Per the request on another thread, here’s a DIY recipe for doing 96 well microplate normalization to a common volume. A disclaimer – I’m an engineer, not a scientist; I’ve worked on a few normalization projects that have commonality from a mechatronic perspective, but I haven’t paid attention to the actual chemistry being done. Consider it a start point; I’m happy to take it further in this discussion, or refer other people that have experience building complete systems around this concept.
To build a fast system you need a non-contact level sensor. Contact sensing is very time consuming. Anything using light – cameras, laser, or optical sensors can be near impossible to tune because of a wide variety of transparency and reflectivity issues. Ultrasonic sensors seem to be good choice, but not all of them. you want to try any sensor out and make sure it works in your application before you build a system around it. Baumer makes a range specifically for microplate sensing. I’m sure there are others, but this is the only one I know of with a beam focus that can aim down a microplate well.
Next, a dispense system. Where feasible, dispensing from bulk saves a huge amount of time over pipetting. My experience is primarily with solenoid/pressure systems which have the advantage of being adjustable for dispense velocity (based on system pressure and tip size) separate from volume (solenoid time). This can be done with any pressurized flask for the diluent, a regulator and any liquid valve for dispense time. You want very repeatable pressure regulation, and a very repeatable solenoid, ideally one designed around dispensing. You also need a good tip with a well defined edge for good separation.
For a simple system all you really need is a single axis of motion. Alternatively, this wouldn’t be hard to set up on a standard liquid handler, I imagine. Repurposing an old 3d printer could work too. Rigidly mount the sensor at the prescribed level above the plate, the dispense valve tip in line with that. Use the single axis to locate the first well under the sensor and get a read (or multiple reads and average for better accuracy) then move to the next well. Do your math to derive the proper solenoid open time to dispense diluent and run the dispensing in parallel with subsequent level readings. 100ms total per well isn’t unrealistic. The valves/sensors can all be located on 9mm pitch, so its not hard to put 8, or 12 of them in a row for higher throughput.
Typically Festo (my employer) side towards industrial controllers, but we’ve made 3D gantry systems running all this with a raspberry pi and a 3D printer board for motion control for one customer. We did not do the application programming as they wanted to control that for proprietary reasons, but they seemed very happy with the results.