Forums

Ambiq "Ultra" Low Power ARM MCUs

Started by rickman January 12, 2016
This part seems to have extremely low power consumption compared to 
other CM4 devices.  They cite 34 uA/MHz using built in buck regulators 
with inductors that appear to be about 0603 size.  When running from 1.8 
volts the current is specified as 51 uA/MHz when I assume the buck 
regulators aren't running or maybe they give less advantage.

The down side seems to be the part only runs at 24 MHz max using an 
internal RC oscillator.  The only crystal is the 32.768 kHz osc.  They 
have means of periodically calibrating the RC against the crystal osc.

Any thoughts?

-- 

Rick
On 1/12/2016 1:46 PM, rickman wrote:
> This part seems to have extremely low power consumption compared to > other CM4 devices. They cite 34 uA/MHz using built in buck regulators > with inductors that appear to be about 0603 size. When running from 1.8 > volts the current is specified as 51 uA/MHz when I assume the buck > regulators aren't running or maybe they give less advantage. > > The down side seems to be the part only runs at 24 MHz max using an > internal RC oscillator. The only crystal is the 32.768 kHz osc. They > have means of periodically calibrating the RC against the crystal osc. > > Any thoughts?
For comparison the other CM4 devices I find are more like 100 uA/MHz but run in the 100 MHz ballpark. I expect this is not using a switching regulator, so I wonder just how much lower the actual current is in the Ambiq parts. I think the Ambiq parts are a bit odd in that they are actually competing against the CM0 parts by having the lowest power consumption and the max clock speed of 24 MHz prevents them from competing effectively against the other CM4 devices which are much faster at the top end. So a CM4F that competes against the CM0s of the world, interesting. While digging around I see that Freescale and NXP are merging or have merged. Lots of overlap in their MCU lines. I wonder how that will shake out. -- Rick
On Tuesday, January 12, 2016 at 10:09:39 PM UTC+1, rickman wrote:
> On 1/12/2016 1:46 PM, rickman wrote: > > This part seems to have extremely low power consumption compared to > > other CM4 devices. They cite 34 uA/MHz using built in buck regulators > > with inductors that appear to be about 0603 size. When running from 1.8 > > volts the current is specified as 51 uA/MHz when I assume the buck > > regulators aren't running or maybe they give less advantage. > > > > The down side seems to be the part only runs at 24 MHz max using an > > internal RC oscillator. The only crystal is the 32.768 kHz osc. They > > have means of periodically calibrating the RC against the crystal osc. > > > > Any thoughts? > > For comparison the other CM4 devices I find are more like 100 uA/MHz but > run in the 100 MHz ballpark. I expect this is not using a switching > regulator, so I wonder just how much lower the actual current is in the > Ambiq parts. > > I think the Ambiq parts are a bit odd in that they are actually > competing against the CM0 parts by having the lowest power consumption > and the max clock speed of 24 MHz prevents them from competing > effectively against the other CM4 devices which are much faster at the > top end. So a CM4F that competes against the CM0s of the world, > interesting. > > While digging around I see that Freescale and NXP are merging or have > merged. Lots of overlap in their MCU lines. I wonder how that will > shake out. >
As far as I can see it's real. Ambiq are running on the linear portion of the transistor, so they do not spend energy saturating the transistor. Really nice idea they have If you look at the EMBCC charts, they have blown the scale, while the others just trying to incrementally outperforming each other But, it's a small company. I wouldn't try to buy from them, to risky Cheers Klaus
> >Any thoughts? > >--
I have no experience on the device at all, but FWIW Jack Ganssle has written two blogs on the device http://www.embedded.com/electronics-blogs/break-points/4441091/Subthreshold-transistors-and-MCUs http://www.embedded.com/electronics-blogs/break-points/4441180/The-Cortex-MPU --------------------------------------- Posted through http://www.EmbeddedRelated.com
On 1/13/2016 8:14 AM, Klaus Kragelund wrote:
> On Tuesday, January 12, 2016 at 10:09:39 PM UTC+1, rickman wrote: >> On 1/12/2016 1:46 PM, rickman wrote: >>> This part seems to have extremely low power consumption compared to >>> other CM4 devices. They cite 34 uA/MHz using built in buck regulators >>> with inductors that appear to be about 0603 size. When running from 1.8 >>> volts the current is specified as 51 uA/MHz when I assume the buck >>> regulators aren't running or maybe they give less advantage. >>> >>> The down side seems to be the part only runs at 24 MHz max using an >>> internal RC oscillator. The only crystal is the 32.768 kHz osc. They >>> have means of periodically calibrating the RC against the crystal osc. >>> >>> Any thoughts? >> >> For comparison the other CM4 devices I find are more like 100 uA/MHz but >> run in the 100 MHz ballpark. I expect this is not using a switching >> regulator, so I wonder just how much lower the actual current is in the >> Ambiq parts. >> >> I think the Ambiq parts are a bit odd in that they are actually >> competing against the CM0 parts by having the lowest power consumption >> and the max clock speed of 24 MHz prevents them from competing >> effectively against the other CM4 devices which are much faster at the >> top end. So a CM4F that competes against the CM0s of the world, >> interesting. >> >> While digging around I see that Freescale and NXP are merging or have >> merged. Lots of overlap in their MCU lines. I wonder how that will >> shake out. >> > > As far as I can see it's real. > > Ambiq are running on the linear portion of the transistor, so they do not spend energy saturating the transistor. Really nice idea they have > > If you look at the EMBCC charts, they have blown the scale, while the others just trying to incrementally outperforming each other > > But, it's a small company. I wouldn't try to buy from them, to risky
I'm not sure they are so much better than the other chips. They are getting a big current improvement from the switcher. So I guess the bare chip actually is about the same current, but because of the lower internal operating voltage they can be more power efficient. So if you don't want the noise of a switching supply in your design this may be an issue. To be honest, we internal supplies dropping below 1.8 volts, I'm surprised other MCUs don't come with an internal switcher. I wonder how much of the chip real estate is for the switch transistor? -- Rick