Cold spray, where do you buy?
in [Design]
|
Prev: Mosquito killer
Next: Finally!
From: John Larkin on 10 Apr 2010 14:16 On Sat, 10 Apr 2010 10:53:32 -0700, RST Engineering <jweir43(a)gmail.com> wrote: >On Sat, 10 Apr 2010 10:08:13 -0700, John Larkin ><jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: > > >>I use freeze spray and heat guns for FPGA timing verification. Prop >>delays are a radical function of temperature. > >I believe you, John, but being an analog kind of guy I can't reason >why propagation delays are a radical function of temperature. Can you >explain? > >Jim Well, that's CMOS. I don't know the physics, but mosfet channel resistance and interconnect resistance have pretty strong positive TCs, so things get slow when they get hot. A reasonable path in and out of a Xilinx FPGA might take 5 ns, and its TC might be in the ballpark of 5 or 10 ps/K. In our delay generators, we generally have to compensate. Gluing a heat sink to the FPGA helps, and adds an air of mystery too. One of my guys is going to engineering school part-time, and as a project he measured the prop delay TCs of some Xilinx chips in various situations. I'll post it if I can find it. As I recall, TTL also slows down with temperature. ECL hardly changes at all, under 1 ps/K for a typical EclipsLite gate. John
From: Joerg on 10 Apr 2010 14:23 John Larkin wrote: > On Sat, 10 Apr 2010 10:53:32 -0700, RST Engineering > <jweir43(a)gmail.com> wrote: > >> On Sat, 10 Apr 2010 10:08:13 -0700, John Larkin >> <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >> >> >>> I use freeze spray and heat guns for FPGA timing verification. Prop >>> delays are a radical function of temperature. I prefer the heat gun. Less of a stench :-) >> I believe you, John, but being an analog kind of guy I can't reason >> why propagation delays are a radical function of temperature. Can you >> explain? >> >> Jim > > Well, that's CMOS. > > I don't know the physics, but mosfet channel resistance and > interconnect resistance have pretty strong positive TCs, so things get > slow when they get hot. A reasonable path in and out of a Xilinx FPGA > might take 5 ns, and its TC might be in the ballpark of 5 or 10 ps/K. That's roughly what I had on CMOS switches as well. So I servoed them (somewhat), using another one on the same die in a DC loop. Worked quite well but it didn't have to be perfect in my cases. [...] -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM.
From: John Larkin on 10 Apr 2010 15:31 On Sat, 10 Apr 2010 11:23:48 -0700, Joerg <invalid(a)invalid.invalid> wrote: >John Larkin wrote: >> On Sat, 10 Apr 2010 10:53:32 -0700, RST Engineering >> <jweir43(a)gmail.com> wrote: >> >>> On Sat, 10 Apr 2010 10:08:13 -0700, John Larkin >>> <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >>> >>> >>>> I use freeze spray and heat guns for FPGA timing verification. Prop >>>> delays are a radical function of temperature. > > >I prefer the heat gun. Less of a stench :-) You must be buying cheap French freeze spray. "Winter in Paris" maybe. > >>> I believe you, John, but being an analog kind of guy I can't reason >>> why propagation delays are a radical function of temperature. Can you >>> explain? >>> >>> Jim >> >> Well, that's CMOS. >> >> I don't know the physics, but mosfet channel resistance and >> interconnect resistance have pretty strong positive TCs, so things get >> slow when they get hot. A reasonable path in and out of a Xilinx FPGA >> might take 5 ns, and its TC might be in the ballpark of 5 or 10 ps/K. > > >That's roughly what I had on CMOS switches as well. So I servoed them >(somewhat), using another one on the same die in a DC loop. Worked quite >well but it didn't have to be perfect in my cases. > >[...] We generally use an LM45 or an LM71 temp sensor and do the tc comps in software, occasionally analog in hardware. Reducing prop delay tc by maybe 5:1 is reasonable without temperature cycling production units. John
From: langwadt on 10 Apr 2010 15:32 On 9 Apr., 01:34, Spehro Pefhany <speffS...(a)interlogDOTyou.knowwhat> wrote: > On Thu, 08 Apr 2010 15:32:27 -0700, the renowned John Larkin > > > > <jjlar...(a)highNOTlandTHIStechnologyPART.com> wrote: > >On Thu, 08 Apr 2010 14:19:52 -0700, Joerg <inva...(a)invalid.invalid> > >wrote: > > >>Gents, > > >>Running low on cold spray. Digikey doesn't sell it anymore, maybe > >>because of shipment safety rules and regs. RadioShack is also dry. > > >>Where can one get that stuff locally? > > >I can ship you a couple of cans. > > >Incidentally, "duster" is the same as "cooler"... just invert! > > >John > > The current 3M ones have kind of a sweet taste to them. > 1-1 difluoroethane, looks fairly benign. > > http://specialtygasesofamerica.com/msds/difluoroethane-msds.pdf > r152a refrigerant, looks like it is atleast somewhat flammable the nonflammble freeze spray is r134a refrigerant. considering that is has a gpw in the hundreds(r152a) and thousands(r134a), it is amazing that you can still buy the stuff for use as a freeze spray -Lasse
From: Joerg on 10 Apr 2010 15:36
John Larkin wrote: > On Sat, 10 Apr 2010 11:23:48 -0700, Joerg <invalid(a)invalid.invalid> > wrote: > >> John Larkin wrote: >>> On Sat, 10 Apr 2010 10:53:32 -0700, RST Engineering >>> <jweir43(a)gmail.com> wrote: >>> >>>> On Sat, 10 Apr 2010 10:08:13 -0700, John Larkin >>>> <jjlarkin(a)highNOTlandTHIStechnologyPART.com> wrote: >>>> >>>> >>>>> I use freeze spray and heat guns for FPGA timing verification. Prop >>>>> delays are a radical function of temperature. >> >> I prefer the heat gun. Less of a stench :-) > > > You must be buying cheap French freeze spray. "Winter in Paris" maybe. > Now if they'd put a "Harley Davidson" aftershave scent in it, that would be ok. > >>>> I believe you, John, but being an analog kind of guy I can't reason >>>> why propagation delays are a radical function of temperature. Can you >>>> explain? >>>> >>>> Jim >>> Well, that's CMOS. >>> >>> I don't know the physics, but mosfet channel resistance and >>> interconnect resistance have pretty strong positive TCs, so things get >>> slow when they get hot. A reasonable path in and out of a Xilinx FPGA >>> might take 5 ns, and its TC might be in the ballpark of 5 or 10 ps/K. >> >> That's roughly what I had on CMOS switches as well. So I servoed them >> (somewhat), using another one on the same die in a DC loop. Worked quite >> well but it didn't have to be perfect in my cases. >> >> [...] > > > We generally use an LM45 or an LM71 temp sensor and do the tc comps in > software, occasionally analog in hardware. Reducing prop delay tc by > maybe 5:1 is reasonable without temperature cycling production units. > I always wondered why FPGA don't have a feature to program an on-chip temperature sensor onto one pin, or at least a diode path. uC often have that, for example the MSP430 series. -- Regards, Joerg http://www.analogconsultants.com/ "gmail" domain blocked because of excessive spam. Use another domain or send PM. |