[time-nuts] Power supply for time source concerns

[Taka Kamiya]

Most commercially manufactured time and frequency sources use switching power supply.  However, when contacted, tech support for SRS says, in regards to PRS-10, use linear power supply.  I am guessing, when manufacturers design complete units, they take switching noise into consideration and deal with it with filtering, etc.
I've seen data on TVB's site and others that demonstrates effects of power supply noise.
Now, when I make DIY timing sources, I started including linear power supplies.  My PRS-10/GPS unit has 2.4A 24V linear supply for PRS10 and switching supply for stuff that really doesn't care.  It seem to work well.  But the problem is HEAT!  Almost all timing devices are heat sensitive and drift accordingly.  Open frame linear supplies has efficiency of somewhere around 65 to 70%.  I could, for example, partition inside the case and fan cool the P/S only but doing so for every single DIY unit becomes tedious.
First question to the group is, how do YOU manage this problem?
My thought now is, what about making an external linear supply ONLY and supply all the voltages connected units could possibly need.  I could fan cool that!  I happen to have a rack mounted UPS unit that is well beyond service life.  I can gut it and stuff open frame supplies to 24V, +/-12V, and 5V.  Does anyone see issue with this?  Of course, on timing device side, I will have to put large enough cap to decouple.  
PS.  I was bit by a telecom surplus time source bug.  I have various one already running and more on the way.

--------------------------------------- 
(Mr.) Taka Kamiya
KB4EMF / ex JF2DKG