%(qi-html-head)s %(qi-html-body-top)s

Anelok: power tree and new boost converter

Werner Almesberger werner at almesberger.net
Tue Mar 22 22:09:08 UTC 2016


Wow, almost two months since starting the last thread !

I've started updating the power distribution diagram for the new
design and added more details, including maximum current flows.
This is the current draft:

http://downloads.qi-hardware.com/people/werner/anelok/tmp/power-tree-20160322.pdf

That's when I realized that the boost converter I had carried over
from the CR2032 design (TI TLV61220) wouldn't be able to provide
nearly enough current with the now much lower battery voltage.

Thus, the search began ...


The basic idea is that the regulator should be able to provide a
continuous output (i.e., at 3.3 V) current of at least 100 mA,
preferably 120 mA (i.e., like the LDO in the MCU).

Anelok normally doesn't need all that, but the memory card can
draw 75 mA during 1/4 second, which I'd consider close enough to
"forever" as far as such things are concerned.

The AAA battery has a nominal voltage of 1.5 V but this decreases
with use and load. A reasonable cut-off point would be around 1.0 V.
There can be further voltage losses in the path between battery and
boost converter, e.g., for battery reversal protection. (To be
covered in a future post.) I'm calculating a loss budget of about
100 mV for this, so the input voltage for the boost converter would
be 0.9 V.

AAA batteries can deliver remarkably high current, as this experiment
shows:

http://www.varta-microbattery.com/applications/mb_data/DOCUMENTS/GRAPHS/04103/G_04103_O_S01_en.pdf

So battery current will not be a problem. But we need a boost converter
that can turn 0.9 V into 3.3 V at 120 mA. It should also have not too
unreasonable quiescent consumption, i.e., while Anelok is in standby.
A chip with all these characteristics is surprisingly difficult to
find.


I looked at a lot of data sheets, and I found no more than two chip
families that can do this with a reasonably simple circuit:
TI TPS6102x and Exar/Sipex SP6641B and SP6648.

Data sheets:
http://www.ti.com/lit/ds/symlink/tps61020.pdf
http://www.exar.com/common/content/document.ashx?id=20805
http://www.exar.com/Common/Content/Document.ashx?id=729

First, the various family members:

Part			Voltage		Switch current	Vin(max)
-----------------------	---------------	---------------	--------
TI TPS61020		Adjustable	1.5 A		5.5 V
TI TPS61025		3.3 V		1.5 A		5.5 V
TI TPS61029		Adjustable	1.8 A		6.5 V
Exar/Sipex SP6641B	3.3 V		1 A (inductor)	4.5 V
Exar/Sipex SP6648	Adjustable	2 A (inductor)	4.5 V

Switch/inductor current is much higher than continuous output current
because 1) Iin = Iout * Vout / Vin (long-term average, for an ideal
converter), 2) the converter has some losses, 3) the inductor is both
charged and discharged, so the peak current is well above the average
current.


It seems that, with a regulator that can handle a switch/inductor
current of about 1 A, we're on the safe side:

Part			Vbat	Iout(max)
-----------------------	-------	---------
TI TLV61220		1.0 V	~ 60 mA		for comparison
TI TPS6102x		0.9 V	~250 mA
Exar/Sipex SP6641B	1.0 V	~160 mA
Exar/Sipex SP6648	1.0 V	~130 mA		Iin(peak) = 0.85 A

For some reason Exar/Sipex specify the maximum current of the SP6648
for a peak of 0.85 A, not for the chip's limit of 1.6-2.0 A.

I've included the TLV61220 to show that it's not quite up to the job.


Next, quiescent current. Note that this is not the "off" current of
the regulator. Since the battery voltage is too low to keep an MCU
running, we never turn off the regulator.

When idle, these regulators draw current on both sides, input and
output. I also added the ams AS1310 and the Skyworks AAT1217ICA, to
illustrate the range of quiescent currents one finds:

Part			Iin	Iout
-----------------------	------- -------
TI TLV61220		500 nA	  5 uA
TI TPS6102x		  1 uA	 30 uA	(from figure 10)
Exar/Sipex SP6641B	250 nA	 10 uA
Exar/Sipex SP6648	250 nA	 12 uA
ams AS1310		100 nA	  1 uA	look at this, TI and Exar !
Skyworks AAT1217ICA-3.3	 ?	300 uA	from the CR2032 design

None of the new regulators quite beats the TLV61220, let alone the
ams AS1310 (which unfortunately can only deliver a meager 40 mA) but
they're at least in the general ballpark of the standby current I've
once measured for the rest of the Anelok circuit:

http://downloads.qi-hardware.com/people/werner/anelok/tmp/current-session.png


Now, let's look at the package and the surrounding circuit:

Part number		Package (size)		Circuit
			
-----------------------	-----------------------	----------------------
TI TPS6102xDRCR		10-DFN (3 x 3 mm)	Cout = large, high ESR
Exar/Sipex SP6641BEK	SOT-23-5 (2.8 x 2.6 mm)	Diode
Exar/Sipex SP6648EU	10-MSOP (4.9 x 3 mm)	Cin/out = large, ESR
Exar/Sipex SP6648ER	10-DFN (3 x 3 mm)	=

They're all available in nice and small packages. The SP6641 needs an
external diode while the other chips have it integrated.

TPS6102x and SP6648 both are designed for tantalum capacitors and
require fairly large capacitances (1 x 47 uF for the TPS6102x and even
2 x 47 uF for the SP6648). Not nice.

TI indicate that one could use ceramic capacitors if adding a series
resistor to introduce a bit of ESR (at least 30 mOhm).


Now, let's look at sourcing:

Part number		Price (USD)	Stock
			@1000		DigiKey	Mouser	Newark
-----------------------	---------------	-------	-------	-------
TI TPS61020DRCR		 0.90		13654	 8643	3308
TI TPS61025DRCR		 0.90		 7899	 3747	4670
TI TPS61029DRCR		 0.956		 7909	~1900	  56
Exar/Sipex SP6641BEK*3-3 0.45		  181	 (300)	 158
Exar/Sipex SP6648EU	 0.747		  813	  218	   -
Exar/Sipex SP6648ER	 0.695 at 3000	    0	  432	   -

  (n) = on order, no stock

This doesn't look too good for Exar/Sipex. The TPS6102x is a little
pricey but it should be a safe sourcing choice.


Does this sound reasonable so far ?

- Werner



More information about the discussion mailing list
%(qi-html-body-bottom)s