Teensy 4.0 external power burning hot

Status
Not open for further replies.

ghostintranslation

Well-known member
Hi,

I'm working on a complex board for a Eurorack module, that includes a 7805DT power regulator that can take 12v and convert it to 5v. I'm so powering the board and the Teensy 4.0 with this 5v.

Everything on the board works, it runs the program, I can address all chips etc..., but after 10s the surrounding of the power regulator gets burning hot.

If I remove the Teensy from the board though, it doesn't get hot at all.

So I am guessing there is a short somewhere around the regulator but when I mesure voltages, I get the expected 12v or 5v. For example the Teensy 5v pin is getting 5v.

Also when powered via USB it all works fine. (and I'm not trying to have USB + power at the same time here).

Here is the schematics and pcb:
Motherboard rev 6.jpg Motherboard rev 6 pcb.jpg

See the power part on the schematics is on the bottom left, and the regulator is on the bottom right on the pcb.


I don't know how to find the source of this kind of issue so any suggestion would help.

Thanks
 
How much current does the board use ? Try running from 9V, and yes a heat sink might be needed.
 
If your program doesn't need the full 600 MHz speed, try setting Tools > CPU speed lower. Even just dropping down to 528 MHz makes a substantial reduction in current, since that's the threshold where the CPU internally runs at 1.15V rather than 1.25V.
 
You appear to have a DPak or DPak2 linear regulator without any thermal plane? The T4 can easily take 0.25A at full speed,
plus you have other circuitry, so lets estimate 0.3A, so dropping from 12V to 5V at 0.3A = (12-5)*0.3 = 2.1W, indicating
a heatsink is mandatory unless you have a multi-layer board with lots of thermal vias to connect the regulator to some
heat-spreading planes.

Much easier to use a buck regulator and avoid the thermal issues perhaps ?
 
How much current does the board use ? Try running from 9V, and yes a heat sink might be needed.

My multimeter seem not to want to measure current... it's a cheap AstroAI AM33D, when I set it on "200m" it's going over the limit, if I set it to "10" it's reading 0.00... so between 200mA and 10A...
Also even if that would be nice to be able to support anything in between 5V and 35V, I need to focus on 12V as this is the Eurorack standard.

If your program doesn't need the full 600 MHz speed, try setting Tools > CPU speed lower. Even just dropping down to 528 MHz makes a substantial reduction in current, since that's the threshold where the CPU internally runs at 1.15V rather than 1.25V.

Good idea but I tried 528MHz, and even went down to 150MHz in hopes, but it still gets hot enough to burn my fingers.

You appear to have a DPak or DPak2 linear regulator without any thermal plane? The T4 can easily take 0.25A at full speed,
plus you have other circuitry, so lets estimate 0.3A, so dropping from 12V to 5V at 0.3A = (12-5)*0.3 = 2.1W, indicating
a heatsink is mandatory unless you have a multi-layer board with lots of thermal vias to connect the regulator to some
heat-spreading planes.

Much easier to use a buck regulator and avoid the thermal issues perhaps ?

Well I thought the power regulator was a common simple solution to support 12V but also protect from any over powering (since it supports up to 35V) and get a stabilized power.

By buck generator do you mean a DC-DC converter like the TR10? I'm not familiar with those, what's the difference with a voltage regulator like the 7805?

Now I see a few possibilities:
- Keep the regulator in the same SMT package and add vias under it so that it transfers heat and that opens the possibility of adding a small heat sink on the other side
- Keep the regulator but in through hole package so that it doesn't touch the PCB and it's easy to add a heat sink, but that takes up more depth space
- Keep the regulator but add a resistor divider to go from 12V to 9V before to spread the dissipation and keep the advantages of the regulator
- Replace the regulator with a resistor divider to go from 12V to 5V, and add a zener diode maybe to add some little protection? I'm guessing this will heat as much as the regulator?
- DC-DC converter?

Also now that I'm thinking again about this, I would be ok to trust the external power to be stable enough and be no more than 12V, so I would not necessarily need to regulate it but I'd still need to drop from 12V to 5V. I'm not sure if that opens possibilities?
 
DC-DC switching converters repeatedly switch on and off at very high frequency. The output is smoothed by output capacitors. Because they are either on or off they are very efficient but the output can be noisy.

Linear regulators like the 7805 effectively act as a variable resistor, dropping the voltage across them. They dissipate a lot of power and are not very efficient, thought the output tends to be very smooth.

For the most efficiency and smoothest output you could combine both types, using a low drop out linear regulator. I did a search and found this:-LP3872.

So use the DC-DC to drop the 12V to 5.5V and the linear regulator to drop the 5.5V to 5V.

Some will say it's an overkill, but it uses the best of both technologies.
 
You are definitely stressing the regulator. A linear regulator effectively drops the voltage by converting excess energy into heat (slightly over-simplified).

Let's say you are pulling 0.25A at 5V for the Teensy. This is 1.25W (P=IV so 5V * 0.25A). You are feeding 12V into the regular, it will pull the same 0.25A current which means total power delivered to the system is 4W (12V * 0.25A).

This means 1.25W of power is going into the load (Teensy board) and 2.75W is being dissipated as heat in the regular. That's a LOT of heat for a regulator to dissipate. If you feed 9V instead, at 0.25A this would be 2.25W. Since 1.25W goes to load, it would only be 1W dissipated as heat in the regulator. Thus dropping your input voltage from 12V to 9V would reduce regulator heat by 64%!

You can slap a big heat sink on the regular, reduce the input voltage, or replace it with a switching regulator. Note the Teensy already has a 3.3V LDO on board so it's designed to take noisy switching 5V sources like USB.

A switching regular converts voltage by charging up a capacitor periodically (slightly oversimplified). It does generate some heat from various losses but it's not DIRECTLY converting energy to heat like a linear does. So a switching regulator would be about the same heat converting 12V to 5V as it would 20V to 5V, etc. This means that switching regulators have very high efficiencies.
 
You are feeding 12V into the regular, it will pull the same 0.25A current which means total power delivered to the system is 4W (12V * 0.25A).

3W, not 4W, so 1.75W in the regulator, but as I say still far too much without good thermal design (heatsinking).
 
3W, not 4W, so 1.75W in the regulator, but as I say still far too much without good thermal design (heatsinking).

Thanks for checking the math! I'm way too tired today to get simple multiplication right. Agreed, still too hot!
 
DC-DC switching converters repeatedly switch on and off at very high frequency. The output is smoothed by output capacitors. Because they are either on or off they are very efficient but the output can be noisy.

Linear regulators like the 7805 effectively act as a variable resistor, dropping the voltage across them. They dissipate a lot of power and are not very efficient, thought the output tends to be very smooth.

Interesting, I didn't know about those. The down point is that it seems bigger and it require a few more components than just the 2 capacitors of the 7805, which I'm not sure I could fit on my board. But if it can fit that would be an option.

For the most efficiency and smoothest output you could combine both types, using a low drop out linear regulator. I did a search and found this:-LP3872.

So use the DC-DC to drop the 12V to 5.5V and the linear regulator to drop the 5.5V to 5V.

Some will say it's an overkill, but it uses the best of both technologies.

My only concern then would be missing space on the board.

You are definitely stressing the regulator. A linear regulator effectively drops the voltage by converting excess energy into heat (slightly over-simplified).

Let's say you are pulling 0.25A at 5V for the Teensy. This is 1.25W (P=IV so 5V * 0.25A). You are feeding 12V into the regular, it will pull the same 0.25A current which means total power delivered to the system is 4W (12V * 0.25A).

This means 1.25W of power is going into the load (Teensy board) and 2.75W is being dissipated as heat in the regular. That's a LOT of heat for a regulator to dissipate. If you feed 9V instead, at 0.25A this would be 2.25W. Since 1.25W goes to load, it would only be 1W dissipated as heat in the regulator. Thus dropping your input voltage from 12V to 9V would reduce regulator heat by 64%!

You can slap a big heat sink on the regular, reduce the input voltage, or replace it with a switching regulator. Note the Teensy already has a 3.3V LDO on board so it's designed to take noisy switching 5V sources like USB.

A switching regular converts voltage by charging up a capacitor periodically (slightly oversimplified). It does generate some heat from various losses but it's not DIRECTLY converting energy to heat like a linear does. So a switching regulator would be about the same heat converting 12V to 5V as it would 20V to 5V, etc. This means that switching regulators have very high efficiencies.

3W, not 4W, so 1.75W in the regulator, but as I say still far too much without good thermal design (heatsinking).

Thank you for the detailed explanation.

I didn't think about checking junction temperature and power when I did that because it seemed to be really common, but now that I'm looking again at the datasheets I see it says the junction to air is 80 ºC/W, would that mean with 1.75W in the regulator it gets to 140ºC which also exceeds the 125 ºC operating temp?

The datasheet of that precise regulator, I'm using the TO-252-2L package:
https://datasheet.lcsc.com/lcsc/1811052003_Changjiang-Electronics-Tech-CJ-CJ7805-3_C21706.pdf

Now I also noticed that actually maybe a more common package used for that type of circuit is the through hole TO-220-3L, this one:
https://lcsc.com/product-detail/Dro...gjiang-Electronics-Tech-CJ-CJ7805_C77895.html

This one has a 66 ºC/W junction to air which means it would get to 115ºC right? And I could add a heatsink to it but I've seen Eurorack commercial modules not even adding heatsink to it... And I am guessing that I could bend it so it doesn't take too much space in the back...

I'm trying to see if just changing the package would make it better before looking at replacing and redesigning completely this part of the board. Do you think the TO-220-3L would be good to try?
 
You can also look at drop in pin-compatible switching modules to replace LM7805. This will often be the most expensive option but can be the easiest.

Take a look at this article.

http://danielelectronics.com/2016/02/09/testing-a-dc-dc-converter-module/

Very interesting and probably a good alternative to not redesign anything. It is pricy though, $9.41 CAD as opposed to $0.2738 CAD for the 7805.

I guess I have a few options to decide on and try now.

Thanks for the help!
 
I would order one now.
At that price I would want to test it fully before committing.
Buy it now and you will have it when you are ready to test.
I agree that the AliExpress delivery is appalling.
If you can avoid any Cainiao option do so, they are the worst.

Here is another option, combines switching and linear regulators.
 
I would order one now.
At that price I would want to test it fully before committing.
Buy it now and you will have it when you are ready to test.
I agree that the AliExpress delivery is appalling.
If you can avoid any Cainiao option do so, they are the worst.

True, well I'm ready to test now, I only need to unsolder the smt 7805 on the PCB and solder wires to be able to test from a breadboard. Because then I will order a new PCB with the new footprint for that part and that order will cost $100+ and take some time again so I need to make sure this will work before making it.
 
I suspect that you can solder the TO220 style package by bending the two outer pins through 90degrees with the central ground pin bent 90degrees the other way.
Looking at the footprint it looks like it might work.
 
I suspect that you can solder the TO220 style package by bending the two outer pins through 90degrees with the central ground pin bent 90degrees the other way.
Looking at the footprint it looks like it might work.

Yes that one might be fine, but I want to test also the dc converter that is pin compatible, but this one is a SIP package so it wouldn't work directly on that pcb
 
Ok, just a thought.
Good luck with it.
All the switching regulators seem to have a soft start. Something to be aware of.
 
Would a soft start be problematic with Teensy?

I seem to remember reading something here, but I think that was to do with powering through 3.3v only.
I just did a quick search and there certainly was a problem with T3.6 some time ago, but Paul said that he was going to fix that.

Just something to be aware of I guess.
 
So use the DC-DC to drop the 12V to 5.5V and the linear regulator to drop the 5.5V to 5V.

Some will say it's an overkill, but it uses the best of both technologies.

Wouldnt work this was as most linear converters need about 2V Headspace. If you want to go that way first drop to something like 7V.

Besides this i would suspect something wrong on the teensy side, most probably some bad soldering at the pins?
 
Checking the Board Pic from above where are the usual huge Caps for the regulator? I just can spot 2 tiny SMT ones, i doubt they will be the needed 22uF+ ones. 0,1uF will certainly get the LM78 going but wont allow for any significant load.

I use the attached schematics to power 3.2s in most of my modules, having a OLED on the teensys 3,3V rail with no issues regarding noise or temperature (gets barely warm).
PSU schematics.JPG
 
Status
Not open for further replies.
Back
Top