Where to put fan in new test equipment design?

[fcb]

I'm working on the design of a couple of bits T&M gear and would like some opinions as to fan cooling.

The gear is quite small (170W x 55H x 220D) specialist T&M equipment and has various internal regulators, MOSFET's etc... that need cooling.  As I want to make the gear really reliable I want to keep the heat down and will be using a quality 40mm fan and a heatsink block.

The external casing is metal and so are the front and back panels.

My prototype design (12V brick powered) just had a fan on the back panel and vents in the front panel, the main power component being mounted on the rear panel.  Works OK, but I'm going mains powered and would like to have a more 'generic' design as there are several pieces of kit in the series.

I'm looking at using the 100mm long LAM4K heatsink to mount any the TO220 parts to - this would be bolted to the PCB with a couple of tapped holes in the bottom of the heatsink. http://www.fischerelektronik.de/web_fischer/en_GB/heatsinks/D02/Miniature%20cooling%20aggregates/PR/LAM4K_/index.xhtml

My thought is to secure the heatsink directly to the rear panel to help support the PCB, have a hole in the back panel for airflow and then mount the fan in-board on the end of the heatsink (so the fan is in the centre of the product).  Vents in the rear panel to allow cooling air in to the product, and the fan blows this air out down the LAM4K. At the moment I prefer this idea to sandwiching the fan between the rear panel and the LAM4K - due to rigidity concerns.

I don't recall seeing this arrangement in any of the T&M gear I own though...

Pics attached of the prototype and the LAM4K.

[nctnico]

It will be better for the fan if it handles cold air instead of hot air.

[fcb]

I agree nctnico.

The extra heat picked up by the air entering the case from the PCB should be fairly minimal (perhaps another 5C), the bulk of the heat generated in the design is from the parts strapped to the LAM4K.

[alm]

The HP/Agilent/Keysight 1/2 rack mobile communication supplies (e.g. 66309/66312/66311/66319/66321) have a similar arrangement to what you describe. See pics here.

Downsides would be more difficult to filter air, since it will come in through multiple holes/gaps in the case. Obviously it depends on the intended environment, fan speed and sensitivity to dust whether filtering is necessary. Also if there are any components that need airflow that are not on the heat sink, then you would need careful planning of the airflow (and may need some ducting) to get sufficient air flow over those components.

If reliability is important, then I would consider going fanless if feasible. Fans are mechanical devices that will fail after a couple of years, especially when exposed to heat. Look at the Keithley 2400 SMU and at the HP/Agilent E361xa power supplies for ideas how this could work.

As a user, I definitely prefer fanless equipment both for reliability and noise. And for noise, if there is a large difference between minimum and maximum dissipation (think a linear power supply), please make the fan temperature controlled.

[fcb]

Thanks alm,

The filtering is an interesting point.  As this is mainly lab equipment, then I'm not planning on putting any filtering on it (none of my test gear - and I have fair bit - has dust filtering, a few bits - VNA, SA, etc -  have EMC mesh over the fans.)

The problem with a passive heat-sink is that it's difficult to stack up, also the extruded case would be expensive to modify like the SMU.  I did briefly consider a heatsink on the back panel, problem is that it would chew up alot of real-estate if it was effective, also, the circuit layout would be more complex and less-ideal in some of the other kit.

The fan will be speed controlled, the prototype had that.

[Tomorokoshi]

How many watts do you need to dissipate?

What is the maximum rated ambient temperature of the equipment?

What are the maximum rated temperatures of the major heat dissipating components?

[fcb]

When running hard, about 30W.

Max rated ambient: 30-40C (indoor lab/workshop).  Most parts are either commercial temp rated or industrial.

[floobydust]

I'm confused what parts you are cooling.
In the pic the fan is on the power resistors and SSR's/SIP DC-DC's? and the back panel IC cannot dissipate much even with airflow through the holes. Not 30W on a flat piece.

By using that heatsink it would only be the power IC getting serious cooling- not the transformer/mosfets, power resistors and SSR's (unless you have a way to also have enclosure airflow).

Your fan holes look really small, much less area than the fan so I think that is restrictive. I would consider a U turn via the back panel i.e have the outlet behind the transformer as I would not like the (tiny) slots blowing air out in my face. But the heatsink would have to be in the path too.

If the unit has hazardous voltages, then there is a safety issue having large vents nearby.

[fcb]

Hi floobydust,

The pix are of a prototype only. The part attached to the back panel is a power amplifier IC and in the prototype doesn't dissipate too much - hence a major reason for the design changes.

I posted a pic on the this thread of the proposed airflow & the tunnel type heatsink.

I've been considering a stainless steel mesh for the slots/fan on the back of the unit - but this reduces airflow and is a sod to fix in-place (requires spot welding) and I've got plenty of T&M gear on the shelf that you can definitely poke stuff through the sides and touch unknown voltages. So probably IP20 (12.5mm finger proof) not IP30 (2.5mm tool/thick wire) or IP40 (1mm slender wire).

[fcb]

So rather than start a fresh thread as it's the same project.

I'm looking at switching this piece of equipment on/off, I struggling to decide between the following:

1. No on/off switch, unit wakes up if USB is connected.
2. On/off switch on the rear panel - probably a rocker type, SPST or DPST?
3. Rocker on/off switch on the front, wires go from mains inlet to rocker and back to the PSU section towards the back.
4. Latching push-button type - mounted on PCB towards the front.
5. As 4 above, but switch mounted towards the rear inside the unit - some sort of long rod connecting a button on the front panel.
6. Soft power-switch.  A small 6x6 r/a push button poking through the front panel - puts electronics to sleep, unit still draws small amount (0.1-1W) if off.

I've seen all of these on bench test equipment - #2 is my least favourite. #5 possibly my most favourite, but I don't have the budget to injection mould a spindly thing like you see in some test gear.

[slurry]

Try to use the casing as a cooler as much as possible, it saves you a bit of the cooling problem, it may even be sufficient enough to let you lower the fan speed and save some hours on the fans lifespan.

Regarding power switch, mechanical solutions is expensive and unnecessary complicated if not needed,
You may just use the power from the USB-cable in sleep and a soft-button to wake up the rest of the circuit?

[fcb]

I'm using as much of the casing as I can access. The device only uses relatively high powers on some settings - so there's built in a temperature sensor on the main heat-sink element to control the fan speed (off/low/high).

One of the issues is that i've chosen stainless steel front and rear panels, SS is a bad conductor of heat (around 10 times worse than the aluminium extrusion - but mounting the devices to the extrusion would be very difficult and make it tough to service.

[fcb]

As for the power switch - I can't quite bring myself to eliminating it altogether.  There are several products in the range and one of them is likely to have a small display on it, so leaving it on might not be an option.

Perhaps I'm just getting old, but I don't tend to leave my test equipment on (with the exception of my favorite scope - which stays on all day if i'm using it) - would like to hear others opinions on power switches, soft vs hard switches, front vs rear switches etc..

[alm]

My preferences for power switch would probably be 5, 3/4, 6, 2, 1. I am not aware of any off-the-shelf plastic push rods, but that does not mean they do not exist. The big instrument manufacturers even install a push rod for instruments where they clearly cheaped out on the front panel (e.g. Keithley 2304 with its low contrast 16x2 display and low number of tactile buttons), but obviously they make them in large volume.

A switch on the back is not something I would expect to be operated in daily use, but only for maintenance or something (similar to the switch on ATX computer power supplies). The only time a switch on the back makes sense to me is if the unit is likely to be operated with the back facing the user, for example the Agilent 16900 series logic analyzers had a soft power switch on the front and back, so people operating it reversed (to have easier access to the cables) could switch it on and off from there.

[floobydust]

If the output can contribute to smoke/pain/electrocution, keep the on/off switch on the front panel. I'm not sure what this box does.
I could see an auto-sleep function, based on USB (what if PC goes to sleep?) or maybe the D-connector port. The display is kind of a power on indicator.

Otherwise, I think this enclosure is already maxxed out for size. I would push back (to management) and use a larger enclosure or option for a second PCB, to stack/mezzanine to get more room inside. Right now it leaves no room for future mods or changes, better cooling etc. I've always found it to be bad when a new product design is already stuffed in the enclosure- you get boxed in and it costs more in the long run.

[fcb]

Thanks both ALM and floobydust.

I'd love to push back (to management), but we are a small business and I'm the boss & chief designer 

The actual enclosure is the same extrusion as shown, just twice as long. Which bizarrely causes more problems if I want a push-button on/off.  In all my years of taking apart test equipment and watching tear-downs I haven't seen a push rod that wasn't custom made (mainly injection molded). So I think that #6 looks increasingly unfeasible unless I machine one, so probably option #3 or #4.

[alm]

Something that I have seen occasionally in equipment that has a PCB-mounted switch on the front is that no traces would connect to the pads on the PCB, and the wires are soldered to the top of the switch. This way avoids having to run mains traces all the way over the PCB, while being faster to assemble than a bulk-mounted power switch.

[bson]

Somewhat tangentially, are there any other sources for forced-air blocks like this?

[alm]

Fischer makes them. According to octopart, Allied and RS Components sell them.

[bson]

Cool, ordered a Fischer LAM 5 to take a look at from Allied... they claim 10 day availability.  I need to dissipate up to 100W from a pair of TO-3Ps (2SC5242) so at 0.4K/W this should be alright with a speed controlled fan.  (Using a MAX31740 and SMD switching MOSFET TBD.)

https://www.alliedelec.com/fischer-lam-5-100-24/70748565/

[floobydust]

In the UK, I found ABL Heatsinks has neat offerings.
In North America, Wakefield-Vette and Aavid.
I found Aavid stuff never achieved their datasheet claims and don't use them.
These are higher priced companies, as china was dumping aluminium for a long time and thus the cheapest.

[fcb]

In my experience the Fischer heatsinks are ALOT cheaper if you ditch their suggested fan and buy your own.

I laughed pretty hard when they sent me a quote with a fan, but it was very reasonable when they removed it from the quote.  I first started playing with the LAM4K extrusion (as per the pictures) but found that the clips they sent (I had 4/5 different variants) are great if you never want to be able disassemble the unit - almost impossible to remove in-situ.

TME also carry these.