Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
STM32 power consumption
1,321 views
Skip to first unread message
Paul Rubin
Nov 9, 2013, 4:15:10 PM11/9/13
to
I'm interested in building a mini gadget that needs a fair amount of
computing speed, and I'm thinking of using this Cortex M4 board, powered
by a small lipo cell:
www.mouser.com/ProductDetail/mikroElektronika/MIKROE-1367
I notice in the data sheet that the maximum power drain is 800 mA at 3.3
volts, i.e. around 2.5 watts. That sounds awfully high--even a fancy
smartphone with a multi-ghz processor doesn't use that much power.
What's going on here? Is that just the amount of power available from
the on board voltage regulator, to poewr off-board stuff? Any idea how
much power a board like this really draws? That said, I don't yet know
much speed I really need. It's a DSP-ish computation that will run
continuously, but let's guess it needs only 50 mhz of the available max
frequency (something like 170 mhz). What's reasonable?
Also, are there particular issues with running a board like this from a
lipo battery (3.5-4.2v, more or less)? The 3-terminal regular seems to
want 5v in....
computing speed, and I'm thinking of using this Cortex M4 board, powered
by a small lipo cell:
www.mouser.com/ProductDetail/mikroElektronika/MIKROE-1367
I notice in the data sheet that the maximum power drain is 800 mA at 3.3
volts, i.e. around 2.5 watts. That sounds awfully high--even a fancy
smartphone with a multi-ghz processor doesn't use that much power.
What's going on here? Is that just the amount of power available from
the on board voltage regulator, to poewr off-board stuff? Any idea how
much power a board like this really draws? That said, I don't yet know
much speed I really need. It's a DSP-ish computation that will run
continuously, but let's guess it needs only 50 mhz of the available max
frequency (something like 170 mhz). What's reasonable?
Also, are there particular issues with running a board like this from a
lipo battery (3.5-4.2v, more or less)? The 3-terminal regular seems to
want 5v in....
Don Y
Nov 9, 2013, 4:30:46 PM11/9/13
to
Hi Paul,
On 11/9/2013 2:15 PM, Paul Rubin wrote:
> I'm interested in building a mini gadget that needs a fair amount of
> computing speed, and I'm thinking of using this Cortex M4 board, powered
> by a small lipo cell:
>
> www.mouser.com/ProductDetail/mikroElektronika/MIKROE-1367
>
> I notice in the data sheet that the maximum power drain is 800 mA at 3.3
> volts, i.e. around 2.5 watts. That sounds awfully high--even a fancy
> smartphone with a multi-ghz processor doesn't use that much power.
No, that doesn't sound right. Most of these little widgett boards are
little more than an MCU on a "carrier" of sorts -- maybe some I/O
connectors. Is yours any different?
Are you sure you aren't seeing a spec that includes inrush current?
E.g., most small MCUs are in the 1/2W ballpark due to thermal
limitations of their packages.
> What's going on here? Is that just the amount of power available from
> the on board voltage regulator, to poewr off-board stuff? Any idea how
> much power a board like this really draws? That said, I don't yet know
> much speed I really need. It's a DSP-ish computation that will run
> continuously, but let's guess it needs only 50 mhz of the available max
> frequency (something like 170 mhz). What's reasonable?
If your board is "just an MCU" (i.e., if you can hand-wave away the
other I/Os that may be on the board as having negligible power
costs -- like switches, etc.), then go to the datasheet for the
MCU. It will give you a better idea what the worst-case power
consumption is (at various temperatures and with specific load
characteristics). More importantly, it will also tell you what
you can TYPICALLY expect to see. The difference will be dramatic
(often a factor of 2 or more; in low power modes, often an
order of magnitude!!!)
Regulator adds some inefficiency (esp if it is a linear -- as is
probably the case on a tiny widget). So, if MCU is 500mW, regulator
might throw away 100mW (depending on input voltage, etc.).
Power consumed in the MCU is related to how many signals are switching
and how fast. So, power at 50MHz should be 1/3 power at 170MHz
ASSUMING THE DEVICE IS SPENDING 100% of its time doing that "same
thing" at both speeds.
Signals switching at the *pins* cost more because they have to drive
"pins" ("C"). Toggling an output pin a X Hz costs far more than
toggling a bit in a register at the same rate!
> Also, are there particular issues with running a board like this from a
> lipo battery (3.5-4.2v, more or less)? The 3-terminal regular seems to
> want 5v in....
Many MCUs will run at ~2V. 4.2V might be high for them without an
LDO regulator in front. Depending on expected battery life and
your cost/size, you might just opt to drop the 4.2 down with a diode
or transistor instead of a real LDO (assuming the I/Os that you
want to drive can also handle the sagging supply.
Any "typical applications" you can review to give you an
idea of how they expect it to be deployed?
On 11/9/2013 2:15 PM, Paul Rubin wrote:
> I'm interested in building a mini gadget that needs a fair amount of
> computing speed, and I'm thinking of using this Cortex M4 board, powered
> by a small lipo cell:
>
> www.mouser.com/ProductDetail/mikroElektronika/MIKROE-1367
>
> I notice in the data sheet that the maximum power drain is 800 mA at 3.3
> volts, i.e. around 2.5 watts. That sounds awfully high--even a fancy
> smartphone with a multi-ghz processor doesn't use that much power.
little more than an MCU on a "carrier" of sorts -- maybe some I/O
connectors. Is yours any different?
Are you sure you aren't seeing a spec that includes inrush current?
E.g., most small MCUs are in the 1/2W ballpark due to thermal
limitations of their packages.
> What's going on here? Is that just the amount of power available from
> the on board voltage regulator, to poewr off-board stuff? Any idea how
> much power a board like this really draws? That said, I don't yet know
> much speed I really need. It's a DSP-ish computation that will run
> continuously, but let's guess it needs only 50 mhz of the available max
> frequency (something like 170 mhz). What's reasonable?
other I/Os that may be on the board as having negligible power
costs -- like switches, etc.), then go to the datasheet for the
MCU. It will give you a better idea what the worst-case power
consumption is (at various temperatures and with specific load
characteristics). More importantly, it will also tell you what
you can TYPICALLY expect to see. The difference will be dramatic
(often a factor of 2 or more; in low power modes, often an
order of magnitude!!!)
Regulator adds some inefficiency (esp if it is a linear -- as is
probably the case on a tiny widget). So, if MCU is 500mW, regulator
might throw away 100mW (depending on input voltage, etc.).
Power consumed in the MCU is related to how many signals are switching
and how fast. So, power at 50MHz should be 1/3 power at 170MHz
ASSUMING THE DEVICE IS SPENDING 100% of its time doing that "same
thing" at both speeds.
Signals switching at the *pins* cost more because they have to drive
"pins" ("C"). Toggling an output pin a X Hz costs far more than
toggling a bit in a register at the same rate!
> Also, are there particular issues with running a board like this from a
> lipo battery (3.5-4.2v, more or less)? The 3-terminal regular seems to
> want 5v in....
LDO regulator in front. Depending on expected battery life and
your cost/size, you might just opt to drop the 4.2 down with a diode
or transistor instead of a real LDO (assuming the I/Os that you
want to drive can also handle the sagging supply.
Any "typical applications" you can review to give you an
idea of how they expect it to be deployed?
Paul Rubin
Nov 9, 2013, 5:02:45 PM11/9/13
to
Don Y <Thi...@not.Me> writes:
> No, that doesn't sound right. Most of these little widgett boards are
> little more than an MCU on a "carrier" of sorts -- maybe some I/O
> connectors. Is yours any different?
The board looks like it has the MCU, a 3-terminal voltage regulator
> No, that doesn't sound right. Most of these little widgett boards are
> little more than an MCU on a "carrier" of sorts -- maybe some I/O
> connectors. Is yours any different?
(don't know if it's LDO), a USB connector, three LED's, and a few
other parts like R's and C's.
> Are you sure you aren't seeing a spec that includes inrush current?
> E.g., most small MCUs are in the 1/2W ballpark due to thermal
> limitations of their packages.
SOC's on boards like the Raspberry Pi are in the 2 watt range (and they
run hot). If this chip uses 0.5W at full speed then (per frequency
scaling as you describe) it's in the 1/6W range at 1/3 speed, which I
guess is acceptable for my purpose though I'd prefer lower.
> you might just opt to drop the 4.2 down with a diode or transistor
> instead of a real LDO (assuming the I/Os that you want to drive can
> also handle the sagging supply.
experiment a little.
Don Y
Nov 9, 2013, 5:09:08 PM11/9/13
to
Hi Paul,
On 11/9/2013 3:02 PM, Paul Rubin wrote:
> Don Y<Thi...@not.Me> writes:
>> No, that doesn't sound right. Most of these little widgett boards are
>> little more than an MCU on a "carrier" of sorts -- maybe some I/O
>> connectors. Is yours any different?
>
> The board looks like it has the MCU, a 3-terminal voltage regulator
> (don't know if it's LDO), a USB connector, three LED's, and a few
> other parts like R's and C's.
OK, I found it at
<http://www.mouser.com/new/mikroelektronika/mikroelektronika-MINIM4STM32/>
Nowadays, LEDs don't take any power (assuming they are just little
status LEDs and not trying to illuminate a room!) Ages ago, you might
dump 20mW into an LED and another 30mW in a series resistor driving
it. Nowadays, I'd put that at 2mW (and only when it is on).
Regulator will dissipate power depending on efficiency. How that
relates to input *current* can vary depending on the type of
regulator used. Looks like the regulator on that board is just
a linear:
"It operates on a 3.3V power supply, and an on-board voltage
regulator allows the board to be powered directly from a USB
cable."
So, that only comes into play when powered from that cable.
It also suggests that the board runs on < 500mA as that's all
you can count on getting from a USB cable!
I see the 800mA reference:
"Power consumption: depends on MCU state (max current into
3.3V pad is 800mA)"
but, can't see anything that would *use* that sort of power!
(also, the USB condition, above)
Schematic (in "User Manual") confirms this assessment. OTOH,
User Manual also repeats the 800mA claim! <frown>
Datasheet for the MCU itself claims 117mA @ 168MHz, 3.6V and 105C
with all peripherals humming away. But *typically* 93mA in those
conditions. And, at 50MHz it looks like 32mA, typ (~60mA at the
extremes cited above)
Of course, if you aren't using certain peripherals, you save.
E.g., same conditions but with all peripherals DISABLED drops you
to the 20mA typ range (almost half the power was dissipated in
those peripherals).
Cost you $30 to find out! :>
>> Are you sure you aren't seeing a spec that includes inrush current?
>
> No idea, that's why I'm asking ;-).
All I can think is inrush current. I.e., if your supply can't deliver
that "for some period of time", the board might not start. <shrug>
>> E.g., most small MCUs are in the 1/2W ballpark due to thermal
>> limitations of their packages.
>
> Hmm, I guess that fits the picture, though I have the impression the
> SOC's on boards like the Raspberry Pi are in the 2 watt range (and they
> run hot). If this chip uses 0.5W at full speed then (per frequency
> scaling as you describe) it's in the 1/6W range at 1/3 speed, which I
> guess is acceptable for my purpose though I'd prefer lower.
AFAIK, rpi has lots of other cruft on board. E.g., it can drive
an HDMI TV! Has *external* RAM, etc.
>> you might just opt to drop the 4.2 down with a diode or transistor
>> instead of a real LDO (assuming the I/Os that you want to drive can
>> also handle the sagging supply.
>
> Thanks, good point. I'll have to check this and I guess I can
> experiment a little.
Remember diodes need current flowing through them to give you a drop!
:>
On 11/9/2013 3:02 PM, Paul Rubin wrote:
> Don Y<Thi...@not.Me> writes:
>> No, that doesn't sound right. Most of these little widgett boards are
>> little more than an MCU on a "carrier" of sorts -- maybe some I/O
>> connectors. Is yours any different?
>
> The board looks like it has the MCU, a 3-terminal voltage regulator
> (don't know if it's LDO), a USB connector, three LED's, and a few
> other parts like R's and C's.
<http://www.mouser.com/new/mikroelektronika/mikroelektronika-MINIM4STM32/>
Nowadays, LEDs don't take any power (assuming they are just little
status LEDs and not trying to illuminate a room!) Ages ago, you might
dump 20mW into an LED and another 30mW in a series resistor driving
it. Nowadays, I'd put that at 2mW (and only when it is on).
Regulator will dissipate power depending on efficiency. How that
relates to input *current* can vary depending on the type of
regulator used. Looks like the regulator on that board is just
a linear:
"It operates on a 3.3V power supply, and an on-board voltage
regulator allows the board to be powered directly from a USB
cable."
So, that only comes into play when powered from that cable.
It also suggests that the board runs on < 500mA as that's all
you can count on getting from a USB cable!
I see the 800mA reference:
"Power consumption: depends on MCU state (max current into
3.3V pad is 800mA)"
but, can't see anything that would *use* that sort of power!
(also, the USB condition, above)
Schematic (in "User Manual") confirms this assessment. OTOH,
User Manual also repeats the 800mA claim! <frown>
Datasheet for the MCU itself claims 117mA @ 168MHz, 3.6V and 105C
with all peripherals humming away. But *typically* 93mA in those
conditions. And, at 50MHz it looks like 32mA, typ (~60mA at the
extremes cited above)
Of course, if you aren't using certain peripherals, you save.
E.g., same conditions but with all peripherals DISABLED drops you
to the 20mA typ range (almost half the power was dissipated in
those peripherals).
Cost you $30 to find out! :>
>> Are you sure you aren't seeing a spec that includes inrush current?
>
> No idea, that's why I'm asking ;-).
that "for some period of time", the board might not start. <shrug>
>> E.g., most small MCUs are in the 1/2W ballpark due to thermal
>> limitations of their packages.
>
> Hmm, I guess that fits the picture, though I have the impression the
> SOC's on boards like the Raspberry Pi are in the 2 watt range (and they
> run hot). If this chip uses 0.5W at full speed then (per frequency
> scaling as you describe) it's in the 1/6W range at 1/3 speed, which I
> guess is acceptable for my purpose though I'd prefer lower.
an HDMI TV! Has *external* RAM, etc.
>> you might just opt to drop the 4.2 down with a diode or transistor
>> instead of a real LDO (assuming the I/Os that you want to drive can
>> also handle the sagging supply.
>
> Thanks, good point. I'll have to check this and I guess I can
> experiment a little.
:>
hamilton
Nov 9, 2013, 5:39:51 PM11/9/13
to
On 11/9/2013 3:09 PM, Don Y wrote:
> Hi Paul,
>
> On 11/9/2013 3:02 PM, Paul Rubin wrote:
>> Don Y<Thi...@not.Me> writes:
>>> No, that doesn't sound right. Most of these little widgett boards are
>>> little more than an MCU on a "carrier" of sorts -- maybe some I/O
>>> connectors. Is yours any different?
>>
>> The board looks like it has the MCU, a 3-terminal voltage regulator
>> (don't know if it's LDO), a USB connector, three LED's, and a few
>> other parts like R's and C's.
>
> OK, I found it at
> <http://www.mouser.com/new/mikroelektronika/mikroelektronika-MINIM4STM32/>
Looking at the pics on this page, I see a SOT89-3 REG1.
> Hi Paul,
>
> On 11/9/2013 3:02 PM, Paul Rubin wrote:
>> Don Y<Thi...@not.Me> writes:
>>> No, that doesn't sound right. Most of these little widgett boards are
>>> little more than an MCU on a "carrier" of sorts -- maybe some I/O
>>> connectors. Is yours any different?
>>
>> The board looks like it has the MCU, a 3-terminal voltage regulator
>> (don't know if it's LDO), a USB connector, three LED's, and a few
>> other parts like R's and C's.
>
> OK, I found it at
> <http://www.mouser.com/new/mikroelektronika/mikroelektronika-MINIM4STM32/>
I'd bet that this reg will handle 800mA before it cuts out.
{searching LD29080DT33]
Schematic shows REG is:
http://www.mouser.com/ProductDetail/STMicroelectronics/LD29080DT33/?qs=sGAEpiMZZMsx3af2HZet3AC0XwQwbUDj
And its an 800mA regulator.
hamilton
Don Y
Nov 9, 2013, 6:46:40 PM11/9/13
to
Hi Hamilton,
On 11/9/2013 3:39 PM, hamilton wrote:
>>> The board looks like it has the MCU, a 3-terminal voltage regulator
>>> (don't know if it's LDO), a USB connector, three LED's, and a few
>>> other parts like R's and C's.
>>
>> OK, I found it at
>> <http://www.mouser.com/new/mikroelektronika/mikroelektronika-MINIM4STM32/>
>
> Looking at the pics on this page, I see a SOT89-3 REG1.
>
> I'd bet that this reg will handle 800mA before it cuts out.
> {searching LD29080DT33]
Yes, but the regulator is *powering* the device when -- and only
when -- it is drawing it's supply from the USB connector. That's
like saying "the circuit that my TV is on is fused at 15A" -- and
claiming the TV *draws* 15A!
Even if a user wanted to take advantage of the on-board regulator
to power OFF BOARD loads (letting the +3.3V "power" pins on the "DIP"
act to SOURCE power), you would still need to know how much of that
regulator output is consumed BY THE BOARD! (and, would have to assume
your USB host could source > 800mA if you were intent on using that
capability off-board.
I.e., there is at least one piece of data missing in order to make
*any* sense of that 800mA spec -- what the board *itself* draws
from a 3.3V supply (imposed on the pins of the DIP) or from a
5V supply (sourced by the USB connector) -- regardless of loads!
On 11/9/2013 3:39 PM, hamilton wrote:
>>> The board looks like it has the MCU, a 3-terminal voltage regulator
>>> (don't know if it's LDO), a USB connector, three LED's, and a few
>>> other parts like R's and C's.
>>
>> OK, I found it at
>> <http://www.mouser.com/new/mikroelektronika/mikroelektronika-MINIM4STM32/>
>
> Looking at the pics on this page, I see a SOT89-3 REG1.
>
> I'd bet that this reg will handle 800mA before it cuts out.
> {searching LD29080DT33]
when -- it is drawing it's supply from the USB connector. That's
like saying "the circuit that my TV is on is fused at 15A" -- and
claiming the TV *draws* 15A!
Even if a user wanted to take advantage of the on-board regulator
to power OFF BOARD loads (letting the +3.3V "power" pins on the "DIP"
act to SOURCE power), you would still need to know how much of that
regulator output is consumed BY THE BOARD! (and, would have to assume
your USB host could source > 800mA if you were intent on using that
capability off-board.
I.e., there is at least one piece of data missing in order to make
*any* sense of that 800mA spec -- what the board *itself* draws
from a 3.3V supply (imposed on the pins of the DIP) or from a
5V supply (sourced by the USB connector) -- regardless of loads!
edward....@gmail.com
Nov 9, 2013, 7:17:13 PM11/9/13
to
> I.e., there is at least one piece of data missing in order to make
> *any* sense of that 800mA spec -- what the board *itself* draws
> from a 3.3V supply (imposed on the pins of the DIP) or from a
> 5V supply (sourced by the USB connector) -- regardless of loads!
Richard Damon
Nov 9, 2013, 10:11:41 PM11/9/13
to
power consumption can be tricky. The rating is probably based on the
maximum usage under worse case conditions, i.e. the processor doing
everything it possibly can. Under lighter loads, it can take a lot less
power.
Looking at your cell phone example, compare the battery life for the
unit just sitting there waiting (but display still on), and the battery
life when playing a fancy game, these can be VASTLY different, and the
difference can be primarily the processor load.
hamilton
Nov 9, 2013, 11:35:26 PM11/9/13
to
This web page states:
"Power consumption: depends on MCU state
(max current into 3.3V pad is 800mA)"
can supply 3.3v to outside circuitry.
Even if the USB port can not, the board can.
If a USB charger is used, then 800mA would be possible.
hamilton
0 new messages