Teensy 4.1 updated LDO causing power supply problems

F

felixer

Member
We have a board using the MAX17501B switching supply which had been working great with the Teensy 4.1. Recently built another board and installed a new 4.1 manufactured after March 2022 with the new NCV8186AMN330TAG LDO on it. The new LDO seems to trip the MAX17501B's overcurrent detection and ends up hiccupping and never fully turning on. The 4.1 is the only thing being supplied power by this regulator and it works perfectly with the older TLV75733P equipped 4.1s. I have tried slowing down the startup on the MAX17501B regulator and adding 100uF+ of capacitance and haven't had any lucky so far. But it seems like the NCV8186AMN330TAG LDO has a MUCH faster turn on time than the old TLV75733P, ~25uS vs ~1.3ms per their data sheets. Any suggestions? More capacitance?

Per this note on the teensy 4.1 page:
"Teensy 4.1 manufactured after March 2022 has TLV75733P (U4) replaced by NCV8186AMN330TAG"

here is a trace with the slower soft start 1.5ms ish on the MAX17501B
1500us.20ufTeensy.jpg
 
the max17501 is suppling 5v
usb is not connected, but if I connect USB first and THEN put the teensy on the board everything works fine.
there is <50 mA on the teensy's 3v3 rail from external components


I can get the teensy to boot and not drop out the 5v reg if I use a LPF at about 30kHz or less. Both RC and LC implementations worked. This isn't really ideal though and I would rather the Teensy LDO just no freak out the 5v supply.
 
There was an issue on some Teensy 4.0 and 4.1 boards where startup with lower input voltage would fail. On Teensy 4.1, resistor R7 was changed from 100K to 470K to allow the new regulator to start up with low input voltage. With the 100K resistor, the startup problem would happen with less than about 3.8V. The problem was worse with the new regulator on Teensy 4.0, having startup issues sometimes with input voltage as much as 4.1V. Details here:

https://forum.pjrc.com/threads/70636-Teensy-4-0-doesn-t-power-on-with-less-than-4-10V

Maybe the Teensy 4.1 you're using was one of the boards made with the new regulator but before the R7 resistor change?

That resistor is only really needed to keep the power on during the 15-sec flash wipe+restore process. If you just remove it completely and the only loss is attempting the flash wipe by holding the pushbutton for 15 seconds will just reboot (and keep whatever was in the flash memory) rather than perform the flash erase and rewrite with known good LED blink program.
 
felixer said:
the max17501 is suppling 5v
usb is not connected, but if I connect USB first and THEN put the teensy on the board everything works fine.
there is <50 mA on the teensy's 3v3 rail from external components


I can get the teensy to boot and not drop out the 5v reg if I use a LPF at about 30kHz or less. Both RC and LC implementations worked. This isn't really ideal though and I would rather the Teensy LDO just no freak out the 5v supply.
Click to expand...

How much capacitance (how many uF) do you have on the 3V3 rail (so over and above what's already on the Teensy board itself)?
 
The Teensy MCU isn't what is failing to work in this case. It is the 5v supply (mine using the MAX17501) being tripped by the new 3v3 regulator that has been installed since march ish? The MCU never has a chance to turn on since the 5v supply starts hiccupping.
Once I get the Teensy running with either an LPF or by starting it with USB power then plugging it into the board then disconnecting the usb everything works perfectly
 
Any where from 10uF (minimum recommended) up to 200uf is what I have tried. I have also tried a number of smaller caps in parallel thinking the ESR might be a problem but it didn't seem to have any effect. I think maybe if I got insane with the capacitance it might work but it seemed like kind of a bodge since the old LDO equipped 4.1's work just perfect! Same thinking for the LPF which I know works.
 
Here's the startup waveform for the old TI part.

ss3.png

But looking at the waveforms OnSemi publishes, I can't see any sort of soft start behavior.

ss4.png

But it seems the new OnSemi part is *not* using its soft start circuit when the startup is controlled by the enable pin. On at least not a soft start that lasts a millisecond... only a small fraction of 200 us! :(

ss5.png

Not sure what can be done about this, if anything, or really even if this truly is the cause of the problem. Just trying to look at the specs and make some guesses that may or may not really help...
 
Looking at the MAX17501 datasheet briefly, I see it has a "Overcurrent Protection/HICCUP Mode" described on page 13.

Maybe the OnSemi regulator's not-so-soft startup current is triggering Hiccup mode?

Here's a crazy idea... Try adding a resistor, approx 7 ohms, between the MAX17501 output and Teensy's input. Maybe also add an extra 100uF capacitor too, on the Teensy side. During the startup moment, even if Teensy acts like a direct short to GND, the current can't possibly go higher than 714 mA because of the resistor. That's close but not quite to the Hiccup mode trigger level. Maybe use 8 ohms for some safety margin.

Of course, having a resistor isn't ideal. If Teensy draws 100 mA, you'll lose 0.8 volts across the resistor. But Teensy can run on 4.2 volts input, and power dissipated in the resistor would have been dissipated by Teensy's linear regulator, so even if it looks ugly it's not any less efficient. Of course, if you're using a SD card or the ethernet, your power usage might end up higher, probably too high for that resistor. Then you'd need something like a mosfet to short it before you turn on the SD card and ethernet.

But maybe give the 7-8 ohm resistor idea a try (without a SD card and programs which don't turn on the ethernet), just to check whether this theory about OnSemi's not-so-soft start is on the right track or turns out to be just a totally wrong guess.
 
And FWIW, we're planing to eventually switch back to the original TI TLV75733P part. When chip shortages were picking up pace Feb-Mar 2021, Robin bought all the TLV75733P parts we could. Those lasted until early 2022. But eventually we used them all and had to switch to other parts. Only just in the last few weeks has TI finally started offering sample-only quantities of that part. Hopefully sometime in 2023 they'll become readily available again. That TLV75733P part really was perfect, but it's been completely unavailable since March 2021.
 
Oh, also be aware that turning on the USB host port can also cause a brief current spike. It's supposed to be current limited by the TPD3S014 chip, but the limit is 850 mA nominal can the specs allow a pretty wide variance (as is common with current limit on pretty many any load switch chips that don't use a real resistor for current reference).

If you're using USB host, don't forget to test whether the MAX17501 is able to handle the moment power gets delivered to the 100 uF capacitor (C33) and whatever USB devices are connected.

Maybe I should also mention we've used a couple different capacitors for C33.... because more parts shortages. The ones we're using on Teensy 4.1 are rated for 10V and perform somewhat better capacitance-wise when charged up to 5 volts. Older Teensy 4.1 had a 6.3V rated part which worked, but wasn't as good as the C33 capacitor we're using now. If you're right near the edge of triggering that MAX17501 Hiccup circuit with USB host powerup, best to test with a new Teensy 4.1.
 
Last edited:
PaulStoffregen said:
Looking at the MAX17501 datasheet briefly, I see it has a "Overcurrent Protection/HICCUP Mode" described on page 13.

Maybe the OnSemi regulator's not-so-soft startup current is triggering Hiccup mode?

Here's a crazy idea... Try adding a resistor, approx 7 ohms, between the MAX17501 output and Teensy's input. Maybe also add an extra 100uF capacitor too, on the Teensy side. During the startup moment, even if Teensy acts like a direct short to GND, the current can't possibly go higher than 714 mA because of the resistor. That's close but not quite to the Hiccup mode trigger level. Maybe use 8 ohms for some safety margin.

Of course, having a resistor isn't ideal. If Teensy draws 100 mA, you'll lose 0.8 volts across the resistor. But Teensy can run on 4.2 volts input, and power dissipated in the resistor would have been dissipated by Teensy's linear regulator, so even if it looks ugly it's not any less efficient. Of course, if you're using a SD card or the ethernet, your power usage might end up higher, probably too high for that resistor. Then you'd need something like a mosfet to short it before you turn on the SD card and ethernet.

But maybe give the 7-8 ohm resistor idea a try (without a SD card and programs which don't turn on the ethernet), just to check whether this theory about OnSemi's not-so-soft start is on the right track or turns out to be just a totally wrong guess.
Click to expand...

Yeah the low pass filters (LPF) with an RC circuit definitely works (LC too FYI), but I was hoping for a solution that didn't expose the teensy to brown outs just like you mentioned!

I was also disappointed in the "soft start" of the new chip being a fast start. *le sigh* I am talking with Maxim about a potential solution but even when slowing down the turn on time for the MAX pretty dramatically the failure still happens.

We aren't using the USB host, ethernet, or SD card on the teensy at the moment so worse case scenario (when we run out of older 4.1's) we will likely end up going with an RC filter for the teensy power. At minimum going forward all new boards utilizing the teensy will have a 0 ohm resistor and DNP cap spot placed on that board so we can work around LDO problems.
 
PaulStoffregen said:
Are you locked into MAX17501?

A different regulator without that Hiccup mode would probably work much better.
Click to expand...

Not locked in but I love the Maxim stuff because they have the best datasheets, customer support and their stuff seems to require the least external components. They actually respond to emails! But no, we can switch if it really comes to that. Though I do believe the hiccup mode is fairly standard on switching regulators. I have seen it been disable-able on a few though.
 
You must log in or register to reply here.
Back
Top