Prune to Improve Signal-to-Noise Ratio
Thinking like a gardener can help to maximize the SNR in your life
Hi y’all. This week’s newsletter will be shorter than the normal dose since we spent some time this weekend in the hospital welcoming our third kiddo to the world. We’re grateful and thankful for a healthy baby and mom and for the newest addition to our family of five. If you want to be an inaugural contributor to the youngest kid’s college fund, consider becoming a paid subscriber!
There have been a few posts recently that have got me thinking again about one of my favorite concepts in engineering: signal-to-noise ratio (aka SNR). Not only is this one of the most important ideas across engineering but it’s also a great metric to keep in mind when you’re thinking about optimizing processes across the different domains of your life.
But oftentimes we don’t really focus on SNR, we focus on signal. This is a problem. While in the early stages of growth increasing the signal can be the easiest and best way to improve SNR, this often changes once you continue to develop and grab all the low hanging fruit. Once you hit this wall, you’ve got to stop chasing signal and instead refocus on the metric that really matters, signal-to-noise ratio.
That leads me to the thesis for the rest of today’s post: to maximize SNR, you’ve got to focus on suppressing the noise. Often this looks a lot like pruning.
The lure of the numerator
SNR is a very simple and yet powerful idea. Mathematically, it is exactly what the name suggests: the signal divided by the noise.
The reason that engineers and scientists are obsessed with SNR is that it is one of the most important metrics of a system.
Want to know how far away you can be from the nearest cell tower and still access the internet? SNR.
Want to know how well your camera will perform in low-light situations? SNR.
Want to know how accurately your bathroom scale can measure your weight? SNR.
SNR is everywhere and is one of the most important specifications of any measurement system.
Of course, the idea of SNR extends beyond the walls of an engineering laboratory. The concept of SNR is also a helpful way to analyze any type of information transfer.
Did the presentation you gave recently really convey the information you wanted as clearly as possible?
Did that email you sent last week really say what you wanted to say without any additional fluff or distraction?
The Christmas Tree Effect
The problem is that we often get confused between SNR and the signal itself. The first piece that got me thinking about this is from one of my favorite writers,
. David is the author of one of my most-recommended reads over the past few years, Range, and the man behind the excellent Substack newsletter.In a recent post, David shared an idea that he calls “The Christmas Tree Effect.” The name is inspired by a quote from an Army officer. What this Army officer noticed is a common approach across many different disciplines: make things better by adding something more.
[F]or years, every adjustment had involved adding something — like hanging ever more ornaments on a Christmas tree — until the armor outweighed some smaller soldiers.
In this mode, we’re continually putting bandaids on our system to correct for the inadequacies of the system that came before it. This approach looks at SNR and tries to improve it by increasing the signal.
The limits of chasing signal gains
Not a bad strategy, at least at first. Do you want to be able to connect to your cell tower from further away? No problem, just crank up the transmit power. Want to be able to see better in the dark? Just turn on the flash for some more light or increase the exposure time.
But there is only so far that this will take you. The relationship between signal and SNR is a linear one. Double the signal and you double the SNR. Improve it by 10x and your SNR also goes up by 10x. You get the picture.
This means that this strategy often hits a wall, and quickly. You crank up the cell tower and now it’s drawing tons of power. Other constraints begin to quickly limit your ability to just boost the power. It’s the equivalent of adding more and more armor to your soldiers. Now they can hardly walk, forget about deftly maneuvering.
But if we return to our equation, we can see that there is yet another way to improve the SNR: decrease the noise. Cutting the noise in half is mathematically equivalent to doubling the signal. Not only does this get you the same SNR improvement, but you also get your improvement without needing to change anything about your signal. You can just move twice as far away from your cell tower and still stay connected. By reducing the noise you can get the same image quality without needing to change the lighting conditions or lengthen your exposure time.
Addition by subtraction
What I want to double-click on from David’s post is the idea of addition by subtraction. As Leidy Klotz writes in his book Subtract, we are naturally hardwired to try to improve the systems in our lives by adding more. It’s so natural for us to try and solve problems by adding something. But ultimately we run up against constraints like limited time, money, or resources.
On the other hand, it often costs far less to prune. I think the framing of pruning rather than cutting here is important. The idea behind pruning is to remove in order to stimulate more growth. It’s not about cutting just for the sake of cutting. To help illustrate, here are a few examples of pruning outside of engineering systems where SNR can be improved by pruning noise.
1. Making good slides: Jean-luc Doumont
The first one is from my all-time favorite workshop on making good presentation slides from Jean-luc Doumont. Jean-luc has been a big inspiration for my own presentation style and this video is one I point all my students to.
One of the most initially jarring parts of Jean-luc’s slides is how sparse they are. This is not the standard slide deck from the last conference you attended with a logo, institutional affiliation, and website URL on every slide. Instead, Doumont recommends thinking about how to best convey the information you want with the least amount of noise.
Another point of Doumont’s method that I have wholeheartedly adopted in my own practice is using the title of each slide to state the thesis of the slide. Don’t just put something bland like “Results.” That’s noise! Use that precious visual real estate to actually write the conclusion drawn from the results.
It feels strange at first (and especially so because you’ll need to fight against PowerPoint’s default template the whole way), but I guarantee this one tip will make your presentations twice as effective and cost you almost nothing. All you needed to do was prune a bit.
I highly recommend this video from Doumont giving a talk for Stanford’s Center for Teaching and Learning about his process for creating effective slides.
2. Improving life in higher education:
Just last week
published a piece with a similar thesis as Epstein’s:Most things get better when they get smaller.
He highlights three examples (and a bonus) of how pruning helps to create a more effective experience or product.
Simplifying bass guitar lines
Cutting paragraphs from writing
Trimming forms of assessment in a course
Check out Robert’s full post below.
3. Pruning by Blackout:
If you’ve never seen it, you’ve got to check out Austin Kleon’s Newspaper Blackout or any of the blackout poetry he frequently shares on his Substack newsletter. The basic premise is to build a message by subtraction; take a paragraph of text and black out all but a few words to remain for your message.
Prototype for the week: look for opportunities to prune
We’re all guilty of adding another thing to our already full lives. Whether it’s buying more clothes to try and cram into our already packed closets and dressers or adding another “leisure” activity to our schedule to help us unwind, we’re all caught up in the cult of addition.
My challenge to myself and to you this week is to walk around with a pair of hypothetical pruning shears. Let’s look for signs of overgrowth and see how cutting back on the noise can help us to continue to grow.
The Book Nook
The Checklist Manifesto by accomplished surgeon Atul Gawande was a quick but very illuminating read. The central thesis of Dr. Gawande’s book was that the simple technology of a checklist can have a significant impact, even when we can’t imagine that it could. In the book, Dr. Gawande uses a beautiful narrative style to share how he came up with the idea for the checklists he developed, the sources of inspiration, and the results of the studies he and his team ran.
Checklists are a big part of my alternative grading experiment this semester. Essentially every assignment that I grade is a big checklist with two different tiers: proficiency and mastery. After reading Dr. Gawande’s book I have a better idea about why the checklists have been so effective and am even more convinced to try and incorporate more checklists into my practice as a professor.
The Professor Is In
This week is the final lab for MicroPs. In it, students implement a hardware co-processor on their FPGAs that implements the Advanced Encryption Standard. It’s a pretty neat lab and they get practice building a non-trivial algorithm in hardware. If you want to check out more about the lab (or try it yourself) the details and starter code are all available on the course website.
As an aside, this website has a pretty cool illustration/demo of the AES cipher!
Leisure Line
Before there was pizza making, there was coffee roasting. A few photos from a batch of beans I roasted last week.
Still Life
Took the kids to the LA Zoo last Tuesday and had a great time. The baby giraffe and the tiger were two of the highlights.
Josh, that coffee looks amazing. I'm having pourovers here on "vacation" right now!
Your piece is really timely. I want to somehow see a piece that marries this concept with the stuff I just wrote about, horror vacui: https://goatfury.substack.com/p/horror-vacui
Congratulations on the new addition to the family!
Great read. I am writing about presentations at the moment for Breaking the Rules and your section on presentations was great to read. In my years as an academic, I clearly noticed my number_of_slides_per_lecture decrease significantly with years