Well, the other option is a lower clock. With this chip, it's gonna pay off to have lower speed capable objects.
I'm going to try and make all my objects and code work in a wide range of clock speed.
I get the feeling that clock settings are going to be even less standardized then with the prop1 from board to board, application to application.
Well, the other option is a lower clock. With this chip, it's gonna pay off to have lower speed capable objects.
I'm going to try and make all my objects and code work in a wide range of clock speed.
I get the feeling that clock settings are going to be even less standardized then with the prop1 from board to board, application to application.
Indeed, there will be many applications where I would not need the speed, just the power of the smartpins and a few cogs. In other applications such as networking and signal processing I will want speed. There is no easy way to change the clock dynamically though since we affect the whole chip but then that's not really a problem in most applications I'm thinking of.
There is no easy way to change the clock dynamically though since we affect the whole chip
That's a good point.
Dynamically changing clock speeds, I wonder if there is a way to detect what the current clock speed is at run time.
Not a show stopper, but would be nice to do.
I was going to write my DSP objects to use some sort of constant calculation at compile time or assembling to calculate fixed timing values.
I'm assuming some people are going to run at 180Mhz, others at <100Mhz, and some overclockers at around 250Mhz+.
When you do a CLKSET from Spin, I believe it also updates that long. The idea was that speed can change at runtime.
That's awesome, this is something we should carry over to Spin2, C, etc.
PropGCC also does this mainly because any PropGCC program is really also a Spin program. It isn't possible to get code running on a P1 without executing at least a few Spin byte codes.
WRT temperature and heat sinks etc. Peters tests results don't include the power input reading so we don't know precisely how much power is being dissipated but if he was able to noticeable affect the temperature in short order with the touch of a finger than we are not talking about a great deal of power dissipation.
The goal is to get the heat away form the chip and keep the die temperature down. Airflow across a bare chip is one way, a small heatsink is another, airflow across a heatsink is even better yet. Now, airflow does not have to mean a fan and a heatsink does not have to mean a thing you glue to the top of the chip. If you can use a vented case then a heatsink situated in such a way to take advantage of natural convection would likely do the trick. You can also use the enclosure itself as a heatsink or a combination of these.
I build a 4 axis stepper drive box using a Gecko G540 which uses its case as a heatsink. The 320W power supply I use has a built in fan. Using a vented enclosure (10"x10"x4") with vents top and bottom combined with a shelf that sets about 10mm above the power supply and has a hole for the PS fan exhaust and a large cut out at the back to provide clearance around the stepper drive unit provides all the cooling that is required. Cool air is drawn in the bottom by the power supply (creating a cool draft at the bottom of the stepper drive) and exhausted out the top vents, the air has to go through the hole in the shelf so it also helps draw air past the stepper drive module housing. For hot climates I fit a small passive heatsink to the stepper drive but that is all.
If you need a sealed enclosure and are using a NEMA 4x type then maybe you can heatsink to the back panel? That gives you ample surface area and you are not dissipating that much heat with the P2 alone to need to worry about overheating as the air will heat the housing and the housing will dissipate to the atmosphere. Or you can mount the P2 PCB in such a way as to be able to use the enclosure itself as the heatsink.
Years ago I installed an early PC based CNC controller (P90 running Windows NT with a real time kernel extension) and given that it was designed for very nasty environments that built an enclosure where the back wall acted as the heatsink and an outer covering over this acted like a duct that a smallish fan directed air through. This was very effective and allowed the electronics to be in a sealed box.
The point is the P2 is not dissipating a lot of heat itself and there are lots of good ways to deal with it (other than pointing a fan at it.)
For reference, running the 1920x1080 vga test code yesterday at 180 or 182.5 MHz, was consuming 0.24A from the 5v supply (so including switching reg losses). While this was only running 3 cogs, the remainder can be had for not much more current.
This setup, including DACs driving VGA monitor was taking 1.2 watts, at around rated clock speed.
The P2D2 is designed so that the reverse side can be mounted direct to a metal case or via a shim and this way we can have a totally sealed enclosure and still have "effective" heatsinking over a prolonged period,both through the bulk of the enclosure and its external surface area.
I know there are situations wHere fans and internal heatsinking won't help enough and when enclosures are out in the hot sun such as HVAC but I also have units mounted out on the side of highways that cop a beating from the weather (and lightning).
BTW, the P2 is the main source of heat (1.8V x 1A + I/O) as the switching regs are very efficient plus I have done measurements around the power supply area as well. Worst case power dissipation of the 1.8V reg I'm using is around 400mW but that is with the P2 well overclocked. Normally it is more like 150mW.
If we were Intel, someone would spill liquid nitrogen over the P2 and report the result. And there would be all kinds of product announcements for liquid cooling your P2.
We aren't. We don't.
BUT I actually would be interested in what is available if we chill the P2 with something like dry ice.
and...
I can't do it, but it would be nice if the fan had LED's that acted as a display. Something simple like "P2" or "250MHz" would be nice:)
If we were Intel, someone would spill liquid nitrogen over the P2 and report the result. And there would be all kinds of product announcements for liquid cooling your P2.
We aren't. We don't.
BUT I actually would be interested in what is available if we chill the P2 with something like dry ice.
and...
I can't do it, but it would be nice if the fan had LED's that acted as a display. Something simple like "P2" or "250MHz" would be nice:)
Great work.
Thanks
You are a bit late to the party. Ozpropdev & tubular have already tested the P2 with dry ice.
You'd think going lower than that would come for free, but it increases hold-time requirements and causes more delay buffers to be inserted so that data doesn't disappear from flipflop inputs before the clock transitions.
A nice heatsink and fan for just a few dollars at Goodwill...
Lots of forgotten tech at the second-hand store. This Super8 editor was made in Chicago. This is all way more interesting than $400 wireless noise-cancelling headphones that only play rap music.
My fan and power supply came from the garage just before I trashed them. Now digging for a nice big heatsink and fan from an old junk pc. All parts free of charge and also can be had by dumpster diving.
Then again, none of its necessary!!!!!!!!!
Comments
I'm going to try and make all my objects and code work in a wide range of clock speed.
I get the feeling that clock settings are going to be even less standardized then with the prop1 from board to board, application to application.
Indeed, there will be many applications where I would not need the speed, just the power of the smartpins and a few cogs. In other applications such as networking and signal processing I will want speed. There is no easy way to change the clock dynamically though since we affect the whole chip but then that's not really a problem in most applications I'm thinking of.
That's a good point.
Dynamically changing clock speeds, I wonder if there is a way to detect what the current clock speed is at run time.
Not a show stopper, but would be nice to do.
I was going to write my DSP objects to use some sort of constant calculation at compile time or assembling to calculate fixed timing values.
I'm assuming some people are going to run at 180Mhz, others at <100Mhz, and some overclockers at around 250Mhz+.
After all these years, I never knew that!
I always thought the clock frequency was just a compile time constant.
When you do a CLKSET from Spin, I believe it also updates that long. The idea was that speed can change at runtime.
That's awesome, this is something we should carry over to Spin2, C, etc.
Yes.
This chip can do a lot per unit cycle! It is worth thinking about when heat or power are limited somehow.
The goal is to get the heat away form the chip and keep the die temperature down. Airflow across a bare chip is one way, a small heatsink is another, airflow across a heatsink is even better yet. Now, airflow does not have to mean a fan and a heatsink does not have to mean a thing you glue to the top of the chip. If you can use a vented case then a heatsink situated in such a way to take advantage of natural convection would likely do the trick. You can also use the enclosure itself as a heatsink or a combination of these.
I build a 4 axis stepper drive box using a Gecko G540 which uses its case as a heatsink. The 320W power supply I use has a built in fan. Using a vented enclosure (10"x10"x4") with vents top and bottom combined with a shelf that sets about 10mm above the power supply and has a hole for the PS fan exhaust and a large cut out at the back to provide clearance around the stepper drive unit provides all the cooling that is required. Cool air is drawn in the bottom by the power supply (creating a cool draft at the bottom of the stepper drive) and exhausted out the top vents, the air has to go through the hole in the shelf so it also helps draw air past the stepper drive module housing. For hot climates I fit a small passive heatsink to the stepper drive but that is all.
If you need a sealed enclosure and are using a NEMA 4x type then maybe you can heatsink to the back panel? That gives you ample surface area and you are not dissipating that much heat with the P2 alone to need to worry about overheating as the air will heat the housing and the housing will dissipate to the atmosphere. Or you can mount the P2 PCB in such a way as to be able to use the enclosure itself as the heatsink.
Years ago I installed an early PC based CNC controller (P90 running Windows NT with a real time kernel extension) and given that it was designed for very nasty environments that built an enclosure where the back wall acted as the heatsink and an outer covering over this acted like a duct that a smallish fan directed air through. This was very effective and allowed the electronics to be in a sealed box.
The point is the P2 is not dissipating a lot of heat itself and there are lots of good ways to deal with it (other than pointing a fan at it.)
For reference, running the 1920x1080 vga test code yesterday at 180 or 182.5 MHz, was consuming 0.24A from the 5v supply (so including switching reg losses). While this was only running 3 cogs, the remainder can be had for not much more current.
This setup, including DACs driving VGA monitor was taking 1.2 watts, at around rated clock speed.
I know there are situations wHere fans and internal heatsinking won't help enough and when enclosures are out in the hot sun such as HVAC but I also have units mounted out on the side of highways that cop a beating from the weather (and lightning).
BTW, the P2 is the main source of heat (1.8V x 1A + I/O) as the switching regs are very efficient plus I have done measurements around the power supply area as well. Worst case power dissipation of the 1.8V reg I'm using is around 400mW but that is with the P2 well overclocked. Normally it is more like 150mW.
We aren't. We don't.
BUT I actually would be interested in what is available if we chill the P2 with something like dry ice.
and...
I can't do it, but it would be nice if the fan had LED's that acted as a display. Something simple like "P2" or "250MHz" would be nice:)
Great work.
Thanks
You are a bit late to the party. Ozpropdev & tubular have already tested the P2 with dry ice.
P2D2 polar edition
Yes thats a frozen TO-92, DS18B20 temperature sensor at the top, hooked up to the helpful P58 pullup
I think it was reading about -57 C, although its only meant to go to -55 I think
You'd think going lower than that would come for free, but it increases hold-time requirements and causes more delay buffers to be inserted so that data doesn't disappear from flipflop inputs before the clock transitions.
I am one who will want to be passive as much as possible.
Lots of forgotten tech at the second-hand store. This Super8 editor was made in Chicago. This is all way more interesting than $400 wireless noise-cancelling headphones that only play rap music.
Then again, none of its necessary!!!!!!!!!