RF small power thermal sensor

[Navarro]

Hi everyone,
I was tinkering here about the idea of measuring low RF powers that isn’t possible with the RF diode bridge way.

One idea came into my mind based that some RF Power Heads are made measuring temperature. 

Anyone ever thought about the idea of making a small machined enclosure with a 50R resistor (maybe a Vishay VPG with 0.1% and 0.2ppm/ºC) and a PTC? I was thinking if it was possible to design something this way so we can measure low power (maybe 0.1nW -70dbm?).

Maybe with some calibration it’s possible to came out with a nice DC-Ghz power sensor that can measure really small power.

[EmmanuelFaure]

They did : http://www.analog.com/media/en/technical-documentation/data-sheets/lt1088.pdf
On the forum : https://www.eevblog.com/forum/projects/anyone-ever-attempt-a-diy-lt1088/

[Gyro]

... and here: https://www.eevblog.com/forum/metrology/diy-precision-ac-rms-to-dc-transfer-standard/

[rhb]

There are very nice thermocouple sensors which Keysight uses.  But low cost they are not.  I've got a 438A, 8462A, 8461D and accessories in transit.  It was not cheap.  There is an eBay listing for the sensor dice asking $7200 for 20 dice for use in repairing blown HP sensors.

One might be able to make a usable sensor by thermally bonding a 50 ohm SMD resistor and a thermistor with a little epoxy.  Solder the resistor on a microstrip board and use magnet wire to pick off the thermistor.  Use a large SMD resistor and a small thermistor epoxied to the top

The kit I am getting was calibrated just prior to shipment.  I've got a 34401A and an 8642C, though these are not professionally calibrated.  The 34401A has been checked with a DMMCheck Plus and is spot on.  I can only go up to 3 GHz, but it might be a good alternative to what I spent which is way too much unless you're old, rich or need it for business.

My main interest is quality low cost OSHW test gear.  In particular I want to try using a small MCU and display to mathematically correct errors rather  than use high precision parts. For example, for a resistance reference, read the temperature, usage history, age ,etc and calculate based on an initial calibration period.  I've got some $3 Chinese voltage references in transit for an initial experiment.

If you'd like to take a crack at it, PM me and we can work out the details.  @svenskelectronik is interested in promoting OSHW and owns a PCB assembly company.  I had to wait unitl I was 65 to be able to have a good electronics lab.  I'd like to make it less expensive so people don't have to wait so long.

[Navarro]

There are very nice thermocouple sensors which Keysight uses.  But low cost they are not.  I've got a 438A, 8462A, 8461D and accessories in transit.  It was not cheap.  There is an eBay listing for the sensor dice asking $7200 for 20 dice for use in repairing blown HP sensors.

One might be able to make a usable sensor by thermally bonding a 50 ohm SMD resistor and a thermistor with a little epoxy.  Solder the resistor on a microstrip board and use magnet wire to pick off the thermistor.  Use a large SMD resistor and a small thermistor epoxied to the top

The kit I am getting was calibrated just prior to shipment.  I've got a 34401A and an 8642C, though these are not professionally calibrated.  The 34401A has been checked with a DMMCheck Plus and is spot on.  I can only go up to 3 GHz, but it might be a good alternative to what I spent which is way too much unless you're old, rich or need it for business.

My main interest is quality low cost OSHW test gear.  In particular I want to try using a small MCU and display to mathematically correct errors rather  than use high precision parts. For example, for a resistance reference, read the temperature, usage history, age ,etc and calculate based on an initial calibration period.  I've got some $3 Chinese voltage references in transit for an initial experiment.

If you'd like to take a crack at it, PM me and we can work out the details.  @svenskelectronik is interested in promoting OSHW and owns a PCB assembly company.  I had to wait unitl I was 65 to be able to have a good electronics lab.  I'd like to make it less expensive so people don't have to wait so long.

This was exactly what I was thinking about.

Get this resistor https://www.digikey.com/product-detail/en/vishay-foil-resistors-division-of-vishay-precision-group/Y201550R0000T9L/804-1034-ND/5028231 and this PTC: https://www.digikey.com/product-detail/en/te-connectivity-measurement-specialties/NB-PTCO-058/223-1797-ND/5272161 and bond them together inside a small aluminium case. The resistor goes into a PCB that are help inside the case and wired-up to the connector with the magnetic wire. On the other side we place a connector so we can 4-Wire the PTC to a microcontroller assembly.

My biggest doubt is:
Is it necessary to do further thermal isolation inside the box?

The nice thing about the thermal sensor is that we can calibrate it with DC and the wire and other parasitic capacitances wouldn't effect the measurement until 100Mhz (my guess).

[Navarro]

They did : http://www.analog.com/media/en/technical-documentation/data-sheets/lt1088.pdf
On the forum : https://www.eevblog.com/forum/projects/anyone-ever-attempt-a-diy-lt1088/

Thanks! That's a nice one, I will dig thought it.

... and here: https://www.eevblog.com/forum/metrology/diy-precision-ac-rms-to-dc-transfer-standard/

That's interesting and I can try this easily here in the lab! 0.01% is an awesome result!

[rhb]

To get to GHz range it needs to be built as an edge launch microstrip with RF in at one end and two of the temperature sensors.  One on the resistor and one nearby.  The enclosure needs to be as small as possible to avoid resonances.  I'm planning on forming enclosures with a 20 ton hydraulic press and a custom die set.  I have a milling machine to make the dies.

Read Franco Rota's article on some diode noise sources he built:

https://www.rf-microwave.com/app/resources/uploads/diodes/VHFComm_NW303.pdf

That will give you a pretty good idea of what the issues are.  In this case with only a single RF component it's a good bit simpler.  Svenskelectronik has agreed to create  board files for the noise source and I'm going to make formed aluminum enclosures per the dimensions of Roca's milled enclosures.

The heat flow part is where it gets tricky.  If it were perfectly insulated the temperature would rise as long as power was applied and never go down.  I suspect that the Keysight sensors pulse modulate the power to the sensor.  I'll find out shortly as I shall be reading a lot of manuals.  It's probably a heat equation analog to an RC circuit.  A small contact to a large heatsink is the resistor and the capacitor in this case is the load  resistor.  You charge it with heat and measure the curve as it rises and then measure the curve as it cools off.  The second sensor measures the build up of heat in the heatsink. 

I spent several months modeling the 1D heat equation a few years ago.  So I have code to solve the problem.  It will require a tweak or two as I was doing fluid flow in porous media, but the math is the same except that the cool side boundary condition is changing as the heatsink warms up.  But that's not a problem to handle.    I was using sparse L1 pursuits which are too compute intensive for an MCU, but it would make great sense as a USB attached instrument.

I'm an old guy with no PCB design tool skills and lots of other projects.  But if you'll create a set of design files, I'll build some and experiment with making an enclosure.  I have a VNWA 3E, so I can only get to 1.3 GHz, but if we have something that works to 1 GHz I think we can find someone with a higher frequency VNA to help.

It would be useful to get a blown HP sensor to teardown or find a teardown video online.

[LaserSteve]

Alan Yates to the rescue:

http://www.vk2zay.net/article/210

You'll have to dig thru a few of his blog pages to fully flesh it out.

Steve

[rhb]

Far more elegant and much easier to construct. Thanks for the link.

[Bud]

Look here, the article begins on page 97 of the PDF

https://ia800502.us.archive.org/30/items/fea_RSGB-Microwave_Projects/RSGB-Microwave%20Projects.pdf

[rhb]

I've been considering this as a thermal flux problem and have an idea I'd like to put up for discussion

   ===============
   R|m1|s1|m2|s2|m3|s3|m4|P
   ===============

  === - insulation
  R -  resistor with adjustable DC bias and RF input
   m1  - small thermal mass
   s1   - temperature sensor
   m1 = m2 = m3
    s1 = s2 = s3
   m4 - large thermal mass
   m4 >> m1
    P - Peltier junction maintaining s3 at constant temperature

In operation a DC bias is applied to R.  When an RF input is present, s1 will rise above s2 causing the device to reduce the DC bias until s1 and s2 match. The RF power is equal to the reduction in the DC bias.

The idea is that when s1 = s2 the heat flux through m1 is equal to the flux through m2 and m3.  m1-3 are as small an electrically insulating thermal conductor as possible with identical dimensions. The Peltier device provides a constant thermal "ground" for the circuit.
Based on comments by Yates identically dimensioned  SMD resistor and diodes using very fine magnet wire for the diode connections should make the most sensitive sensors. m1-3 would be the same dimension, but thinner if possible. 

The construction would be the resistor on a microstrip board with as short a feed from the connector as possible.  Then s1-3 and m1-3 would be stacked in order with  magnet wire leads coming off to standoffs on the sides with thermal grease on all contact surfaces.  A shield/form would be placed over s3 and the stack insulated with spray foam insulation.  After the insulation hardens the shield is removed and m4 and the Peltier junction installed and secured to the PCB.

[Navarro]

I did a small test yesterday:



It's a 100R resistor and a 2n3904 as temperature sensor. All glued with cyanoacrylate glue. The wire is just one of the cores of other wires.

This test sensor was placed inside a Hammond box, the smaller one for guitar pedals that I have laying around here.

How I connected it? 0.5v with a 66332A in the resitor. For the temperature reading I used my 34401A reading as resistance (I know, it isn't the best way to do but it's a 30min test).

Every time I powered the sensor up, the resistance reading in the 34401A got a instant feedback. The turn back to my reference zero was very fast too.

PS: This SMT breadboard is Open-Source "Hardware" https://github.com/PY1CX/SMT-Breadboard

[rhb]

Here are the relevant sections from:

Conduction of Heat in Solids
Carslaw and Jaeger
Oxford Science Publications 1959

This is the canonical reference for the mathematical physics of heat flow.  Section 3.5 treats 1D flow in a slab of infinite extent which is the same as a rod perfectly insulated on the sides.  Section 4.5 treats a rod with the ends held at different temperatures and losses through the sides  which is what I have suggested as a means of implementing a sensor.   Section 4.1 and 4.2 are included to explain the notation used in 4.5.

In the case of the slab, the temperature at any point between the faces is a linear function of distance if the conductivity is constant.  If it is not constant the temperature will have a flatter slope in good conductors and a steeper slope in poor conductors.

I've not actually dealt with the rod case before, so it will require some study.  My expectation is that with good insulation the result will be very similar to the infinite slab case.  I should note that we don't actually need to solve these equations.  The heat flux and hence power, will actually be measured by the change in DC bias required to maintain constant temperature differentials across the 3 sensors.

I thought it worthwhile to document the mathematical physics simply for the sake of completeness.

[rhb]

I finally ordered SMD 2N3904s and 100 ohm resistors to try out the thermal sensor design and got a binocular microscope so I can see what I'm doing.   But I also looked at the datasheet for the AD8307.

"92 dB dynamic range: −75 dBm to +17 dBm to −90 dBm using matching network"

These are available as modules on eBay for very little.  They don't go to 1 GHz, but the LT5537 does
 
"–76dBm to 14dBm, single-ended 50Ω" 

at the price of slightly less dynamic range.  And you have to build it yourself.  Another interesting device is the ADL5513.

With an external attenuator it should not be difficult to reach +40 to +70 dBm. 

I think it worth noting, that these devices have vastly better pulse response than anything that can be built using a thermal sensor which would be functionally  CW only

It seems to me that an OSHW power meter for the sub 1 GHz ISM bands should be reasonably tractable project and generally useful to quite a few people for testing Chinese and other COTS ISM modules.

So how about preparing draft requirements?  Max power, frequency range, resolution, price point, etc.  There are some very interesting 0.25 dB attenuator modules built around Pergrine PE43xxx chips.

I should be equipped to start some experiments by the end of the week.

[rhb]

Have you done any more tests?  In particular looked at the diode drops for three transistors separated by a small wire spacer when a DC pulse is applied to the resistor?  The other side of the third transistor should be glued to a small heatsink.

That's the most important test.  The spacers should be a 3-5 mm length of aluminum or copper  wire roughly the same diameter as the transistor's largest dimension, filed and lapped as flat and square as possible and glued between the transistors with as little glue as possible.  Drill a hole in a block of aluminum or plastic as close to perpendicular as you can and lap the ends of the wire spacers on a wetted piece of 240 or 320  grit wet or dry sand paper placed on a flat surface (piece of 6 mm  plate glass, stone counter top, surface plate).  The water on the back side holds the paper down and the water on the front carries away the material removed.  Stretch the wire with a vise and pliers until it goes slack then cut it with a hacksaw.  The faces of the transistors should also be lapped.  Use a figure 8 motion with very little pressure to lap the wire and transistors.

What you should see is the first transistor change, then the second transistor should change, but not as much.  The third transistor should not change if the duty cycle for the pulses is low.. With good insulation, the 2nd transistor will change more.   Without insulation, some of the heat is radiated before it reaches the second transistor.  The fundamental concept is to apply RF to the resistor via a blocking capacitor and DC to the resistor so that there is a constant difference in the diode drops on the  transistors.  The amount that the DC bias has to be changed to achieve that is the measure of the RF power.  With aerogel powder insulation around the thermal stack it should be quite sensitive and accurate.

I've got  most of my parts, but right now I'm playing with some QRE1113 optical sensors.  I've almost completed the fixtures for gluing them to 6d finish nails and then lapping the faces flat. I *think* I might get as good as 1-2 micron resolution  over a range of 300 microns.  For a $1 part that's pretty cool.  Then I'll have to come up with more uses for it besides the precision square checker.

I still need to order the aerogel powder and I'm waiting on my brass tubing to arrive from China.

[edpalmer42]

What specs are you hoping to achieve with this sensor?

I have a Boonton 4200 Microwattmeter with a 4200-E sensor.  The sensor uses a double-diode circuit and the meter/sensor combination has a measurement floor of -70 dBm (100 pW) and a rated frequency range of 100 KHz to 18 GHz.

My understanding is that most thermal sensors don't come close to those specs.  In particular, they can't reach power levels that low.

I bought a dead sensor and just for giggles replaced the unknown diodes with a couple of Schottky diodes scavenged from some random board.  They weren't RF diodes, just regular Schottky diodes.  Imagine my surprise when the probe not only worked, but was within 2 dB (without calibration) at levels of -30 dBm or greater.  It died below about -35 dBm, but still...  So if I input the calibration factors to my meter, I could use the probe over levels of -30 dBm to whatever without really caring if I fried the diodes.  Just toss in another couple and go.  By the way, at 0 dBm, the frequency response is within 1 dB at 1 GHz.

Ed

[rhb]

I might have to try repairing the capsule of my Millivac 823B if I can figure out how to get it apart without destroying it.

HP uses thermocouples in their sensors .  The 8482A is 0.3 uW to 100 mW to 4.2 GHz.  The 8481D is 100 pW to 10 uW  to 18 GHz.

The thermal mass of a couple of 100 ohm 0805 resistors in parallel is very small, so it requires very little power to heat it. If all that heat is conducted away through 3 transistors and two spacers as I outlined with good thermal insulation, then it is the thermal equivalent of measuring the voltage drop across a couple of resistors.

The objective is a low cost RF power sensor.  I got involved because I've spent a lot of time with the heat equation and recently bought a calibrated 438A with the sensors I quoted.  I got a good deal, but it was not cheap. I thought Navarro's idea worth playing around with.  With a small Peltier cooler to keep the temperature of the 3rd transistor constant and low thermal losses through the sides it should be exquisitely sensitive.

I also plan to evaluate the logarithmic power detectors that AD sells.  There are a number of interesting options in the form of Chinese copies of the eval board designs which are very cheap.  I bought a bunch of expensive tools so I could try designing cheap ones.