Expected Teensy 4.1 operating temperature

strud

Well-known member
I have a custom board with numerous peripherals connected to the Teensy 4.1 and some of them being supplied by the on-board 3.3V reg and they are using 80mA (ADC, 16MHz clock, line driver and receiver, opto isolated inputs)

The concern I have is that the steady state of the MCU is in the order of 61degC (as reported by the internal temp sensor) and the 3.3V reg is also very warm to the touch.

The clock is the std 600MHz.

Is this temperature exceptionally high for this clock speed?

Should I be chasing down I/O current draw/sink ie could this be an issue?
 
That seems warm. You might drop to 528 MHz and see if you get the needed performance and measure the temp and current diff.

Normal T_4.1 current is closer to 100 mA - so that temp at 80 mA suggests the board isn't shedding heat as it might - esp. compared to T_4.1 board that is probably smaller?
 
That 80mA is external devices running off the Teensy 3.3V, I have not measured the teensy current draw as yet.

I am using the std Teensy 4.1 board mounted onto my custom board so thermal performance should be normal.
 
It looks as though the teensy is using about 100mA itself initially (total current draw 180mA) but after a little while total current draw on 5V input goes to 265mA.

I'll isolate some functions in the code to see what peripherals might be using this power.
 
So I put some stops in my setup code to try and work out what was chewing up the extra current.

- setup pin mapping and wait for serial port to initialise : 161mA
- allow external ADC to init and configure : 162mA
- initialise SD card lib: 163mA
- init USB lib : 170mA
- begin Ehternet (QNEthernet) 207mA (no ethernet client)
- start up main state machine thread: 224mA
- start remaining threads: 224mA
- IO expansion: 224mA
- enable Quadrature counter: 224mA
- allow loop to run: 234mA (no ethernet client), 283mA (with ethernet client connected)


It looks like Ethernet support is using quite a lot of power, I assume this is normal?
 
Not seen Ethernet power called out before.

283 mA explains the temp of the 3.3V regulator.

Was 528 MHz build tried? It may still have the needed speed and makes a measurable difference in internal processor voltage.
 
Ok tried 528MHz and total current draw is down to 262mA on the 5V feed with ethernet running.

CPU internal temp looks to be down to just under 54degC, down from about 62 so that is significant.

Will be removing pretty much all the external draws from the on-board 3.3V reg.

I've ordered a bunch of little heat sinks too to keep the main processor at a lower temp, will report back on that once they come in (from Aliexpress....)

Maybe I can run the processor at this speed generally and crank it up for short periods when required.... I assume it can be changed on the fly?
 
That's a helpful current drop - and a good/better temp drop.

Heat sinks can help if there is good air transfer - fans are better esp together.

Getting excess waste heat from the 3.3V reg will help.

Hopefully the 528 MHz gives enough speed.

This is available and generally works well:
extern "C" uint32_t set_arm_clock(uint32_t frequency);
 
The Ethernet DP83825I transceiver chip can draw quite a bit of power when active - up to 41mA per the spec sheet.

With a decent heatsink and a little air, you might see 8-10C drop.

Besides what has already been mentioned, you can pull some heat out of the 3.3V regulator by running the 5V VIN input at a lower voltage since the excess voltage is dropped as heat. The VIN can be dropped as low as 3.6V if that is an option for you to consider. If stuck with 5V, a series silicon rectifier diode or two will drop about 0.6V or 1.2V and the waste heat for that portion of the voltage drop will be dissipated in the diode(s) away from the Teensy 4.1.
 
100 to 120 mA is pretty typical for 100 Mbit/sec ethernet. The datasheet spec is somewhat misleading, showing only the power dissipated inside the chip. Power is also dissipated in the magnetics, some of which ends up as the actual transmitted signal.
 
The DP83825I chip has a low energy ethernet mode. So far I've not tried using it. Maybe it's time to try some experiments....
 
As this topic on power consumption might effect me to, I would like to jump in and ask Paul: Those max 250mA on the 3.3V pin is what is available by the voltage regulator after teensy is powered, right? Specs of the regulator talks about a maximum of 1A, but I would like to make sure, that I understand that correctly. My Board (OLED and 69 WS2812B) take up between 20 and 90mA measured on the 5V Input, thats roughly calculated up to 140mA in 3.3V. That would be totally fine inside the specs, right? (I stay under 0.2A on 5V all the time. I could consume more, but even at that +90mA by the RGB LEDS, I need to have them white and way brighter, that I would like to use the device with.)
 
The Ethernet DP83825I transceiver chip can draw quite a bit of power when active - up to 41mA per the spec sheet.

If stuck with 5V, a series silicon rectifier diode or two will drop about 0.6V or 1.2V and the waste heat for that portion of the voltage drop will be dissipated in the diode(s) away from the Teensy 4.1.

Hi KenHahn, That is a great suggestion and easy one for me to implement !

I currently have a Schottky diode in series with the 5V input to prevent any chance of the USB supply feeding the rest of my board, so I can change that to a non-schottky type with a more significant forward drop.
 
Those max 250mA on the 3.3V pin is what is available by the voltage regulator after teensy is powered, right?

Yes, that is the concept, though it's really only a recommendation based on testing at room temperature with free air exposure.

If you continuously draw 250 mA and ethernet is on and a SD card is regularly writing data, you can end up pushing the regulator close to its max temperature. Adding about 100 mA more can cause thermal shutdown. All testing was done on a solderless breadboard with the regulator side facing up. Access to air for cooling make a huge difference. Even a small air flow greatly improves maximum current before risking thermal shut down. Restricting air flow usually means you can't get all this current.

I'm considering reducing the number to 200 mA on future printed cards.


My Board (OLED and 69 WS2812B) take up between 20 and 90mA measured on the 5V Input, thats roughly calculated up to 140mA in 3.3V. That would be totally fine inside the specs, right?

Yes, of course 90 is less than 250.

However, WS2812B LEDs are normally powered by 5 volts, not 3.3 volts, so that wouldn't apply to the 3.3V load.

Also some type of WS2812B consume almost 1 mA per LED when dark (just to power up the controller) and when all 3 colors are at maximum is usually between 12 to 20 mA. Unless you operate the LEDs with only the settings, I'm pretty sure you'll find 69 addressable LEDs end up using more than 90 mA. But again, that would be from the 5V line, not the 3.3V line where the voltage regulator comes into play.
 
Back
Top