An SEM needs a really tiny electron source. Hot and warm emitters are hard to make really small, and cold microtips and nanotubes don't last long.
Of course, I'd need a photocathode material that could stand being exposed to air.
A fiber-coupled laser can be had for under $20.
John
There really isn't a hardware architecture. Possibly Analog Devices latest blackfin DSP processor might be able to do this, but a 3 us servo loop seems pretty far out there. Unless you are regulating the temperature of a milligram object with tens of Watts of heating or cooling, it is pretty much uncalled for to run it so fast. What is the rate of temperature rise/fall? Will the temperature change by more than a few percent in 3 us? If not, there is no sense in running the loop that fast.
You will need a temperature sensor that can read the temperature in
3 us. Sounds like you are looking for a PIC-like processor, none of those have an A/D converter than can convert at 300 KHz, I don't think. Completing the entire loop calculation in 3 us seems impossible on single-chip processors.Jon
Or a caesium source nearby to reactivate it after pumpdown.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Did you see the PMT pics I posted?
That wire rig looks like the evaporative source for the big photocathode. I wonder how they coat the dynodes.
John
I have worked with the Beagle Board, which has a TI ARM CPU running at 600 MHz to 1 GHz, depending on model. But, the GPIO is limited to one I/O change every 240 ns. That makes it hard to do a lot of stuff with it at high rate. The CPU freezes for that 240 ns, apparently. I have never found out if other ARM implementations have this same I/O scheme.
Jon
If there are no multiplies needed, it may be possible to do this with a Xilinx FPGA, or maybe even a CoolRunner II CPLD. The FPGAs can be had in relatively small SQFP packages for around $10, the CPLDs can be as low as a buck or so. Then, 300 KHz would be trivial if the ADC can get you the samples that fast.
Jon
It depends on the I/O clocking scheme. For example, PIC32 claims to be able to toggle every clock cyle, for a 80MHz rate.
I believe the first lpc210Xs had the same issue, the reason for it is that the peripheral bus and/or the peripherals often run on a slower clock. getting a cpu and memory to run that fast is hard enough
-Lasse
Thanks.
The heater and sensor are that fast--potentially very much faster, in fact. The relevant thermal masses are small, and the actuator power is a few watts. However, it's bandwidth rather than slew rate I care about, because that's what limits the attainable loop gain down at frequencies where there's significant thermal forcing.
It's thermal conduction rather than integrator time constants that limits the speed of temperature control loops--if the plant were just a pure integrator, in principle I could always add gain to make it go as fast as I liked.
I'm not wedded to an ADC-based loop. A very fast thermostat would be accurate enough, assuming its set point and hysteresis can be controlled by the processor, and I may try building it that way. I've built two uC-based temperature controllers in my life, and neither of them needed to be very fast, which is why I'm seeking the wisdom of the August Sagacious Multitude of SED.
First go will use a Kepco BOP with Rs and Cs.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Which PIC? PIC24H can do 500,000 samples per second and PIC32 can do
1,000,000 samples per second.Depends on what he need to do. At 80 MIPs, 240 instructions can certainly do a lot.
est
be
e .ma?
re
It is certainly doable, but not under a buck. Probably around ten bucks.
I'm starting from what I'd like, namely a 40-cent processor, a 3-cent FET, an 8-cent reference, and a few tenth-of-a-cent resistors, and working up from there if necessary. (Making something amazing with almost no apparatus is great fun, even if it requires a mildly generous definition of amazing.)
Nice. Distys have come up some in the world. I've played with the LPC13xx using the LPCXpresso board and Code Red tools, and I quite like it. ARM Cortex M3 does seem to be a good direction go go in for many things.
There are probably follow-ons, for higher-end applications such as laser control, where that would be an excellent approach, but I need to see some revenue before expanding in the direction of FPGAs.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
We'll see!
Thanks
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
This is comparing a SoC to a microcontroller. Those are (still) two different worlds.
IIRC they have fast GPIO as well. Later devices all have fast GPIO.
-- Failure does not prove something is impossible, failure simply indicates you are not using the right tools... nico@nctdevpuntnl (punt=.) --------------------------------------------------------------
I agree. I actually have ARM controller based designs in the field which use a 300kHz software controlled SMPS. That takes about 50% of the available CPU power which leaves enough for the rest of the 'application'.
-- Failure does not prove something is impossible, failure simply indicates you are not using the right tools... nico@nctdevpuntnl (punt=.) --------------------------------------------------------------
I just looked at at datasheet for an lpc2106 and it shows the GPIOs on the peripheral bus, the datasheet for e.g. an lpc2148 mentions fast GPIO and it shows that the GPIOs are now on the faster memory bus
-Lasse
me
re
20I would call these mobile phone processors, in BGA and SDRAM (extra wide bus). They have super fast CPU clock, but lousy I/Os. It's hard to build them for less than 50 bucks.
Clock your processor at 50MHz, use 74AHCxx parts as drivers, and you can get almost 7 bits of PWM (assuming you can use a 3.3V heater supply). The faster you can switch the FET, the better...
Have you looked at integrating twice? Basically a proportional-integral- integral loop? That'll give you 40dB/decade coming back from the second controller zero instead of the 20dB/decade that you'd get from a single integrator. Dunno if there's enough headroom to do a nice conservative design and still get tons-o-gain -- but if you want "huge gain at low frequencies" that's the way to go.
So, "processor as analog", basically?
You may be able to get there from here, after all.
I haven't looked at the Atmel stuff for ages. Assuming you can find a chip that can sample its ADC at 400kHz, and manages at least one MIP/MHz, then you should be able to close a loop at 400kHz if it's hand-written in assembly language on a fast processor, with all the stops pulled for speed and nothing else.
Even then you're still probably looking at one of the Cortex parts rather than an 8-bit part -- I _think_ the ARM Cortex cores have barrel shifters, and I _know_ they have 32-bit signed arithmetic. Both of these things are going to be a huge (and probably necessary) help in achieving a control loop that can complete in a hundred or so processor clocks.
-- www.wescottdesign.com
-- How accurate is "very accurate?" and, how do you know how accurate the measurement is since you're trying to measure the pipe temp through a blanket of steam?
John! John! You have to start thinking like a liberal... "very" is quantitative ;-) ...Jim Thompson
--
| James E.Thompson, CTO | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine. Sometimes I even put it in the food.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.