADI Pluto
Many thanks to Jeff Keyzer (@mightyohm) for helping create these slides!
Thanks to Derek Kozel (@derekkozel) for reviewing this presentation
DC to RF...Starting Where?
By Chris Gammell