Two HC-SR04s side by side but they look totally different - which one would you use?

I've used both kinds on a rain barrel project and honestly the plain one has been the workhorse for me too. Mine's been sitting on top of a barrel in the garage measuring a tank about half full to nearly full for a couple of years now, reading through NewPing into an Uno, and it barely drifts. I check it against a dipstick now and then and it's within a centimeter or so. I did buy one of the multi-mode boards out of curiosity and got it working over the serial pins with a small sketch I found, but it felt like extra fuss for no real gain on a simple level reading. The metal can on the older boards is just part of the timing circuitry, so its absence isn't a fault, the newer chip does that job differently. One thing I'd say for a water tank is that whichever board you pick, mount it dead level and keep the face of the transducers clean of condensation. Condensation on the sensor face was the only thing that ever gave me nonsense readings.

I went the other way with my rainwater cistern and used the dual-mode board over I2C with an ESP32, and it's been solid. My reasoning was that with the trigger and echo style, the timing has to happen on the Arduino side, so my sketch sits there waiting for the pulse to come back. With the serial and I2C version, the little chip on the board does all that internally and just hands you a finished distance reading. That meant my loop kept running the pump logic and the display updates without any awkward waiting around.

The annoying part was documentation. The listing had almost nothing useful, so I ended up finding a datasheet on a Chinese hobbyist site and running it through a translator to figure out the address bytes and the command format. Took an evening of fiddling, whereas the plain sensor would have been running in ten minutes with NewPing. So if your project is just a level readout on an Uno, the standard one really is the path of least resistance.

As for the missing metal part, my understanding is that on the dual-mode boards the onboard micro replaces the timing circuitry the old design needed, so it's not a cheaper build, it's a different design doing the same job internally.

One thing I'd check before committing either way: make sure whatever library you pick handles the max distance sensibly for a tank, because the stock example sketches assume small distances and will give you garbage on a deep tank unless you set the timeout properly.

I was looking at these and I'm wondering if my use case would be too complicated for the standard HC-SR04. I'm planning on using them to measure the distance to an automated feeder while it's actually dispensing food, meaning the distance is constantly shifting. I'm worried that the simple trigger and echo method might introduce too much measurement delay when the target is moving so quickly. I read that the multi-mode chip does the timing inside the sensor, wich is cool. Does having that internal timing help with latency when you're dealing with rapid, changing distance measurements, or is that feature only meant for slow, stable readings like a static water tank? I feel like I might need the I2C option for my setup, but the wiring seems trickier.

I agree with the point that the newer chips handle the timing internally, and that really does simplify things compared to managing a pulse on an Arduino. When I tested a similar setup for a workshop project, where I was tracking the height of a moving pile of wood, the multi-mode board was defintely easier to wire up and code for continuous polling. I found that the power consumption draw seemed marginally lower on those serial setups when they were just waiting for a measurement compared to the trigger and echo pins which seemed to keep cycling the clock even when idle.

Honestly, for a static tank like you are measuring, the difference in chip logic isn't going to save you from a bad install. The standard setup is perfectly reliable, provided you respect the acoustics. I ran a test once on a propane tank setup, and I found that if the sensor is mounted less than 15 centimeters above the liquid surface, the early reflections cause a consistent error of at least 8 centimeters. You need clearance for the sound to travel and bounce back clean. The multi-mode chips don't magically fix poor mounting geometry. Focus on the standoff height.

I'm skeptical about that 15 centimeter limit for mounting height. While early reflections definitely happen, I've found that the error isn't as predictable as a fixed boundary. It seems to depend much more on the surface quality of the liquid and how much ambient sound is around the setup. Sometimes the sensor just manages to filter out those initial weak echoes even when placed close to the water line. I think the sensor's internal filtering does a lot of the heavy lifting there.