Threshold logic in CPU's has been around in ARM processors for quite a while. AmbiQ has something similar that's available in boards from Sparkfun. The data sheet from this claims 6uA/MHz: https://www.sparkfun.com/sparkfun-thing-plus-artemis.html
Great to see new entries into western microcontroller development, especially trying to push the envelope in some dimension as this does.
If they get enough eval boards out at decent prices, and the SDK is halfway usable then it would appear some of the lessons from Espressif are being learned.
How is this new entry into western microcontroller development? Everything points towards Upbeat Technology being a Taiwanese based company licensing a RISC-V core from SiFive. Just like companies were previously making MCUs with cores based on designs made by ARM.
The technology is extremely cool, but I struggle to find any real uses for this.
One demo shows a drone flight controller - but any drone has enough motors and actuators to render the power consumption of the MCU irrelevant.
The operating voltage is a flex, but I'm pretty sure no battery chemistry goes this low while retaining a lot of usable power in it. No power harvesting on board, no RF on board, no integrated buck converter. Which doesn't bode well for sensor uses either.
It's cool that an IC this low power, one that can run on voltage this low, is even possible to make. But it looks like a proof of concept piece more than anything.
> One demo shows a drone flight controller - but any drone has enough motors and actuators to render the power consumption of the MCU irrelevant.
Lower power consumption for the brain-box that controls things always means smaller mass, and/or longer flight time. These are both admirable goals, and it is good to take these advancements where we can get them. Even if the improvements may seem like low-hanging fruit, they do accumulate.
> The operating voltage is a flex, but I'm pretty sure no battery chemistry goes this low while retaining a lot of usable power in it.
Supercapacitors are neat. They're capacitors, so they're really good at being charged and discharged many many times. But one of the things that keeps supercaps from being more broadly used is that they can have a good bit of power remaining in them even when their voltage is too low to directly power electronics, which reduces their maximum utilization (and by extension increases their size and expense in application).
This chip is stated to work at voltages as low as 0.4 volts.
This seems like a fine fit for a solar-powered sensor with fairly simple power management that is powered through the night by a supercap. And by avoiding the issues inherent with common chemical batteries, it may be possible for the sensor-widget to have a lifespan that is measured in decades.
> No power harvesting on board, no RF on board, no integrated buck converter.
It's just an MCU (though they do call it an SoC). Lots and lots of useful MCUs that are out there in the real world don't include these functions on-die. That's OK, isn't it?
(Besides, the UP201 is at the first released rev. The featureset is allowed to expand as time progresses. We walk before we run. If the company doesn't go belly-up, I'm sure it will include a kitchen sink and mail reader in due course.)
In a drone, how much the controllers eat is straight up irrelevant. Because you already have to keep 20A worth of motors powered or die trying. You aren't going to make a dent in your power problems by going from 40mA to 1mA on the flight controller MCU.
> Lots and lots of useful MCUs that are out there in the real world don't include these functions on-die. That's OK, isn't it?
No. It's embedded, and chips live and die by integration. You either get really good at being an easy to develop general purpose jack of all trades solution, or you get into a lucrative niche and design everything for the niche.
This appears to do neither. It's a specialized piece, but it's not specialized enough. It suggests some very specific applications, but doesn't deliver on being suited to them. Tech demo material.
> Operating at such low voltages introduces timing challenges, including potential setup time violations. Upbeat addresses these with a second key innovation: a proprietary Error Detection and Correction (EDAC) architecture. This system, which includes a patented special flip-flop design, can catch and correct setup time violations that may occur at near-threshold voltages. This allows for reliable operation without sacrificing efficiency gains.
Wouldn't it be simpler to initialise at a higher voltage and then bring down the voltage after stabilisation? Unless of course the errors are always occurring?
Set up time is needed before every clock edge to ensure reliable state transitions by traditional flip flops. It’s not a one time thing. When you’re operating at low voltages it’s more likely that random system noise will violate the stable level required during the setup time. That’s why they need to do this correction all the time.
Threshold logic in CPU's has been around in ARM processors for quite a while. AmbiQ has something similar that's available in boards from Sparkfun. The data sheet from this claims 6uA/MHz: https://www.sparkfun.com/sparkfun-thing-plus-artemis.html
Great to see new entries into western microcontroller development, especially trying to push the envelope in some dimension as this does.
If they get enough eval boards out at decent prices, and the SDK is halfway usable then it would appear some of the lessons from Espressif are being learned.
How is this new entry into western microcontroller development? Everything points towards Upbeat Technology being a Taiwanese based company licensing a RISC-V core from SiFive. Just like companies were previously making MCUs with cores based on designs made by ARM.
The technology is extremely cool, but I struggle to find any real uses for this.
One demo shows a drone flight controller - but any drone has enough motors and actuators to render the power consumption of the MCU irrelevant.
The operating voltage is a flex, but I'm pretty sure no battery chemistry goes this low while retaining a lot of usable power in it. No power harvesting on board, no RF on board, no integrated buck converter. Which doesn't bode well for sensor uses either.
It's cool that an IC this low power, one that can run on voltage this low, is even possible to make. But it looks like a proof of concept piece more than anything.
> One demo shows a drone flight controller - but any drone has enough motors and actuators to render the power consumption of the MCU irrelevant.
Lower power consumption for the brain-box that controls things always means smaller mass, and/or longer flight time. These are both admirable goals, and it is good to take these advancements where we can get them. Even if the improvements may seem like low-hanging fruit, they do accumulate.
> The operating voltage is a flex, but I'm pretty sure no battery chemistry goes this low while retaining a lot of usable power in it.
Supercapacitors are neat. They're capacitors, so they're really good at being charged and discharged many many times. But one of the things that keeps supercaps from being more broadly used is that they can have a good bit of power remaining in them even when their voltage is too low to directly power electronics, which reduces their maximum utilization (and by extension increases their size and expense in application).
This chip is stated to work at voltages as low as 0.4 volts.
This seems like a fine fit for a solar-powered sensor with fairly simple power management that is powered through the night by a supercap. And by avoiding the issues inherent with common chemical batteries, it may be possible for the sensor-widget to have a lifespan that is measured in decades.
> No power harvesting on board, no RF on board, no integrated buck converter.
It's just an MCU (though they do call it an SoC). Lots and lots of useful MCUs that are out there in the real world don't include these functions on-die. That's OK, isn't it?
(Besides, the UP201 is at the first released rev. The featureset is allowed to expand as time progresses. We walk before we run. If the company doesn't go belly-up, I'm sure it will include a kitchen sink and mail reader in due course.)
In a drone, how much the controllers eat is straight up irrelevant. Because you already have to keep 20A worth of motors powered or die trying. You aren't going to make a dent in your power problems by going from 40mA to 1mA on the flight controller MCU.
> Lots and lots of useful MCUs that are out there in the real world don't include these functions on-die. That's OK, isn't it?
No. It's embedded, and chips live and die by integration. You either get really good at being an easy to develop general purpose jack of all trades solution, or you get into a lucrative niche and design everything for the niche.
This appears to do neither. It's a specialized piece, but it's not specialized enough. It suggests some very specific applications, but doesn't deliver on being suited to them. Tech demo material.
> Operating at such low voltages introduces timing challenges, including potential setup time violations. Upbeat addresses these with a second key innovation: a proprietary Error Detection and Correction (EDAC) architecture. This system, which includes a patented special flip-flop design, can catch and correct setup time violations that may occur at near-threshold voltages. This allows for reliable operation without sacrificing efficiency gains.
Wouldn't it be simpler to initialise at a higher voltage and then bring down the voltage after stabilisation? Unless of course the errors are always occurring?
Set up time is needed before every clock edge to ensure reliable state transitions by traditional flip flops. It’s not a one time thing. When you’re operating at low voltages it’s more likely that random system noise will violate the stable level required during the setup time. That’s why they need to do this correction all the time.