DS18B20: one works, one doesn’t on STM32L071 (both OK on Arduino)

[Jevil]

TL;DR: Two “DS18B20” sensors both work on Arduino, but on my STM32L071xx board only one works; the other never responds. I’m generating 1 µs delays from a 32 MHz clock. Suspecting clone/timing or power-mode differences. Looking for gotchas others have hit on STM32/1-Wire.

Context / setup

• MCU: STM32L071xx, 3.3 V logic

• Bus: single DS18B20 per test (no multi-drop)

• Pull-up: 4.7 kΩ to 3.3 V (external)

• Wiring length: 50 cm

• Power mode: 3-wire

• Timing: delays built from 32 MHz system clock, 1 µs granularity (busy-wait / timer)

What I see

• Arduino (5 V): both sensors enumerate, convert, and read fine.

• STM32L071 (3.3 V): Sensor A works flawlessly; Sensor B shows no presence pulse / no ROM code (or occasionally garbage/CRC fail, depending on timing tweaks).

What I’ve tried

• Verified wiring continuity and pinout (DATA/VDD/GND).

• Switched the STM32 GPIO to open-drain, high-speed; internal pulls disabled.

• Tried stronger pull-ups (3.3 kΩ / 2.2 kΩ).

• Stretched/relaxed timings around spec:

  • Reset low: ~520 µs, sample presence at ~70–80 µs after release

  • Write ‘1’: low 6–10 µs, slot ≥60 µs

  • Write ‘0’: low ~60 µs

  • Read: low 2–5 µs, sample at ~12–18 µs, slot ≥60 µs

• Confirmed idle bus sits at 3.3 V.

I'm really stuck here, and tempted to buy an extra official DS18B20 probe to test my hypothesis.

Any Advice apreciated :)

I can share my code if wanted but i followed this tutorial. (only diffrence is clock speed, but that shouldn't matter)

 

[mfgkw]

Hi Jevil,

 

DS18B20 are known to be faked frequently.

Before wasting much time I would buy some from a reliable source.

 

BR Klaus

[Andrew Neil]

---

@mfgkw wrote:

DS18B20 are known to be faked frequently.

---

Indeed.

Although, if they work with an Arduino, they should work with any other microcontroller: the sensor - even a fake/clone - isn't sensitive to different make & models of microcontroller!

However, it is possible that the specs are different - which might explain why it works at 5V (Arduino), but not 3V3 (STM32).

A genuine DS18B20  is specified to work at 3.0V - but who knows about a fake?

Similarly, the timings may be off...

 

@Jevil again, likely that your STM32 implementation is marginal - so one just works, and the other just fails ...

[Andrew Neil]

---

@Jevil wrote:

Two “DS18B20” sensors

---

Why the quotes? Are you suggesting that they may be fakes?

 

---

@Jevil wrote:

• Stretched/relaxed timings around spec:

  • Reset low: ~520 µs, sample presence at ~70–80 µs after release

  • Write ‘1’: low 6–10 µs, slot ≥60 µs

  • Write ‘0’: low ~60 µs

  • Read: low 2–5 µs, sample at ~12–18 µs, slot ≥60 µs

---

How have you verified that you are actually achieving those timings?

On the 1-Wire^TM bus, timing is critical.

Having one sensor work, and the other not suggests that your timings are marginal...

[Jevil]

---

---

Why the quotes? Are you suggesting that they may be fakes?

---

Yes that is my assumption, but i'm not sure. The DS18B20 that does NOT work is from DFRobot.

 

---

How have you verified that you are actually achieving those timings?

On the 1-Wire^TM bus, timing is critical.

Having one sensor work, and the other not suggests that your timings are marginal...

---

I do not have a logic analyzer on-hand but will buy one. Is this one a good one?

I also know for a fact that if i decrease my clock speed to a lower number i.e. 16MHz instead of 32MHz that the DS18B20 that does work also won't work with the decreased clock speed (TIM Prescaler also lowered ofc).

So i'm also a bit afraid that my HSI is not very accurate thus also having a less accurate microsecond delay. \

I also tested it with the DHT22 and it works flawlessly on both clock speeds (but has less strict timings, if recall correctly)

Thanks!

[Andrew Neil]

---

@Jevil wrote:
I do not have a logic analyzer on-hand but will buy one. Is this one a good one?

---

That should be adequate.

Do you have an oscilloscope? That should be sufficient to see what's happening...

In fact, it's probably worth looking first on an oscilloscope - too see that there aren't any analogue issues...