[time-nuts] 5070B once more.... (actually 5370A fans)
richard at karlquist.com
Fri May 22 17:39:59 UTC 2009
Tom Van Baak wrote:
> One comment, and one idea on airflow...
> I've noticed that none of my high-end frequency standards use
> fans at all. That includes every Rb, Cs, and H-maser. This does
> not mean they are all cool to the touch, but my guess is it's better
> to allow temperature gradients to exist than use fans in an attempt
> to even then out.
> I'm sure this requires no small effort in thermal engineering on
> their part. But it also has the side effect of 100% quiet instruments
> and no mechanical wear-out mechanism. Of course, you don't
> want to exceed a high temp limit on any one component, but the
> point is passive heat transfer with metal seems to be the better
> choice than active transfer with air and fans.
Maybe I can put this into perspective with my experience
on the 5071A team. First of all, the 5071A has to be able to
run on a battery, so you can do the flying clock experiment.
This limits the power consumption to a few dozen watts. Considering
the size of the 5071A, that little amount of power consumption
has plenty of heat sinking and doesn't need a fan. The power
supply uses highly efficient Vicor modules. The one thing that
is probably not so well designed is the AC power transformer.
However, it gets a pass, because it doesn't affect power consumption
when on battery, and it is perfectly OK for the transformer to
cook itself as long as the insulation is rated for the temperature
rise. Other than that, there is really no way a transformer can
fail. Another thing is that if any fan were used, it would raise
the issue of microphonics in the 10811 oscillator, which is a serious
issue in the 8662A signal generator. Len Cutler would have
insisted on a microphonic witch hunt if we had used a fan, so we
were all glad we didn't have to go through that. The RF circuits
were in machined boxes that conducted away heat pretty well. They
had a modest internal temperature rise.
Regarding the comments below on the 5370: there are always two
questions with temperature: meeting spec and reliability. Instruments
vary as to which is a bigger issue. Some have temperature
proof measurement techniques that will work virtually until
something burns up, so you can get lulled into a false sense of
security just because the instrument meets spec.
Rick Karlquist N6RK
Member of 5071A design team, circa 1992.
> Now, a suggestion for the 5370 question.
> I'm sure we could discuss for some time what airflow is too much
> or too little or where to place the thermal probes or how to take IR
> scans or how then interpret them. But it seems to me the bottom
> line is not what the temp is inside, but how the unit performs; how
> stable the time interval counter is. voltco; tempco; airco?
> It doesn't matter if you measure a +10C change inside but it has
> little effect on the jitter or accuracy of the counter. Similarly, if a
> tiny +1 C change at some special point causes many picoseconds
> of drift then that's something more to worry about.
> So my suggestion is for someone to take a 5370 and set it up to
> take continuous phase measurements between two stable inputs
> (or delay generator, or a tee from a common reference) and then,
> vary the airflow from way too low to way too fast. You could do
> this in steps every 10 minutes, or linearly over hours, etc.
> The plot you get is then simply accuracy (or jitter or drift) versus
> air flow. This plot, more than one or more raw temperature probe
> measurements or IR pictures, will tell you what the ideal range of
> air flow is. I have no idea what you'd find. But it would settle the
> question about fan replacement for the 5370.
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to
> and follow the instructions there.
More information about the time-nuts