Inside the Propeller
Humanoido
Posts: 5,770
Is there anything inside the Propeller chip
that can act like a sensor?
(or a technique to program a sensor inside the chip?)
No, not sense the outside world, but sense the inside of the chip.
The project is to crudely image the inside of the chip,
by mapping electrical fields, or other method
with parts and code inside the chip.
Thanks sincerely
that can act like a sensor?
(or a technique to program a sensor inside the chip?)
No, not sense the outside world, but sense the inside of the chip.
The project is to crudely image the inside of the chip,
by mapping electrical fields, or other method
with parts and code inside the chip.
Thanks sincerely
Comments
The answer to the first question is maybe or "theoretically, but not in any meaningful practical way". The various sections of the chip and their wiring on the chip have capacitance and inductance and this can be measured in a fashion. Semiconductors can detect light and other forms of radiation and the difference between a memory array and a camera chip is more one of intent than structure. That said ...
The answer to the second question is NO. There's a lot going on inside the chip. Everything it does involves moving charges around and that will affect how much noise and other artifacts there might be on any attempt to measure things on the chip. The fact that there are things inside the chip that theoretically might act as sensors doesn't mean that they can be used in any meaningful way as such.
Even though you can use RAM arrays as crude photo sensors, the chip package is designed to prevent this from happening. You'd have to grind or chemically etch the package material off the chip to expose a portion of the RAM array to light to use it. You can't just take off the whole package, because then you couldn't have a program stay in un-altered form in memory to process the rest of the memory contents.
I don't know how the metallization was done over the memory arrays. It could interfere with this effect. Remember that the chip layout is sometimes done specifically to prevent portions of the chip from reacting to electrical fields and ionizing radiation since that's not something you want to happen when you can avoid it.
I believe that's how Micron's CMOS imaging division got its start. The very first imager I worked with in the early '80s included a converted CMOS RAM chip as its sensor. It wasn't very good, though, exhibiting a wide range in sensitivity from pixel to pixel.
-Phil
CON _clkmode = xtal1 + pll16x _xinfreq = 5_000_000 PIN = 0 PUB start | time, f0, f1 ctra := %00010 << 26 | PIN frqa := $8000_0000 dira[PIN]~~ ctrb := %01010 << 26 | PIN frqb := 1 repeat time := cnt waitcnt(time += clkfreq/ 1000) f0 := phsb waitcnt(time+= clkfreq) f1 := phsb f1 -= f0 '.debug "Frequency is: ", dec f1, #CRHere is some sample output:
-Phil
That's brilliant. You have just turned an 8 core 32 bit processor into a thermometer:)
Well, given some calibration this may well be useful sometime.
One thing I can directly can see it can be used to is --- Correction of RTC's controlled by Propeller frequency
Could that work as a thermal-protection feature? The Prop can sense it's about to over heat, so it shuts down something or turns on a fan???
What I mean is... it takes approximately 30kV per centimeter to break down air for a spark or a short circuit to occur. Scaling that down to chip level, that's about 3V per micron. In order to get past Moore's law, great efforts have been made to introduce dielectric materials that increase the permeability between adjacent metal wires and layers at the sub-micron level. The act of decapping the chip and exposing it to air could disrupt the permeability.
CON _clkmode = xtal1 + pll16x _xinfreq = 5_000_000 PIN = 0 COGS = 1 PUB start | time, f0, f1 ctra := %00010 << 26 | PIN frqa := $8000_0000 dira[PIN]~~ ctrb := %01010 << 26 | PIN frqb := 1 repeat COGS - 1 cognew(@heater, 0) repeat time := cnt waitcnt(time += clkfreq/ 1000) f0 := phsb waitcnt(time+= clkfreq) f1 := phsb f1 -= f0 '.debug "Frequency is: ", dec f1, #CR DAT heater jmp #heaterHere are the results:
Because '.debug uses PST and an extra cog, I had to obtain the result for one cog from the spectrum analyzer without '.debug being active. Anyway, the results are as expected: the more that's going on in the chip, the warmer the die.
-Phil
One other thing worth noting is that the Prop has 16 independent VCOs spread across one edge of the chip (PLLA & PLLB below):
So it should be possible to measure temperature gradients across the die or localized heating. Of course, some prior calibration would be required, since the VCOs are sure to have different offsets and gains vs. temperature.
-Phil
Permittivity surely, not permeability? Most of these dielectrics are oxides and very refractory, zirconium and hafnium dioxides and silicates should resist atmospheric oxygen I think
[Also the gate dielectric is at the bottom of the pile and not exposed at all ??]
"Also the gate dielectric is at the bottom of the pile and not exposed at all" ... right but even using a 180nm process, the metal to metal distance is .28um ... A spark gap of 840mV in regular air under 1 atmosphere of pressure.
"The idea is to drive the PLL with a high-enough frequency that the VCO is guaranteed to operate at its free-running frequency, then divide that frequency down so that it can be measured with another counter. The result of counting for 1ms is a number...".
Do you indeed have an external hardware divider in there? The frequencies reported with the 1ms gate are around 1.8MHz.
One thing that generates heat on the die is operating the pin at threshold, as in a sigma-delta converter. So a neighboring counter module used as your temperature sensor should be able to pick that up.
How a PLL/VCO free-runs also depends on the Phase Comparator - they do not all nicely saturate, but may 'hunt' but fail to lock.
Would this also work at the other end of the range ? (that is likely to consume less power itself, and also be more MHz-legal to mesaure) :
The data quotes 64-128MHz as valid for the PLL.VCO, so if you asked for something well under 64MHz would that free-run, or try to lock on some edges, or ?
A free-running PLL VCO will oscillate at about 230 MHz. At a PLL's lowest divide ratio (/8), which is actually VCO/128, the output on the pin will be ~1.8 MHz. That's what I'm measuring with the other counter.
jmg,
With a 40 MHz NCO drive, the PLL is completely befuddled, shows no inclination to hunt, and its oscillation is quite stable.. At the other end of the range (i.e. 0 Hz), the story is quite different. The output of the VCO is very nearly aperiodic with sporadic episodes of oscillation, followed by longer epochs of quiescence.
-Phil
http://en.wikipedia.org/wiki/Global_Consciousness_Project
Sounds like a PLL driven with 0Hz would be decent entropy source for a random number generator.
Back on topic, the internal RC clocks are strongly effected by temperature and supply voltage. Given some stable external timing source, they'd give another view into the chip. (even a simple external RC circuit can be enough)
Lawson
* Pin Diode
* VCO
* Over clocking Different Amounts
* Neighboring Counter Module
* RC Clock
Does this cover all known ways of using the Prop chip to detect its own internal temperature?
Detecting temp in more locations inside the chip will help create a better thermal map.