[pvrusb2] Fun with heat / power [was: (an indescipherable string)]

[Mike Isely]

(see below)

On Fri, 12 Jul 2013, Gary Buhrmaster wrote:

> On Fri, Jul 12, 2013 at 8:24 PM, Jan Brouwer <jan at brewsky.nl> wrote:
> ...
> > My unit does get pretty hot... I didn't think of cooling it before, so
> > maybe that would do the trick!
> 
> Certain Hauppauge devices have a history of getting hot
> and resetting.  Certain Hauppauge devices also have a
> history of having power supplies going (partially) bad
> (especially if they are a few years old).  Either can cause
> "random crashes".  Since it is (usually) the easiest to
> test (unless your computer is next to the refrigerator :-),
> I always swap out the power supply first.
> 

Yes, this is definitely the case.  In fact, back in prehistory when I 
was much more a hardware hacker than a software guy I learned that when 
digital circuits misbehave the FIRST thing one should always suspect is 
the power supply!

To summarize various previous discussions on this topic, the things I've 
seen over the years:

These units do dissipate heat.  If you have it sitting in the open 
ambient air (with the room at a normal comfortable temperature) then it 
generally should not be a problem.  But if the device is in a closed-up 
box where there's no air circulation or if perhaps it's stacked with 
other heat-generating devices, then you may have problems with crashes / 
resets.  I do remember one specific tale a number of years back where 
the user had stacked 3-4 24xxx devices on top of each other and then 
stuffed them in a closed box.  That didn't work out so well :-)

I find that people get frequently fooled by bad power.  The issue here 
is that normal PC power supplies are in fact very forgiving of bad mains 
- typical PC PSUs have large capacitors and a lot of internal 
protection.  They're designed to process/condition 300-700 watts without 
wasting a lot of energy and it takes a good design to do that.  I've had 
PCs actually keep going perfectly even when the building power fails for 
a fraction of a second.  The monitor might wink out for a moment but the 
PC just keeps truckin' along.

USB-powered devices (well, those not connected to a powered hub...) have 
the benefit of drawing power from that nice stable PSU.  But any 
peripheral which is self-powered with, say, it's own little power brick 
is at a disadvantage.  Power spikes, brown-outs, or just plain saggy 
power won't affect the PC but the cheap wall-wart just can't defend 
against crap like that.  It doesn't have the same quality of filtering, 
and it is just too small to have enough hold-up capacitance to survive a 
dropout.  That leads to situations where strangely enough pvrusb2 
devices seem to randomly crash while the PC itself is unscathed.  
People frequently conclude that the device is flakey when in fact it's 
just bad power that can't be compensated by a simplistic power brick.

So people seeing pvrusb2 resets don't immediately suspect bad power 
because, well, the PC is not crashing...

Even a surge protector won't help the situation if the problem is 
dropouts or sags - surge protectors just block damaging spikes and 
filter EMI but they don't store energy to cover glitchy power losses, 
even for a fraction of a second.  For that you need a UPS, and one that 
is fast enough on switchover to keep from glitching the power brick (I 
think most modern UPSes meet that requirement).

One big tip-off of bad power is if you can correlate pvrusb2 crashes 
with external electrical events, like say your refrigerator cycling or a 
nearby A/C compressor turning on/off.  A sump pump could even glitch 
things.  Anything that can suddenly draw (or cease drawing) large loads 
can do this.  And inductive loads - like a motor - can really screw with 
your power.

And yes, I've seen reports of people who have correlated pvrusb2 crashes 
with their A/C compressor cycling on/off.  (Though it has been a few 
years...)

You won't be able to analyze bad power with a voltmeter (analog, DMM, 
DVM, whatever..).  The events you are looking for will be too fast and 
the meter is usually filtered internally for slower (more readable) 
response.  If you really want to look for glitches, use an oscilloscope 
- but BE CAREFUL - you're messing with high voltage + high current which 
can seriously injure or kill you.  You can blow up the scope if you get 
the grounding wrong!  Actually it would be safer instead to 'scope it 
indirectly - say look at the output of a doorbell transformer wired to 
the mains.  Then at least it's isolated, current-limited 24VAC instead 
of 120VAC or 240VAC (depending obviously on your country's power grid).

Anyway, if you suspect bad power, one easy test for this is just to plug 
the pvrusb2 device's power brick into a UPS.  If you suspect heat 
issues, try just using a simple desk fan to blow air across the top for 
a while.  If either experiment changes the behavior then you know you're 
onto something.

  -Mike


-- 

Mike Isely
isely @ isely (dot) net
PGP: 03 54 43 4D 75 E5 CC 92 71 16 01 E2 B5 F5 C1 E8