Teensy 4.1 skimpy power supply

Status
Not open for further replies.
D

Deleted User

Guest
Recently I saw some strange behaviour of a Teensy 4.1 with Ethernet, both PSRAMs and MicroSD.
No overclock, nothing else connected to the 3.3V rail.
During heavy data processing and transmission, it sporadically rebooted, without any cause in the code.

The casue was the voltage regulator as I've found with a transient recorder and a FLIR! It began to build up heat very slowly and suddenly switched off the Teensy 4.1 for a fraction of a second, before starting up again for a varying span of time, minutes to about an hour, before it turned off again. When the code is more idle or there is less network traffic or less SD operatios, this does not happen.

When it happens during write, the SD filesystem gets shattered. This is with different types of SD, all real SanDisk, so the SD is not the fault. There is not enough buffering capacity to allow the SD to finish block operations, before power dies.

So the voltage regulator is at it's full maximum, or even beyond, if Teensy 4.1 is equipped with all it's possible features and working, even without any peripheral connected to 3.3V.

I suggest to design a better voltage regulator on the next Teensy!
 
Although I have zero mileage / experience using the T4.x I would suggest adding an electrolytic capacitor to the regulated side of the Teensy power grid to reduce the effects of the power transient to see if it lessens the impact. A heat sink on the processor and/or or voltage regulator may help too if it is a thermal event. You may also try underclocking based on processor temperature to help avoid these events too.

Did you try an external power supply different from the USB power that may be limited in ability to supply the power demand being placed on your setup?
 
:rolleyes:
Thank you for advice.
As you might imagine, I have made a workaround by adding a better regulator and a larger capacitor for the SD.

An unmodified Teensy 4.1 has this problem and it could "surprise" users. I think it is a failed design, because the "engineer" did not take into account the power demands of the possible on-board addons.
 
Was USB the power supply to the T_4.1? It can only supply 500 mA? - and the onboard regulator IIRC can convert more than that to 3.3V - but heat dissipation is pushed beyond that.

Was the current drawn by the T_4.1 monitored?
 
did you try a USB3 port which provides 1.5A of current?

provide a demo code that shows the behaviour so we can try to replicate it
 
I ran into a similar problem when using a RA8875/6 displays. Solution was to use a Powered USB3 Hub and no issues. As @tonton81 mentioned did you try a USB3 port?
 
@Frank B
If you were able to read the datasheets, the one of the TLV757P and the DRV package, then compare this to the actual board, you would get the idea, the thermal dissipation is less than optimal, so reducing maximum current quite a lot. Plus, there are transients challenging the TLV757P even more, as they bring overcurrent protrection close to the trigger wall.
But yeah, I am the stupid. Just to please you.

Fact: Design is suboptimal. Placing thermal vias on the side of a package, using a 35µm (at best) copper layer to conduct heat to the thermal vias over a distance of more than 1mm is plain [...], at best.
Max current will so not be possible.
Fact: Buffering caps to MicroSD are too far away from the card and too small. See specs.

Actually I do not plan to respond anymore to your constant bashings, Frank. My electronics explore the universe. Since decades. In Space. Perfectly functional and reliable up to now. So do not tell me anything!
I am not going to post thermal scans of a hot TLV75733P on a Teensy 4.1, not posting recorder streams of 3.3V ditching off while 5V rock solid. Water to the sea...
 
You could tell the others which part (with less thermal dissipation, perhaps?) you used or how you fixed it.
I promise, I will not read it, and don't intend to answer.
 
I don't see anyone here bashing, Frank just asked what you changed it with, no need to throw heat his way, people are here to help each other at both software and hardware levels.
 
As a first very quick test, I connected a 10 ohm resistor between 3.3V and GND and ran just a simple LED blink program which counts, so I can see if the board has kept running without any reboots.

I also connected a power supply to VIN & GND and set it to 5.05V. I used a multimeter to check that Teensy 4.1 really had just over 5V power. Here's the setup.

sc.jpg

When I disconnect the USB cable, my power supply says the total current into VIN is 444 or 447 mA as the LED turns on and off. The resistor is very hot to touch, which is a little surprising since it's rated for 10 watts and this is just over 1 watt. The TLV75733P on Teensy 4.1 also feel pretty hot. So far I've done no accurate temperature test, only just this very quick & dirty test with a subjective finger touch.

So far it's been running for nearly 20 minutes.

sc.png

Just to be thorough, here's the code running:

Code:
void setup() {
  pinMode(13, OUTPUT);
}

void loop() {
  static unsigned int count = 0;
  Serial.printf(" count = %d\n", count++);
  digitalWrite(13, HIGH);
  delay(500);
  digitalWrite(13, LOW);
  delay(500);
}

@Deleted User - I hope we can discuss this problem in a calm & civil manner? I do take reports of problems seriously, as you've hopefully seen across so many threads on this forum. You've probably also aware we have a general expectation about posting details to reproduce problems. I believe this message demonstrates Teensy 4.1 is running successfully with 330 mA extra current beyond the base level to run at 600 MHz without sleeping, but also without many peripherals active. At this point, if you're willing to continue talking here, I'd like to ask you for more detail about how to reproduce the problem you encountered?

Meanwhile, I will continue letting this simple test run. I'll also look for another resistor & order parts if needed. While not a highly dynamic load, hopefully this very simple test does at least show thermal performance in a way we can reproduce.
 
Quick update. I found another 10 ohm resistor. Connecting both to 3.3V (for ~660 mA extra load) definitely causes the TLV75733P to go into thermal shutdown after just a few seconds.

I also found a 22 ohm 10W resistor. I've just started another test with both 10 & 22 ohms (for ~480 mA extra load). I measured the voltage at each resistor. Both are at 3.26 volts. With the USB unplugged, my power supply says the total current into VIN is 591 & 594 mA.

So far this test has been running for about 10 minutes without a reboot. Will leave it running overnight....

DSC_1354_web.jpg
 
The test with 480 mA extra current (a 10 ohm and 22 ohm resistor) has been running continuously for over 6 hours.

At this point, I'm going to turn this off and move on to other more urgent matters, since I just don't have a large workbench where I can leave a test like this running without interfering with my ability to do other work.

@Deleted User - I am willing to do more testing, but I really want you to provide more detail of how you tested. Please, if you comment again on this thread or any others, try to keep it constructive and maybe dial the hostile tone down just a bit? You've spent a lot of effort writing about your disapproval of how Teensy is designed. I really do appreciate all honest feedback, even when it's unpleasant. But I hope you can understand how much more constructive this sort of feedback would be if you could put some of that effort into documenting how to tested and giving me more details about how to reproduce the problem.

sc.png
 
The problem seem to be transient loads triggering TLV75733P internal protection. With a constand resistive overload, the Teensy just turns off, stays so for a few seconds and turns on again.
With your test setting, my Teensy 4.1 turned off after 10 minutes.

The problem I found and referenced to in OP is, it turns of for less than a second, which I believe is due to a combination of thermal protection and current sensing. Heat would not be dissipated so quickly in this setting. I do not know the internals of the TLV75733P so any guess here is highly speculative.

Adding a 4,7µF to the MicroSD slot did remedy enough. This for me is proof it is the transient problem, not resisitve continous load.

But for safety I've added a LF33CDT I had in my junkbox with a transistor and resistor as inverter. 33µF low ESR at the output, 4.7µF (standard, non low-ESR for inrush limitation) at the input. A patch wire from EN, 5V, 3V that is all. So it follows the power-on sequence requirements and it has a generous cooling tab for heat dissipation. Ugly, but solid.

Anyway, over time I got allergic to Franks constant counteractions, also on other posts.
He may be a master of C.
I make hardware. Definitely, am not a coder/programmer of C.
Machine code is what I do normally, a dinosaur I am.
So better I leave this here.
Sorry to have caused any inconvenience.
 
Last edited:
Can you say which SD card you're using? If you're not going to show us any code to reproduce the problem, maybe at least describe what your code is doing with that card?
 
Administrative update: This user requested to be deleted. Robin & I did not decide to delete their account. It was a clear request sent to us by private email. We handle delete requests by deleting all personal information.

Our general policy is to avoid censorship. We don't usually delete messages. Messages this person wrote remain on the forum as "Deleted User".

On the tech side, I believe the 480 mA steady current test shows we do have a good amount of extra current capability on the 3.3V power on Teensy 4.1. I'm going to do a little more investigation over time, probably with a thermal camera. I might also try some special high speed write tests with various SD cards. But by general feeling is we're not facing a dire problem as the alarmist tone of these messages suggests.

I'm closing this thread now.
 
Status
Not open for further replies.
Back
Top