Test Gear Design - power supply preference

[fcb]

I'm finishing the design on two bits of test gear for release later this year and I'm in a bit of dilemma about how to get juice into the units.  In a nutshell, I'm trying to decide if I go with mains power (IEC socket) or an external DC adaptor, so here's a brief description - I think it's a finely balanced argument but I'm struggling to

P331 - component signature analyzer (like a Huntron Model 30)
USB port on back, sits on bench.
Extruded alloy case with metal front panel (170mm x 55mm x 90mm HWD)
Power consumption 5 to 20W. Isolated +/- supplies needed inside.

Go with IEC socket?
Internal toroidal transformer.
Voltage selector required (inside probably)
Front panel rocker power switch? Latching power push-button? Rear switch only (in IEC socket)? Soft power button on front?
Heat dissipation will probably require 40mm fan? Temperature control? Vents in front or rear (not possible in sides).
Case at earth.  USB port shield at earth (probably).
Extra £15-20 on BOM compared with external brick.

Go with external brick?
External 12v 30W brick with universal input.  2.1mm centre+ or Kycon.
Case at bricks 0V. USB port 0V at bricks 0V (so in most instances, the USB port defines case earthing).
Fan not required (probably).
Internal fixed frequency, fixed duty cycle push-pull converter for isolated supplies - low noise.

The other project is similar (P328 - audio analyser) and is in a case twice the length and consumes up to 40W (60W toroidal or bigger brick) - so will probably get a fan.

I'm attracted to the brick PSU for it's lower cost, fanlessness and easier construction on the other hand I like the IEC socket as it feels a bit more professional. I want to give both products a family feel, so what I choose for one I kinda want to use on the other

[mariush]

I would go with an external brick.

You can isolate the usb port data wires with an Adum3160 or Adum4160 so that your device will be isolated from the computer or whatever gets plugged into USB.  That is, if you don't use a serial to usb chip which can then be isolated using two optocouplers.

Note that there are ready made switching power supplies in plain pcb form you can buy, like this 12v 25w (40w with a fan cooling it down) psu : http://uk.farnell.com/emerson-network-power/lps23/psu-open-pcb-12v-2-1a/dp/1171481

It works out to about the same price as making a linear psu (unless you find some psu cheaper somewhere else, farnell may not be the cheapest) and you may benefit from the fact that the psu already has all the certifications needed (CE, ul, csa, all that crap)

[Dave]

I would go with an internal power supply, mainly because I like integrated solutions. I really hate having to keep track of all the different power bricks. If you have a universal IEC power connector, you can plug your device in practically anywhere.

As for the heat dissipation, you could just put a convection-cooled heatsink on the rear (just a couple of ribs). Your circuit is relatively low power, and you won't need to dissipate more than a few watts anyway. But I suppose it will depend on the number of connectors you will need to accommodate on the rear panel.
I would try to avoid using a fan, for obvious reasons.

[fcb]

Thanks for the feedback mariush.

The isolation of the USB isn't the issue, both have isolated USB in different forms. I do use the Adum devices in the audio analyser and opto-couplers in the signature analyser (the data rate is MUCH lower).

I'm not keen on an internal off-the-shelf SMPSU, it doesn't save cost, as you point out (although it does solve the 115/230 issue) vs. toroidal. But I guess it makes any certification a little easier (the farnell w/s is down at present).

I'd be interested to see if fellow engineers like/loathe SMPSU bricks (lump-in-cord and other types).

[fcb]

I would go with an internal power supply, mainly because I like integrated solutions. I really hate having to keep track of all the different power bricks. If you have a universal IEC power connector, you can plug your device in practically anywhere.

As for the heat dissipation, you could just put a convection-cooled heatsink on the rear (just a couple of ribs). Your circuit is relatively low power, and you won't need to dissipate more than a few watts anyway. But I suppose it will depend on the number of connectors you will need to accommodate on the rear panel.
I would try to avoid using a fan, for obvious reasons.

Thanks Dave for the feedback - yes, I agree keeping track of the bricks can be annoying (especially if you haven't used a piece of gear for a while), a counter to that would be labeling and also specifying the requirements on the rear of the unit (hence 12V/2.5A, 2.1mm centre+ve) - that way if the brick was lost, it would be easy to replace.

Heatsinking.  Good point - might investigate that again, although the back panel on the audio analyser is a bit too busy. It's interesting that you see more fans on test gear than heatsinks these days.

[IanB]

Disadvantages of power bricks: yes, I dislike having to keep track of a proliferation of bricks and remembering which brick goes with which item. On the other hand this is somewhat mitigated by standardization. If everything uses a standard 12-16 V DC power plug (2.1 mm, centre +ve) then I can mix and match and not worry too much about blowing things up by plugging in the wrong power brick.

Power bricks have the advantage generally of keeping heat away from the device being powered and being fanless. I like them for this reason. A second advantage is it gives people the option of other 12 V DC supplies like car batteries. With the increasing number of off-grid users (OK, maybe not in the UK), devices that take 12 V power are becoming increasingly attractive.

Disadvantages of IEC plug mains cables: generally they are thick, heavy, and inflexible. I dislike them on small devices. Is a smaller Figure-8 cable an alternative? These cables are much lighter and more flexible where the power demand is not great.

[fcb]

Some good point IanB, I hadn't considered the weight/flexibility of an IEC.

It's a metal chassis and the figure of 8 (telefunken?) has no grounding - I did briefly consider the cloverleaf (you see it on laptop adaptors), as that has ground - but it's not much smaller than the IEC, but they are starting to be available.

[free_electron]

No bricks !
Labs have power distribution bars with lots of sockets. Bricks are annoying.

If they are plug-in  they block adjacent power outlets , sometimes they are plugged sideways which means the weight is putting stress on the pins. Shove the equipment backward and there is a risk you hit the brick and it cantilevers and falls out of the power bars.

If they are the bricks like used on laptops ( the brick has an iec plug) it means more cable spaghetti. You need to throw the brick on the bench behind the machine. It will fall off the bench , unplugging itself from the machine or iec plug , destroying the low voltage side wires by yanking em to shreds. If you move the machine you need to move the brick as well.  Every bench has iec cables a plenty. Having to deal with the brick is annoying.

Take a look at all the testequipment from agilent tecktronix keithley ets. Unless it is handheld and rechargeable you will find no bricks. For the reasons above. They are annoying.

Now, why dont you simply buy an open chassis switching supply and put that inside the box. No need to muck about with transformers , voltage settings etc.
You can get suck supplies at lower price than external bricks. Go to digitkey and look for meanwell , mei li , phigong and others. These are well established brands that make switchers with all kind of output voltages and are autoranging 90 to 240 volt input. Some pf these are medical grade (leakage) and cost less than 15....

[fcb]

Thanks free_electron for the feedback.

I've been swaying towards the internal PSU with IEC and a small (temperature controlled) fan.

As electrical noise is a big consideration for the audio analyser unit (less so for the signature analyser), I wouldn't use a variable duty-cycle/frequency SMPSU.