Not up-to-date assembly diagram and hardware

[Chameleon0ff]

Hello! I want to assemble HDS for home use. I already have ESP32-S3 N16, OLED SH1116, MPU6050, HX711 with 2kg loadcell, TTP223 touch button x2. The firmware loads normally, but when I try to turn on the scales in the serial port, an ADS1115 initialization error appears. I also have ADS1115 at work, but I don't understand where to connect it.

Can you please update the connection diagram for self-assembly or provide pins for connection?

[decentjohn]

@Sofronio can you assist, fyi @tadelv

[Chameleon0ff]

Okay, I managed to compile and test my HDS a bit. I configured config.h (uncommented lines 17 and 20) and now everything is great!
I made a diagram with connection to ESP32-S3, but I'm not sure if it will work for everyone

[Sofronio]

@Chameleon0ff I highly suggest using an ADS1115 (or another external ADC) for battery voltage monitoring. The ESP32 series has poor onboard ADCs—each one behaves differently, they’re not very linear, and they’re difficult to calibrate accurately for battery voltage measurement. When I developed the original Open Scale, the ATmega chips had much better ADCs, so this wasn’t a problem back then.
With that said, to support earlier versions, I added automatic ADS1115 support. You just need to remove the #define ADS1115ADC line for the code to ignore it. There’s also a flag called b_ads1115InitFail that will disable the ADC if initialization fails.

[Chameleon0ff]

I use ADS1115 in my assembly. The charge value is not very stable, but maybe it is related to the old battery that I pulled out of the broken scales
I also had to short GPIO10 and EN so that the scales could come out of sleep mode. Connecting the 7th leg of TP4056 I can bring it out of sleep mode only after a significant change in potential (for example, when connecting the charging cable). Therefore, I decided to do without indicating the charging status
Also, my scales have a significant weight drift - about 30g/min without a load on the scales and more than 70g/min with a weight on the platform. At first, I tried to transfer the HX711 to 80Hz mode, but the values were very unstable (the values changed several dozen times within a second), although there was no such drift. I decided to return to 10Hz so that the values were stable, but drift appeared

[Sofronio]

For measuring the battery level, you can also try adding a small capacitor across the battery measurement line to smooth out the noise—it often helps stabilize the readings.
Additionally, the TP4056’s EN pin can be wired through an NPN transistor and controlled by an ESP32 GPIO, allowing charging to be temporarily disabled so the real battery level can be measured. This setup has already been implemented in version 3.0.1.
Weight-drift compensation was introduced in version 3.0.2. I recommend trying it on the develop branch. You can fine-tune the behavior using TRACKING_THRESHOLD and STABLE_OUTPUT_THRESHOLD. On a larger scale with a much beefier load cell, I had significant fluctuations; setting both parameters to 0.3 produced very stable results.
And finally, 80 Hz has never been ideal—the noise is much higher than at 10 Hz.

[tadelv]

@Sofronio would posting the new schematics help us close this ticket?