[time-nuts] Power Supply Noise Affects Thunderbolt 1 PPS

[Eric Lemmon]

Attila,

It turns out that the MeanWell T-30B switching power supply does, indeed,
require a minimum load on each output- but especially so on the +5 volt
output, which powers the Thunderbolt's 10 MHz and 1 PPS output circuitry.
The specs for this power supply are:  +5V, 0.5-3.0A; +12V, 0.1-1.0A; and
-12V, 0.1-0.5A.  The actual steady-state current draws of my Thunderbolt
are:  +5V, 0.23A; +12V, 0.15A; and -12V, 0.003A.

In order to increase the respective loads above the minimum specified
levels, I added a 16 ohm 5 watt resistor on the +5 volt output, and an 82
ohm 5 watt resistor on the -12 volt output.  Voila!  The Thunderbolt settled
down and the 8040C locked up to the 1 PPS.  I really don't like to use
resistors for this purpose, since they do produce some heat and waste
energy- a bit less than 2 watts apiece in this instance- but it did prove
that too little load on a switching power supply can be a bad thing.

Thanks for sharing your thoughts on this matter.  I enjoy reading your many
postings, and I have learned a lot from the subjects discussed on this list.

73, Eric Lemmon WB6FLY
 

-----Original Message-----
From: time-nuts-bounces at febo.com [mailto:time-nuts-bounces at febo.com] On
Behalf Of Attila Kinali
Sent: Monday, February 27, 2012 12:48 AM
To: Discussion of precise time and frequency measurement
Subject: Re: [time-nuts] Power Supply Noise Affects Thunderbolt 1 PPS

On Sat, 18 Feb 2012 18:59:21 -0800
"Eric Lemmon" <wb6fly at verizon.net> wrote:

> I hooked up a linear triple-output bench supply to run the Thunderbolt,
and
> now the 8040C locked up perfectly on the 1 PPS signal.  Since I don't want
> to tie up one of my bench power supplies to run the Thunderbolt, I plan to
> try adding some filters to the outputs of the T-30B switcher.  I'm
thinking
> that some low-ESR tantalum capacitors might do the trick.  Rather than
> re-invent the wheel, I wonder if others found this problem, and what their
> solution was. 

The usual way to filter such low frequency noise components is to
use T or Pi filters. But it also depends on how the noise looks like.
Is it spikes? Is it a sine? Or triangles?

If there are spikes, add some ceramic capacitors as well, depending
on the nature of the spikes something between 10n and 10u. Having
10n, 1u, 10u is mostly a good compromise for high frequency clamping.

If the noise is sine or triangle like, try with the mentioned
T and Pi filters, maybe multiples of those.

> I also wonder if this T-30B power supply requires a minimum
> load on any of its outputs, for it to operate properly.  Hmmm.

Yes, all switching power supplies (and DC/DC converters) have a
minimum load requirement. What it is, depends on the controller used.
Modern controllers change from PWM mode to PFM (aka pulse skip) to
keep the efficiency high. Some (notably chips from Linear) have something
they call burst mode, ie they do a few switching cycles to drive the voltage
up, then sleep for an extended time.
If you can figure out what chip your power supply is using, you can probably
force it to PWM mode, but then you have to ensure that you enhance the
cooling of the switching transistors as they will run much hotter.

Alternatively, you can use a power supply that gives you an intermediate
voltage (somewhere between 10 and 15V) and use high speed DC/DC converters
that work with switching frequencis in the MHz range instead of 100-300kHz
of an AC/DC supply to generate the voltages needed by the thunderbolt.
This should reduce the noise quite a bit.

If you are really time-nutty, you can let the DC/DC converters produce
a voltage about 1V above what you need and use low noise LDOs (ie not
the 78xx or LM317 & Co) to produce the voltages for the thunderbolt.
This should give you a 60-80dB damping of the noice produced by the
DC/DC converters.


			Attila Kinali
-- 
Why does it take years to find the answers to
the questions one should have asked long ago?

_______________________________________________
time-nuts mailing list -- time-nuts at febo.com
To unsubscribe, go to
https://www.febo.com/cgi-bin/mailman/listinfo/time-nuts
and follow the instructions there.