Micro SD power consumption

I'm working on a low-power logger to continuously log
data at fairly low rates (2-3KB/s) for months. I've successfully
used SD and SD micro cards at lower rates, but I'm seeing
a power dissipation problem with the KB rates. When the
card is really idle, the current draw is well under 1mA.
During and after writes it is about 20mA. (With a 2G
Transcend micro SD).

The simple approach is to buffer up a sector (512 bytes) and
write it as needed. However, when I monitor the power,
I find that it takes only a few milliseconds to write the
data, but the power dissipation in the SD card stays high
for ~250mS after the end of data transfer. This 250mS is,
suspiciously, the same as the write timeout interval
for SD cards. My working hypothesis is that the card
is keeping the internal charge pump and other electronics
turned on until the end of this interval.

Is there a way to command the card to shut down the power
hogs ASAP? I'm using the normal Idle command without
result (or at least a long-delayed result).


I've gotten the average current under 5mA by buffering
up about 16KB before writing---but that requires an
expensive SPI FRAM buffer to supplement my MSP430
controller. It's either that or switch to an ARM
variant with 32KB or more of RAM---and face a whole
different set of power and cost issues.


Mark Borgerson

On May 24, 4:03*am, Mark Borgerson <mborger...@comcast.net> wrote:
> I'm working on a low-power logger to continuously log
> data at fairly low rates (2-3KB/s) for months. *I've successfully
> used SD and SD micro cards at lower rates, but I'm seeing
> a power dissipation problem with the KB rates. *When the
> card is really idle, the current draw is well under 1mA.
> During and after writes it is about 20mA. *(With a 2G
> Transcend micro SD).
>
> The simple approach is to buffer up a sector (512 bytes) and
> write it as needed. *However, *when I monitor the power,
> I find that it takes only a few milliseconds to write the
> data, but the power dissipation in the SD card stays high
> for ~250mS after the end of data transfer. *This 250mS is,
> suspiciously, the same as the write timeout interval
> for SD cards. *My working hypothesis is that the card
> is keeping the internal charge pump and other electronics
> turned on until the end of this interval.
>
> Is there a way to command the card to shut down the power
> hogs ASAP? *I'm using the normal Idle command without
> result (or at least a long-delayed result).
>
> I've gotten the average current under 5mA by buffering
> up about 16KB before writing---but that requires an
> expensive SPI FRAM buffer to supplement my MSP430
> controller. *It's either that or switch to an ARM
> variant with 32KB or more of RAM---and face a whole
> different set of power and cost issues.
>
> Mark Borgerson

Cheaper than a FRAM, might be direct power removal aka a 1.5c
transistor

If you know when the device has finished, then simply 'throw the big
switch', and you know the current
you will get then!

-jg

-jg wrote:
> Cheaper than a FRAM, might be direct power removal aka a 1.5c
> transistor
>
> If you know when the device has finished, then simply 'throw the big
> switch', and you know the current
> you will get then!

that is a practical possibility but success depends on
more than that :
- how can one be sure when to cut the power ?
- if SD init takes 250ms then it's not worth it.
- these parameters will change from model to model
the only way to know is to try :-/

> -jg
yg


--
http://ygdes.com / http://yasep.org

In article <7459c310-2148-4997-bd37-b05639ba9fa7
@y10g2000prc.googlegroups.com>, says...
> On May 24, 4:03*am, Mark Borgerson <mborger...@comcast.net> wrote:
> > I'm working on a low-power logger to continuously log
> > data at fairly low rates (2-3KB/s) for months. *I've successfully
> > used SD and SD micro cards at lower rates, but I'm seeing
> > a power dissipation problem with the KB rates. *When the
> > card is really idle, the current draw is well under 1mA.
> > During and after writes it is about 20mA. *(With a 2G
> > Transcend micro SD).
> >
> > The simple approach is to buffer up a sector (512 bytes) and
> > write it as needed. *However, *when I monitor the power,
> > I find that it takes only a few milliseconds to write the
> > data, but the power dissipation in the SD card stays high
> > for ~250mS after the end of data transfer. *This 250mS is,
> > suspiciously, the same as the write timeout interval
> > for SD cards. *My working hypothesis is that the card
> > is keeping the internal charge pump and other electronics
> > turned on until the end of this interval.
> >
> > Is there a way to command the card to shut down the power
> > hogs ASAP? *I'm using the normal Idle command without
> > result (or at least a long-delayed result).
> >
> > I've gotten the average current under 5mA by buffering
> > up about 16KB before writing---but that requires an
> > expensive SPI FRAM buffer to supplement my MSP430
> > controller. *It's either that or switch to an ARM
> > variant with 32KB or more of RAM---and face a whole
> > different set of power and cost issues.
> >
> > Mark Borgerson
>
> Cheaper than a FRAM, might be direct power removal aka a 1.5c
> transistor
>
> If you know when the device has finished, then simply 'throw the big
> switch', and you know the current
> you will get then!
>
There is some uncertainty about when the device is finished.

The other problem is that a cold start on and SD can require
up to a second. The process for waking up the card and
converting it to SPI mode is something that can take
quite a while.


Mark Borgerson

A possiblity exists in selecting a brand of card that supports an automatic low power mode

The following from San Disk tells of an auto sleep feature.

Quote "Automatic Sleep Mode
A unique feature of cards in the SanDisk SD Card Product Family is automatic entrance and exit from sleep mode. Upon completion of an operation, cards enter sleep mode to conserve power if no further commands are received in less than 5 milliseconds (ms). The host does not have to take any action for this to occur. However, in order to achieve the lowest sleep current, the host needs to shut down its clock to the card. In most systems, cards are in sleep mode except when accessed by the host, thus conserving power.When the host is ready to access a card in sleep mode, any command issued to it will cause it to exit sleep, and respond."

Hope this helps