I'm the blog author. I'm glad you guys are getting a kick out of my write-up! I'll answer a few of the questions/comments here, and I'm happy to answer any other questions.

This is why we read Hacker News :)

I like the audio approach. It should make it possible to measure rotation speed under load. Even a Vitamix slows down a bit when powering through stuff. This might provide a way to compare different blenders in terms of quality. A good blender and a cheap blender might have the same top speed, but the good blender will have more torque and slow down less when meeting resistance.

I’m not sure if you saw it, but the last figure in the post is a spectrogram of the motor under load. I'd say the load results were the most interesting because they revealed differences in the speed control of the different models.

You should also be able to see the speed of the blades (or a multiple depending on how the motor is wound) in the back-emf produced by the motor and fed back out the power cord. I've seen many cases where a sewing machine would produce frequencies that interfered with the horizontal and vertical sync of a television set ... in principle, a blender motor should als be noisy.

The simplest way to test this would be to connect an oscilloscope or spectrum analyzer to the AC power line. This can be dangerous, so I'd recommend capacitive-coupling or for an even safer test, inductive coupling. Capacitors are effective high-pass filters, so picking the correct value will reject the 60Hz power while letting you see the high frequencies produced by the blender.

That reminded me of something I have wondered about for a while, I was playing an Atari video game as a kid and the picture quality massively degraded. After a death I went to mess with the cable and found it unplugged, but the game was still somewhat visible. While I never could reproduce the effect I assumed it was probably going though the power lines and not EM. Any thoughts?

PS: I assume the simple nature of the graphics made a difference but that's about it.

An small air gap is generally capacitive and will pass the higher frequencies while rejecting the lower ones. In the old days (when we only had a few channels on cable television), the CATV systems had a lot more loss and ingress (60Hz from the AC and other signals induced onto the center conductor). To support the higher frequencies now used, the cable has to be carefully installed and maintained.

In the '80's we could get full cable TV by pointing a large Yagi antenna at a particularly strong "leak".

What makes you think that TVs receive video signal through the power line? An apartment I was staying in a few weeks ago had terrible cable signal to the TV, whether or not the coaxial cable was connected, but careful arrangement of the cord produced better signal while unplugged than haphazard arrangement while plugged in. Cables are waveguides. I'm surprised you couldn't reproduce the effect.

This is a great hack and writeup, surprising to see it on a blending blog! I love to watch people explore the solution to a problem without expensive tools or significant experience.

Ah, the audio solution is clever!

As far as optical, RC aircraft peeps use something like this tachometer: https://www.youtube.com/watch?v=9OJ4fx5wwlU

You can also download a spectrograph app for your smartphone and see it update live. I did that a few years ago when I was getting into a debate online about blender speeds.

That's a great idea! Did you find a good app? If so I'd like to hear. I just looked at a few free ones, and they didn't have enough control to be able to read off frequencies in the relevant range very well.

Reading your parent, I decided to (on Android) dl and try "Spectrum View" (there's an SpectrumView for remote control of a spectrum analyzer) and "Spectrogram", both from the same publisher. I would say both are useful and fun, particularly Spectrogram. Nice, barebone apps that are free and no ads, while apparently very well made. (no, I have no affiliation with the publisher).

Actually, I got so mesmerized during my bus trip that I missed my stop :) Fun facts: the brakes have a very sharp 5kHz resonance sound, while the beeper ("I'm closing the doors") were spread out over a few fairly sharp bands. Very fun! When I get home I'm gonna check what the freq contents of my infant daughters laugh and yelling looks like. :)

I use SpectrumView but I have not done an exhaustive search of the space.

There are also a few strobe apps available that use the LED flash on your phone. They work surprisingly well.

How about the obvious solution: photo interrupter.

That's essentially what he was talking about at the start, he avoided it largely because of issues with setup. Putting wires into the blender would subject them to being destroyed during the testing process or require putting a hole in the pitcher which also isn't ideal. The strobing idea is actually the way that you'll usually adjust the timing on a car engine, it's a well documented and tested technique, and fairly easy to figure out when you've got a harmonic frequency.

A small (1/2") round mirror under the blade tip, a pen laser and photo detector would work quite well.

That sound one was really cool. I certainly wouldn't have thought of it.

I worked on a BLDC-powered high-end blender prototype for a company that wanted to get into that market, and when we wanted to measure the speed of the motor, we connected to the controller's serial console. :)

On a related note, blender sound testing is an awful, awful task. You blend a half-pitcher or so of what are basically airsoft pellets, and it is damn loud. I forget the numbers, but I wore two kinds of hearing protection at once and still found it fatiguing.

Ultra low-tech: thread. One could conceivably let thin thread wind up around the blade's spindle during a timed run. Measure the wound string length and work your way back to the RPM of the blade.

Not sure how various materials (say, sewing thread vs thin vinyl fish line) would hold up at 10k RPM, but there might be something that could survive at a proper tension?

EDIT: Reading fail. Someone had already commented on this. Still seems like a interesting challenge to attempt.

From the comments: Use a twine, interleave it around the blades and let it run for a second or two. With a long enough twine, the number of knots divided by the number of seconds gives you the RPS :) Like the thought, but at around 380 RPS, I don't think I will find a twine long enough.. plus the twine will offer a very high resistance to the blade after only a few knots..

But we won't know how long time it takes to wind up to constant speed and without running for a while this will attribute to a fairly large error. From the audacity FFT plots we see that it takes a fairly long time (100's of ms) to get to the target speed (ie the rise time of a step response).

OK what you need here is something called a Sirometer, which is a tool for measuring RPM based on vibration / resonance.

Here is a link: http://www.treysit.com/16.html

The audio measurement is neat, but how is it converted to RPM?

Hz is revolutions per second, so RPM = Hz x 60.

But is a single revolution equal to 1 Hz of audio? What if the blade made two click sounds in a single revolution. Wouldn't that double the pitch, compared to a blade that made a single click?

(Blog author here) If you check out the spectrograms you'll see that there are a lot of different frequencies in the audio. I wasn't sure if the most powerful frequency would be the true rotation rate, so that's why I wanted to have a separate method to check. However, as TheLoneWolfing mentioned, it turns out that the most powerful frequency is indeed once per revolution.

True.

But most of the time the fundamental will be once per revolution.

I thought this was about the program Blender ( http://www.blender.org/ ), but no, this is about measuring the rotational speed of a smoothie-making-blender.

Ok, we'll change the title to be less ambiguous.