Adc noise of the rp2040 is huge. If you have the time switch it to the RP2350. They fixed it on that one. Take a look at the Bus pirate they had a hole thing about it.
With the boost driver it's on all the time, just ramping up and down. For the Convoy driver there's a cool down time. So the only proper fix I can think of is synchronised ADCs.
for v sensing you should reflect on 4 wires kelvin to load and supply, if feasible. I would not trust the resistance measurement of the apparatus with banana connectors. Very thorough testing and good looking data traces congrats!!
Thank you! The resistance calibration isn't ideal, but I don't think 4-wire Kelvin will help in this case since we're measuring a current instead of a resistance.
Will you be releasing the kicad and firmware files? One of my TI drivers had current sense resistors just for monitoring led current as feedback. I wonder if accuracy of sensing could also improve efficiency in closed systems. 2:35 marco reps flashbacks
I think it might be pretty common to have a current sense resistor in the driver. My DIY driver has one too. I don't think the current sense accuracy will help with efficiency, but the value of the resistor certainly has an impact.
The plots are made with matplotlib. It's actually a normal scatter plot, but to reduce the mess of overlapping points, I binned the data and took the medians of each bin.
Yes there's some effort required to get the correct footprints with 3D models. If you're using commonly used parts, then I found componentsearchengine.com to be reasonably good. You'll still need to manually check the footprints and pinouts etc, they're not always 100% correct.
I think if you don't need constant current feature, buying reputable 5V power supply and using the adjust potentiometer to make it 4.2V will just do the work.
Thanks for the suggestion, it looks more affordable. I was also looking at the RD60xx series which are in a similar price range, but I have no idea how good they are.
@@engineerbo The RD60xx are not bad but i really don't like to use switching power supplies for these types of tests. For just a efficiency test it should be fine. It might be able to find the power supply for lower under the chinese name QJ3020E. I like it because it is linear but also high current and has sense lines so i can make sure the voltage is right at the load.
Adc noise of the rp2040 is huge. If you have the time switch it to the RP2350. They fixed it on that one. Take a look at the Bus pirate they had a hole thing about it.
I'm using external ADCs instead of the RP2040 ADCs for this project.
Run calibration data points in pseudorandom order so that error from the last points checked cancels out during regression 😁
Do you mean to make the power supply and boost driver randomly jump around in voltage and brightness? Probably has a non-zero chance of working 😆
@@engineerbo Yeah. I figured, if you are doing one second off time anyway, right? So any concerns about settling time are already out the window.
With the boost driver it's on all the time, just ramping up and down. For the Convoy driver there's a cool down time. So the only proper fix I can think of is synchronised ADCs.
This is a really useful suggestion and something I'll keep in mind for later projects
for v sensing you should reflect on 4 wires kelvin to load and supply, if feasible. I would not trust the resistance measurement of the apparatus with banana connectors.
Very thorough testing and good looking data traces congrats!!
Thank you! The resistance calibration isn't ideal, but I don't think 4-wire Kelvin will help in this case since we're measuring a current instead of a resistance.
Excellent work and explanations! Thank you for sharing your project!
Glad you liked it!
Will you be releasing the kicad and firmware files? One of my TI drivers had current sense resistors just for monitoring led current as feedback. I wonder if accuracy of sensing could also improve efficiency in closed systems.
2:35 marco reps flashbacks
I think it might be pretty common to have a current sense resistor in the driver. My DIY driver has one too. I don't think the current sense accuracy will help with efficiency, but the value of the resistor certainly has an impact.
"why spend 1 hour doing it manually when you can spend 1 month almost automating it?"😂😂
How are your producing those beautiful charts at 6:32?
The plots are made with matplotlib. It's actually a normal scatter plot, but to reduce the mess of overlapping points, I binned the data and took the medians of each bin.
Did you spent a lot of time adding the components in Kicad? Is there a simple way to find components footprint and symbols and add it to your project?
Yes there's some effort required to get the correct footprints with 3D models.
If you're using commonly used parts, then I found componentsearchengine.com to be reasonably good. You'll still need to manually check the footprints and pinouts etc, they're not always 100% correct.
I think if you don't need constant current feature, buying reputable 5V power supply and using the adjust potentiometer to make it 4.2V will just do the work.
What program are you using for your plots, and the 3D scatter plot?
The data plots, including the 3D scatter plot, are made using matplotlib.
An affordable bench power supply I use is the CSI3020X. It is really not that expensive.
Thanks for the suggestion, it looks more affordable. I was also looking at the RD60xx series which are in a similar price range, but I have no idea how good they are.
@@engineerbo The RD60xx are not bad but i really don't like to use switching power supplies for these types of tests. For just a efficiency test it should be fine. It might be able to find the power supply for lower under the chinese name QJ3020E. I like it because it is linear but also high current and has sense lines so i can make sure the voltage is right at the load.
I like linear bench power supplies too. Thanks for the information, I'll look into it!
Are you planning on releasing the source code and schematics?
Nice explanation. Which software do you use to visualize the data in 3D ?🤔
Thank you! The 3D plots are made with matplotlib.
Thank you for your prompt response 😊