monitoring heatsink temperatures

Hi,

I think what I need is a "Remote Diode Temperature Sensor" with an SPI interface. I have found some with SMBus, but this is hard to digitally isolate to my FPGA as it is a bidirectional bus.

cheers, Jamie

The mounting method for sensors does not need to be more complicated than mounting methods used for the semiconductors already sharing the interface. These methods are already assumed to be 'acceptible' in production - as are the isolation methods used at this interface.

RL

National sells an LM45 in a TO-220 package.

John

You can solder a short-lead diode to a solder lug and fastened to the heat sink either with the hardware used to hold down the devices generating the heat OR with a M6 screw into the threads on the bottom of the sink.

Buy a batch of diodes sufficient for your production run as diodes from a single batch are generally quite close in their temperature IV curve. At a quarter of a cent apiece in volume you can buy extras and surplus the rest at the end of production.

Jim

-- "It is the mark of an educated mind to be able to entertain a thought without accepting it." --Aristotle

If you need isolation for your SMBus, you can use Analog Devices digital isolators.

-- Mark

Hi,

The heatsink I am mounting to has TO-220 600V mosfets mounted to them with silpads for electrical insulation, is it ok to also attach TO-220 package LM35 temperature sensors to this heatsink that are on the digital (user) ground? The heatsinks are currently electrically not hooked up to anything, should they be tied to earth ground?

The other heatsink is for a planar transformer, the heatsink goes right on the ferrite or can have an insulating pad. I think using an electrically insulating pad and tying the heatsink to earth ground may be the safest option?

Any other TO-220 analog temperature sensors besides the LM35?

cheers, Jamie

The temperature sensitivity of a simple PN3904 transistor is adequate for a cooling fan controller, why pay for a calibrated analog device? Just hold the flat side of the transistor against the heatsink with a spring clip.

"

As a sensor is normally not self-heating, it's easier to isolate electrically, while still showing a fairly accurate report of the temperature of the mounting surface, regardless of the sensor type or reporting method.

Grounding of heatsinks, directly or at RF frequencies, is something that will depend on how your thermal, mechanical, EMC and safety schemes are intended to inter-relate. Heatsinks for HF noise sources should have a well-defined RF noise return, to reduce the area of vertically polarized planes and horizontally polarized loops. This can be established without direct connections, using safety-rated components, if necessary.

The larger or more physically unstable the heatsink, the harder it will be to avoid a physical interface to safety-grounded mechanical supports.

Although you are not responding to other posters' suggestions re diode junction sensors, these and other indicators ( passive PTC/NTC or thermo-mechanical) are available in TO220 or single-screw packages at a price. An empty TO220 body has a price. FPGA port occupancy and interfacing have a price.

If your assembly uses a mixed technology of SMD and PTH components, surface mount semiconductors or passive sensors can also be thermally coupled to heatsink mounting hardware directly, or through an FR4 layer in multilayer PWBs.

Properly selected non-linear passive sensors can be interfaced to the digital inputs of an FPGA or a dedicated local control circuit more simply than a 'linear' sensor, with no apparent reduction in performance. Simple controllers that operate independently of fail-sensing circuitry can actually offer better options for system survival and recoverability under single-fault abnormal conditions, regardless of the absence of accurate spot temperature reports to 'n' decimal places. Don't lose sight of your initial intended aims.

RL

how about SMT160-30, you can get it in to220 so you can screw it onto the heatsink or in to18 so you can just press it into a hole of thew right size.

The output is a ~4KHz with a duty cycle proportinal to temperature, easy to connect to an FPGA, but I guess you need to implement a division to get the temperature but since 4KHz is so slow it shouldn't be too bad.

afair case is grounded

-Lasse

Hi,

The heatsink is electrically insulated from the HF fets, will it still be a HF noise source because of the AC currents through the fet cases?

I don't understand what you mean regarding the RF noise return, the heatsinks are extruded aluminum, similar but larger to those in an ATX PC computer power supply. Can you elaborate on the method of reducing the area of the current loops in the heatsinks? I guess this is primarily to meet RF emission specs?

cheers, Jamie

This looks like a pretty expensive part, $12 at

Didn't see it at any other distributer.

cheers, Jamie

Hi,

I just ended up using a 4 channel SPI interface ADC (ADC084S101 etc) and four LM35's, one to each heatsink, not as cheap as a remote diode temperature sensor IC like the SMBus LM95234 and 4 diodes though, but SPI seems simpler to buffer and send to an FPGA.

cheers, Jamie

Depending on the thickness and dielectric properties of your insulator, capacitance from each TO220 fet drain to the heatsink will range between 12 and 65pf. This turns your fet drain into an RF current source feeding this hardware above ~5MHz. If the heatsink is isolated from safety ground, it's easier to deal with.

With no path for this current to return to the Fet source-connected reference, a voltage will develop on the heatsink, with the current attempting to return to the source through free space or the surounding hardware and enclosures.

A small capacitor, connected between the noise return and nearby conductive heatsink mounting hardware, will provide a local path for return of the current. This effectively reduces noise voltages on the larger hardware (capacitive voltage division) and current in surrounding hardware and enclosures.

If the heatsink IS unavoidably safety-grounded through identifiable connections, these connections might be minimized and made lossy/hiZ at RF. Screens between fet and heatsink can also be effective at shunting capacitive current before it hits the heatsink.

When ever capacitors are added, their contribution to low frequency leakage current has to be taken into account.

If you look more carefully at your ATX computer supply, you may see provision being made in the layout for these or other interference suppression methods, where applicable.

RL

Hi,

So does this mean hook up a small HV cap, ie. 1kV/1nF, from the heatsink to the local ground (not earth ground) of the fets? Are there any other components that should be used as well like a 500Volt TVS diode in parallel with the cap maybe to protect against charge buildup?

Can you elaborate on this? Is this only a concern if the cap goes from the heatsink to earth ground as it could trip a GFI?

Also for a heatsink with HF diodes is a cap to local ground the same method for EMI noise reduction? The diodes aren't isolated from earth ground in this case.

cheers, Jamie

It would be prudent to lay out for a capacitor, if you are unsure. As there is no source for charge build-up (unless you're building this for Starstruck's computer), there is no reason to complicate this interface.

If the heatsink were connected to safety ground, it would contribute to line frequency leakage currents that are limited by safety regulations. The total leakage should not exceed the values permitted in the applicable application.

Primary or secondary, there's no difference in their potential contributions to noise.

The waveforms on the secondary can be expected to have more modest peaks and dv/dt - likely restricted by the driven source impedance (inductive). Local reverse recovery behaviour, or other rectification or active switching artifacts generate their own flora and fauna - which have amplitudes that need looking after. Some configurations do not even show voltages on the heatsunk surface of the rectifier.

The secondary heatsink will normally have fewer restrictions involving safety, though keeping an eye out for noise current paths is still needed. Their location also helps to reduce their contribution to line conducted interference levels. Transformer capacitive coupling currents are not so benign.

RL

"applicable application"

..groan..

RL