Custom board current higher than T4.1

S

slash2

Well-known member
I've been working on a custom board based on the T4.1. During development I have had several issues, but these are mostly resolved. One item though has been really difficult to suss out. The current draw is a good 10mA higher than a stock T4.1. At first I figured it was due to other components I had added, but if I disable them, the problem remains. Additionally, if I look at my board with a FLIR thermal camera, the processor is running noticeably hotter than a stock T4.1 running the same code. In order to reduce the number of variables, I brought out an earlier version of the board that had just the basic Teensy circuitry. I loaded 'Blinky' on that and a T4.1 with the blink rate set to 5 seconds on/off so my voltmeter could provide stable current readings. The T4.1 runs about 102-105mA (off/on), mine runs 112-115 mA. This is consistent over several versions of my design.

The board I'm using to figure this out is a close copy of the T4.1 schematic, using the same parts. Differences:

1. I do not have the Ethernet chip and the T4.1 I have does not have it.

2. My boot chip (from PJRC) is the MKL02Z32 and all the Teensy 4.1s I have use the alternate chip. Could that be the difference, though I don't see heat from the boot chip - just the main processor.

3. I do not connect U1-15 (PTA1) to EMC_01 as I need all EMC0-31 for a port-wide input on my final design. Other threads indicate that is ok and it does work. In any case, my board does have the means to jumper that connection and doing so does not change the current.

4. I had to skip the external memory IO pin setups in setup.c so the EMC pins could be set as inputs, but I use the same code base for both the T4.1 and my board.

5. I do have a QWIIC connector on my board. I did remove the SCL and SDA pullups.

6. I do not have the USB host chip.

7. I do not have the 32Khz oscillator

8. I use only one BAT54 dual diode, as I don't have a battery input and can combine the two arrays into one. I just noticed my schematic is a bit unclear, but pin 1 on the array goes to 3V3.




My schematic is below. The Change notes were from a previous version of my board to the current version.

1771781337912.png
 
Last edited:
Can't read those pins names or designators... Anyway the part number for the processor seems to differ from the T4.1 circuit...
 
Oh sorry, it shows ok on my monitor.

The processors that are installed are identical part numbers though. Good eye on that -fixed. It is probably something that simple.

I'm attaching the Eagle files if that's helpful.
 

Attachments

  • T4test.zip
    69.9 KB · Views: 22
I checked various voltages and they all look similar to the T4.1, except, on my board, NVCC_PLL was 2.52 volts, where on the T4.1 it is 1.11V. I realized I had a 4.7uF cap in parallel with the 0.22. I removed the 4.7 and the voltage dropped to 1.11. Current consumption dropped slightly to 110-114mA, still 8mA higher than the T41.
 
From 1060 Electrical Technical Manual:
In case when high accuracy real time clock are not required system may use internal low frequency ring oscillator. It is recommended to connect RTC_XTALI to GND and keep RTC_XTALO unconnected.
So, maybe this is where some of the mystery current is going (when no RTC crystal).

You could test this by removing RTC crystal and 2 caps from T4.1 and measure current. If current doesn't change, then it's something else. I would expect RTC circuit to be extremely low power (by its very nature).
 
Last edited:
Thanks! I tried it and no change, as you expected. But I'll be sure to add that to the next iteration.

One other difference is my board does not have fuse, ferrite, or FET in the VUSB line. Should not matter, but it is a difference.
 
If your looking for power savings I got rid of the 3.3v linear regualtor and power mine off a 3.3v buck booster TPS63020 (which is supplied by a BQ25798 power management chip) ive had no troubles with power quality

You need too use a couple mosfets to help delay the power on stage. But if you were drawing 100ma with 5v linear regualtor that could drop to 70 to 80ma using switching.
 
Good idea, but in this design, I will be running off a lithium battery (which is why the current difference matters so much). In that case, it’s pretty much a wash in efficiency between the two. I suppose a buck-boost would allow me to drain the battery down past the linear regulator drop out, but there is not a lot of energy left in the battery after that. The regulator in the Teensy has a very small dropout.

I’ll have to keep that in mind though for the future, since, in a 5 volt system, that would make a significant difference.
 
slash2 said:
Good idea, but in this design, I will be running off a lithium battery (which is why the current difference matters so much). In that case, it’s pretty much a wash in efficiency between the two. I suppose a buck-boost would allow me to drain the battery down past the linear regulator drop out, but there is not a lot of energy left in the battery after that. The regulator in the Teensy has a very small dropout.

I’ll have to keep that in mind though for the future, since, in a 5 volt system, that would make a significant difference.
Click to expand...
What temperature range do you need to run over? At room temperature once you get to around 3.5 v the battery is basically dead. But if it's cold you can be well below that and still have over 50% left. So if you need this to operate down close to 0 C (or 32 F in weird units ) then a buck/boost can make a big difference. No help with you main issue but something to keep in mind.
 
Interesting. This will be at room temperature, where I find that the battery ‘knee’ is about 3.2v. So if take it down to about 3.35 or so, the 3.3v regulator can still work and I’ve pulled just about all the energy out. Discharge tests using that criteria pretty much match the battery’s rated mA-hr rating. For example, I did a test on a 70 mA-hr battery yesterday and it ran 33 minutes (to 3.35V) at an average 125mA- which is ~1.8C -a hard use test. That works out to 68mA-hr.
 
slash2 said:
Good idea, but in this design, I will be running off a lithium battery (which is why the current difference matters so much). In that case, it’s pretty much a wash in efficiency between the two. I suppose a buck-boost would allow me to drain the battery down past the linear regulator drop out, but there is not a lot of energy left in the battery after that. The regulator in the Teensy has a very small dropout.

I’ll have to keep that in mind though for the future, since, in a 5 volt system, that would make a significant difference.
Click to expand...

So are you using 18650 cells?

I'm using single cells in parallel (2.8v to 4.2v) using a battery management chip (an easier circuit option is the bq25892) which includes charging as well automatic power path charging and switching from usb and solar. I also use a max17201 fuel guage

The output of the battery management chip goes to a power latching circuit that the microcontroller can turn off when the battery gets low. This goes into an effecient buck booster 3.3v switching converter. This is a lot more effecient than a linear regualtor and gives full range of the battery.

The diy teensy powers straight of this converter at 3.3v, using an n Channel and p channel mosfet to delay the startup (based of the teensy micromod design)
 
Oh no, we are using small circular or rectangular cells. Two cells between 70 and 120 mA-hrs each. They are on the order of a cubic centimeter each. We are severely space and weight constrained, so it's a battle between more complex circuitry and a larger battery. I estimate that simply running the battery into the stock regulator and stopping when it gets to 3.35V, I waste 14% of the battery's power, assuming the battery protection comes in at 3.0V. That is significant, though a boost-buck is probably no better than 95% efficient. So we are talking a 9% difference. Maybe that is worth looking at.
 
It looks like it's the 4.7uH inductor (L1). I used a Murata LQM18PH4R7MFRL to keep the size down as much as possible, but it must have too much series resistance (0.44 ohms). I had an EPCOS T1472K on hand which is a bigger part and half the series resistance and substituted it. The current running the Blink program is now the same as the T4.1. Running my logger program, current dropped from 232 mA to 212 mA.

So, if you are rolling your own board, don't skimp on this part. Wish I knew what the T4.1 uses, as it is pretty compact and, obviously, works well.,
 
I found this from the NXP Hardware Development Guide:

The recommended value for the external inductor is about 4.7u~10uH ，saturation
current >1A，ESR < 0.2Ohm;
 
You must log in or register to reply here.
Back
Top