[DSP-10] Receiver system temperature

Wed Oct 4 09:33:16 EDT 2006

Hi Courtney,

When I worked in Ultrasound Engineering we used to take hot cold noise 
figure measurements
with liquid nitrogen.  Liquid nitrogen was available at the other end of 
the factory, so I would
wheel down one of the hugh stainless steel dewers (from MRI plant) and 
fill it up.  (In fact, that's
how I developed a hernia from pushing those damn dewers so far, that's 
another story.) We used
the hot cold resistor technique described in an old RF design article 
published by some GE folks
at CR&D, when they were designing low noise MRI preamps.  I have the 
article at work, be
aware there is a mistake in the equation, however the nomograph is 
basically correct...  We did
this method a lot and discovered that some manufactures of resistors 
don't change in resistance
as much as other manufactures do.  Besides, after repeated dunking in 
liquid nitrogen some
manufactures resistors become brittle and brake. The leaded resistors 
worked better than the
surface mount 1206 - at that time...  We found the old 1% Allen Bradly 
resistors faired the best
to this "abuse."

Courtney, you are on to something by trying to conduct NF measurement 
using room temp
verses boiling water! We used separate but identical resistor coax 
combinations for measurements.
No gauge R & R here, ha. Be aware, these are tricky measurements! We 
found certain bench
power supplies were nosier than others, GHz oscillations in prototype 
preamps, IF amps with
screw with you mind. #43 ferrite is your friend. We ended up conducting 
all of these measurements
in screen room.

Opps, gotta get dressed and go to work.  Good luck

Dave Garnier - wb9own

Courtney Duncan wrote:
> Perry, thanks for the tips.
>
> They are carbon resistors, 1/8 watt, so they don't have much thermal 
> mass except for the plastic (or whatever it is) packaging.
>
> I didn't really insure even heating of the resistors, and knew that 
> this would be a source of error.  I just put the soldering iron close 
> to one of them, put the thermocouple right in between and kind of 
> waved my hands and said, "Well, if one of them is 80C and the other is 
> 120C, the meter ought to read 100C and the actual temperature should 
> be close +/- 5C, right?"
>
> Maybe not.  It depends on more stuff than I have good visibility into.
>
> I was just wondering if I was doing the right calculation. Currently, 
> I'm adding up the power (from an 8-bit spectrum, CW mode) in the 255 
> frequency bins over about 95 samples, about one minute's worth, and 
> dividing by the N (about 25,000).  Since this gives an answer like 4 
> with a standard deviation of 2, I know it's not very precise.
>
> I know there's ripple in the passband (a big fraction of a dB in 
> places) so was thinking about calculating a temperature for each 
> frequency bin then doing mean and standard deviation on all 255 
> temperatures.  If I'm being hurt by ripple, that might help, but I 
> doubt if it will help a lot.  Something more like 5-10 minutes of data 
> instead of one minute at each temperature would probably help more.
>
> Also I could use LTI mode and get 16 bit spectra.  I doubt that would 
> help a whole lot either, but I should probably be doing both these 
> things anyway, for whatever improvement they're worth, since both are 
> easy.
>
> If it looks like the limiting factor is knowing the precise 
> temperature of the bulk resistor material, I can repeat the experiment 
> in the kitchen (I'm just getting an old E-Bay laptop hooked up for 
> this so it would be possible), in the freezer, the refrigerator, in 
> air, and in warm or hot oil.  My meter only reads accurate to one 
> degree C though, so I'd ultimately be limited by that.
>
> I'll probably end up extending the software to do a least squares fit 
> on all this data too.  An extension I need to do anyway.
>
> I should say that the goal here is to be able to make some sort of 
> rough kitchen-style measurement that verifies performance of the 
> radio, or some preamp that I might buy or build, independently from 
> the spec sheets, to get a warm fuzzy that it's all working as it 
> should.  If I could get a resolution of 25K in the measurement, that 
> would be good enough for this.  Looks like this first outing was more 
> like 250K!
>
> It wouldn't take much to modify this to do a sun-noise measurement, a 
> QRPp one at that.
>
> You know, this experiment isn't high tech by today's standards, but 
> one of the great things about DSP-10 is that you can just go off on a 
> whim and measure stuff like this.  Maybe if I did a brief piece for 
> QEX or something it would get other guys drooling about the 
> possibilities....
>
> (Of course I'd have to have presentable results.)
>
> Thanks again,
>
> Courtney, n5bf/6
>
>> What type resistors did you use?  I suspect you may have a non-linear 
>> noise factor in the resistors coupled with accumulated measurement 
>> errors.  Did you heat the resistors in a "bath"?  If you used hot 
>> air, what did you do to insure even heating of the resistors.
>> You might want to take your samples at 0C and at 20C and see how it 
>> measures up.
>>
>> ----- Original Message ----- From: "Courtney Duncan" 
>> <cbduncan at earthlink.net>
>> To: "DSP-10" <dsp-10 at mailman.qth.net>
>> Sent: Sunday, October 01, 2006 1:44 AM
>> Subject: [DSP-10] Receiver system temperature
>>
>>> I've just measured the receiver temperature of my DSP-10 in the 
>>> following way:
>>>
>>> Solder a pair of 100 ohm resistors in parallel at the end of a piece 
>>> of coax with a BNC on the other end.
>>>
>>> Connect this to the DSP-10 antenna connector in receive mode.
>>>
>>> Measure the bulk temperature of the pair of resistors by inserting 
>>> the thermocouple from a BK Test Bench 390 in between them.
>>>
>>> Captured a minute of data in CW mode with this terminator at 20C 
>>> (293.16K).
>>>
>>> Warmed up the resistor pair to 101C (374.16K) and took another 
>>> minute of data.
>>>
>>> Did two other such data captures, one at 105C (plus or minus 5) and 
>>> another at 20C.
>>>
>>> Sum up all 255 bins of noise over each file and calculate the 
>>> implied receive noise temperature using the formula:
>>>
>>> P2m / P1m = R = ( Tr + T2 ) / ( Tr + T1 )
>>>
>>> Where R is the power ratio between the warmer ( T2 ) and cooler ( T1 
>>> ) measurements and Tr is the unknown receiver temperature.
>>>
>>> I'm assuming that I can just add the temperature of the receiver to 
>>> the temperature of the load like this.
>>>
>>> This reduces to
>>>
>>> Tr = ( T2 - R*T1 ) / ( R - 1 )
>>>
>>> My four measurement files are:
>>>
>>> Temp.   Power average
>>> 20C     3.87855
>>> 20C-2   4.01719
>>> 101C    4.17844
>>> 105C    4.25063
>>>
>>> Using 20C and 101C, I get Tr of 754K.
>>> Using 20C and 105C, I get Tr of 593K.
>>> Using 20C-2 and 105C, I get Tr of 1170K
>>> Using 101C and 105C, I get nonsense.
>>>
>>> This is admittedly this is a "noisy" measurement, but the results 
>>> are not inconsistent with either 600K or 1000K.  Based on this, I 
>>> think I'll keep using 600K for calculations for now.
>>>
>>> Does anyone recall offhand what the spec sheet for the first RF 
>>> amplifier says about its noise figure?
>>>
>>> Courtney, n5bf/6
>>> _______________________________________________
>>> DSP-10 mailing list
>>> DSP-10 at mailman.qth.net
>>> http://mailman.qth.net/mailman/listinfo/dsp-10
>>>
>>
>>
>> _______________________________________________
>> DSP-10 mailing list
>> DSP-10 at mailman.qth.net
>> http://mailman.qth.net/mailman/listinfo/dsp-10
>
> _______________________________________________
> DSP-10 mailing list
> DSP-10 at mailman.qth.net
> http://mailman.qth.net/mailman/listinfo/dsp-10
>