I'm sending 170 buffered samples from an ESP32 with MPU6050 to ThingSpeak, but the total entries increase inconsistently with a few samples missing in each interval.

I'm using an ESP32 to collect data from an MPU6050 sensor. The ESP32 buffers 170 samples at a time and uploads them to ThingSpeak using the bulk update API.

However, the data entries received on ThingSpeak are inconsistent. Here's the pattern of total entries I see on the channel after each upload:

• After 1st upload: 170 entries (correct)
• After 2nd upload: 338 entries (should be 340)
• After 3rd upload: 499 entries (should be 510)
• After 4th upload: 657 entries (should be 680)
• After 5th upload: 818 entries (should be 850)

As you can see, the expected total should increase by 170 each time, but it's losing a few samples in each batch, and the number of missing entries increases over time. The losses are not consistent — it seems like 1 or 2 samples get dropped randomly in each interval.

Some things I’ve considered:

• The ESP32 is collecting and sending the full buffer (verified via serial debug).
• Could it be a payload size limit, HTTP timeout, or formatting issue?
• Maybe ThingSpeak is silently dropping malformed lines from the bulk update?

Has anyone else run into this with buffered uploads to ThingSpeak? Any tips for debugging the exact cause of dropped samples?

Thanks for any help!

[M Ramzan]

I reduced the upload to 100 samples, but I'm still seeing data loss. Here's the pattern of total entries on ThingSpeak:

100, 200, 298, 396, 496, 595, 695

Only the first two uploads are complete — after that, around 2 samples are lost per batch.

Debug log shows:

Sample 99 - MPU1: 0.03, 1.20, -0.05 | MPU2: -0.08, 0.09, -0.90

Total data length sent: 4299

POST data length: 4359

ThingSpeak Response: 202

Payload is under 8KB and timestamps are ISO 8601 formatted (e.g., 2025-06-17T11:24:33Z). All fields are present and properly formatted.

Any idea why lines might still be dropped?

Thanks!

[Vinod]

I'd recommend logging the free memory on the ESP32 to another channel. I suspect there is a memory leak in the device firmware.

Another thing to try is reduce to 50 samples per bulk update. If my memory leak theory holds, you will see that the number of samples lost per batch start off at zero, but over time you will see more samples lost.

[Christopher Stapels]

Is it possible that there are any time stamps overlapping in sucessive groups?