Tip:
Highlight text to annotate it
X
Hi, this is just going to be a quick little aside video
from something I found during the testing
the review of this Atten power supply unit, and I thought
I'd just make a quick separate video about this, rather than include it somewhere
in the middle of this review video, so if you want to check the review of this thing, check it out.
Now, what I'm doing is I'm measuring the noise performance
--the output noise of this, power supply, and
the way I'm doing that is I've just got my BNC to banana
plug adapter there, my-- going straight into
the scope via a coax of course, and I've got my scope
set up for a bandwidth limit, now that's quite important because
the specs usually, of a power supply -- let's take a look at it
um, the ripple here in this case, there it is
20 hertz to 20 megahertz, so it's a 20 megahertz bandwidth
limited range, and that's what the bandwidth limit on your scope does.
And if you turn it off, let's have a look at the display at the moment,
okay, that's with the bandwidth limit on, let's turn it off,
and we'll see that the noise is very significantly -- peak to peak noise -- there
very significantly higher than if you've got the bandwidth
limit down to 20 megahertz. So you definitely want
to measure the performance over that limit. But look what we're getting
here, okay? This is supposed to be a linear power supply.
Quite quiet, okay? Now the spec is
one millivolt, RMS of course. There it is.
One millivolt RMS, now RMS is the key, it doesn't tell you anything about peak-to-peak.
But you look at this and you go, well,
why are we getting this switching effect here, this is not a
switching power supply! So where is it coming from, is it coming
from the circ-- processor inside here, is it coming like--
display refresh or something like that, is it coming from
internal from the power supply, well we'll find out in a second.
Let's take a look at it, 5 microseconds per division, 3 divisions,
that's 66.6 kilohertz or thereabouts, significant switching
you know, component there. It dominates that display.
And if you weren't careful, if you just hooked this up and you didn't know what
you were doing, you might think "well, this is coming out of this power supply"
"Well, this is a *** power supply, look at the switching, it's horrible!"
You know, but is it coming from this power supply?
You may guess the answer. No, it's not!
Let's find out why. Now the first thing we might check is that what happens
when you disconnect it. Okay, you disconnect it, it goes away. Not a problem.
Let's connect one side of it to it, not a problem, connect the other side
well, we're just getting 50 hertz garbage on there.
Let's not worry about that too much. But look, you can see
the switching component, folks, is sti-- oop
oop-- I accidently hit it. The switching component is still
in there. There it is there, you can see it.
So it's coming-- it's definitely coming through this power suppy. So you still
might think "okay, this power supply is the culprit".
But let's switch the output off, okay, so the relay it should disconnect
the output there, so let's [ beeping ] oop--
[ beeping ] switch all of the outputs off. There we go, our outputs are physically
disconnected. It's switched off and it's disconnected those
outputs but it's still there. You'll notice that the noise
really doesn't change much at all, whether or not you've got that output on.
or off. And next up you might think "well, is this BNC
look, here's this coax, move it around". Look, it seems to be a little bit changing
a little bit there, it's sort of, you know, picking up stuff,
so I don't know, is it the BNC? Well let's use a different BNC.
So as you can see, the switching is still there even with this different BNC
cable. So it's not that. It's not picking it up, so [ exasperated sigh ]
What do we try next to try and figure out the source of this
switching frequency because, well, is it the power supply? Because that's the thing
right? If you're doing this-- these sorts of measurements you have to know
exactly where all your noise sources are coming from.
I know for a fact it's not within this power supply. So I'm going to let you try and
guess where it's coming from. We're going to try and hunt it down.
So what we're going to do now is, well, is there any other lab gear
around here, sort of picking up noise -- maybe this coax ain't that good --
right, is it picking up noise somewhere because it's not shielded
all the way, maybe it's picking up noise somewhere else.
Hmm. Let's try a very simple thing, let's just switch the power supply off and see
what happens. Aha, look at that
folks! It's still coming through, it's being picked up.
What do we do next? Let's pull the mains plug on this thing.
Look at that, I just pulled
the mains plug, and it's still coming through! What do we try
next? Well, let's try a real scope probe, you know. I've got my
500 megahertz Agilent scope probe here right, it's a real
fair dinkum probe, and look, I've even got my antenna earth loop there.
And look, we're not picking up anything at all!
Now, I'm using my scope probe to connect up to the power supply here, and look,
it appears to have gone, but let's stop that.
Bingo, it ain't gone, look, it's still there, there it is 3 divisions
1, 2, 3, that's 66.6 kilohertz still there,
So we are getting a better result, because this is a much
higher quality pho--you know, proper shielded
you know, oscilloscope probe, as opposed to just, you know, some regular coax which
may not have 100% coverage, but we're still picking it up so I've gone back to my
regular coax here just so we can see the effect again, and mains cable
or no mains cable on this supply,
powered up or not powered up, it makes no difference. We're still getting that switching noise.
Is this power supply magic? Even when it's not
powered is it magically generating this switching frequency?
No, of course not! It's picking it up somewhere.
Alright, so I suspect it's common-mode noise being picked up
through the mains system, and because this oscilloscope is mains earth referenced,
and as you saw in my-- the mysterious oscilloscope
phenomenon, you can actually get ESD impulses which
jump onto the coax cable-- onto the lead
and then into, because this oscilloscope only has so much common-mode
rejection from the mains, can actually generate
input noise coupled through
the earth system. Now what I've got is this isolation transformer here.
And this physically removes the mains earth and
isolates this. So it effectively turns this oscilloscope int-- it's not mains
earth referenced anymore, so now you can use your scope probe to
probe your circuits and you don't blow them up, etc. Because it physically
removes the earth on this thing, it's not recommended to do this
by the way, power the scope, usually you power your product through this thing,
and not your scope, or you use a proper high-voltage probe.
Eh, that's for a different video. But look, it's physically
changed now, it's still there but look, it's different.
We've got different components picked up-- we're still 5
microseconds per division there, but it's not that
consistent 66.6 kilohertz we saw before, so aha!
We're getting closer to this thing. And you--
you'll see that it'll instantly go back if I plug the proper mains cable
back into this thing, it'll instantly go back to exactly what we saw before, so
we're tracking this thing down. So what we've got is some sort of
switching device somewhere, either in the room or
on the mains distribution system that is causing this thing.
So just to show you that it's not the Agilent oscilloscope doing this, here it is on a Rigol
scope. 5 microseconds per division, exactly the same thing.
happening. So what do you do? You start looking
for things that are either within the direct vicinity
that are switching, or something that is connected to the mains system.
So you start by, well, I've got my lights up here, my LED lights
I'll switch those off. Does it make a difference?
No, nothing, it's not those. Not a problem. Is it the
fluoro lights in the lab? Well, only one way to find out.
Nope, look at that!
Still exactly the same. And it's none of my
gear, I've turned all my gear off on the lab, I've switched
the computer off in the office cubicle
I've got in here, and I still can't find it, so
let's go investigate under here.
Now here is all of my power boards.
They're all connected down to the one, and there's a whole bunch more powering
my electronics bench over there as opposed to my teardown bench, so
these ones here, there's a-- you know-- there's a few things
plugged in. Let's have a look, there's a-- what have we got, no.
We've just got a mains cable that's going off to nothing, nothing,
going up to gear that I know is switched off. Aha!
What is this? What is this?
Hello mister
Qili Power! Hmm.
Well, there's only one way to find out. I've now switched this down to 2 millivolts
per division, you see we're getting the huge noise there, I'm still measuring the output direct
on the power supply by the way. Whoop, it's not switched on, now it's switched
on. There we go, so we're picking up that noise there.
What you would think is noise coming from this power supply,
if you didn't know how to measure things properly.
Let's disconnect this stupid Qili, look at that
switching power supply, look at it go up as I put it near that coax. Look!
***! So let's-- I'm going to pull the cord on this! Just going to yank
it, here we go. Tada! We've found our culprit
folks! One of these switching-- cheap-*** switching
power supplies plugged into the same mains board
as what I was powering my oscilloscope and my Atten power supply
from! Bingo! Big trap for young players.
So there you go, we're now at 5 millivolts
per division and you can see that we're still picking up noise,
and that is [ unintelligible ]-- most likely more common-mode noise
between the earth and the neutral in the mains system, but
we've gotten rid of that huge spike which we were getting before
that was upsetting our measurements. We can try and track down sources of this type of
common-mode noise, we can filter our mains and do all sorts of stuff like that to
reduce it. But I'm pretty happy now that we've actually gotten rid of that
huge 66 kilohertz spike we were getting from that
switching power supply. And of course if we go back to our
original issue and just disconnect it from there. Bingo, [ laughing ] we're no more noise
and we can even go down to 500 microvolts, you know, per
division and we're sweet there. Why is that not updating?
Ahh, bloody firmware in this thing. I haven't got the latest firmware for this
Rigol scope yet, so it has some freezing issues with the horizontal
mode. But there you go, that's 500 microvolts per division.
Switch that and put that in we pick up a bit more, we put it over on our
power supply over here, and we're going to pick up a butt load of
common-mode noise. But that's not coming from our
power supply. So you might be asking "well, why was this thing picking
it up even though it's switch off and disconnected
from the mains like that?" Well, it's because
the internal circuitry and the internal transformer in here is
effectively -- via AC coupling -- is effectively
working as a very effective
you know, pickup antenna so to speak, and that's why this
oscilloscope probe won't pick it up, because this is a relatively high-frequency pickup
coil, okay? It's going to pick up, you know, ESD and lots of high-frequency discharge
and stuff like that, as I've shown in previous videos. So the transformer
inside here and the coupling to it is basically going to
effectively work as a better, lower frequency, pickup
antenna for that stuff. That's why if we disconnect it, bingo, we're gone.
Okay? But we hook it up, this thing is entirely switched off, disconnected from the mains plug
so it's not actually picking it up through the mains earth.
It's still working as a very effective antenna for picking up that common-mode noise.
It's still working as a very effective antenna for picking up that common-mode noise.
And common-mode noise comes in all types folks, it can
come from anywhere, be careful. Watch this, I'm going to grab this coax
with one hand, touch the screen over here, no folks
it ain't magic, it's picking up the noise-- the switching
refresh of the screen there. Look at that, woohoo!
And of course that is one of the claimed uh-- well it is one of
the disadvantages of these digital scopes is that they can be spewing out
stuff which can interfere with low value measurements. So that's why, you know,
a lot of the greybeards frown upon these digital scopes because,
ah, you know, be generating all sorts of crap. You won't get this sort of thing happening with an
analog scope. So what are we going to do when measuring our
Atten power supply here? We know it's a linear supply, it's not spewing out
any switching stuff. So all this high-frequency
peaks in here, common-mode noise
is coming from somewhere else in our measurement system.
So when we're measuring the noise on a linear power supply like this
Atten power supply, we know that these high-frequency switching
components in here are effectively common-mode noise
being picked up somewhere else in the system. So really you want to
chop those out and only look at that in there.
So as you can see, even though we got rid of that mains source we're still picking up, you know
a lot of common-mode noise in here. And unfortunately
that's going to be hard to get rid of. Now, even if I power both this scope
and the power supply through a filter--
a mains input filter board-- so I've got both bits of gear
that's the only thing off that filter, we're still picking up this pain
in the *** common-mode noise here. Look at that!
So, what's that coming from? Well, if we go full-circle
back to something we tested before. Our lights.
Let's turn it off. Look at that!
folks! Bingo. So now we're talking
We've started to eliminate all of our problems here and
getting towards more of the real noise performance
of this power supply. So
we still have-- oop, that's me by the way, be careful--
So that's-- um-- we're getting very very close. There's still one
-- there's still a burst in there that's triggering
off that. So it's obviously oop-- probably we can move our
trigger around and there we go, we can single-shot capture off that.
So there is some-- another burst event coming in there, but
really, umm, that folks now
we can at least get a more decent measurement of our
power supply. You can see how this is not easy, we originally had
a common-mode noise source direct-- a switching power
supply directly on there. We thought we eliminated the lights, but
we didn't. Let's switch those lights back on. Look at that! Unbelievable!
Woohoo! They're the LED lights I've got up the top. And they're not even
supposed to be PWMing, they're supposed to be constantly on at
maximum brightness there. I can turn my other set of LED
lights above does absolutely nothing, but those
lights I've got up there, big switching noise. So
So we came full circle there, and we're getting closer to eliminating everything.
So let's actually look at the differences in the quality of some
coaxes. I've got this particular coax cable here
we're 2 millivolts per division, I'll keep it on that.
That, you can see, we're picking up lots of high-frequency
common-mode noise there, now you'll see that the bulk
of the ripple and noise in there is gonna pretty much stay
consistent between these. Now let me try another
coax cable here. It's roughly
the same length, but it's going to be a different type with a different outer weave.
So here's this other one and you can see that it is particularly cleaner, I mean if you put
it near the screen there, there we go, that's why the other one was picking up
so much crap. The weave wasn't as good, the shield wasn't as good
and it's picking up more of that stuff from the screen.
Now, if we disconnect that and we plug in our
scope probe, our proper scope probe -- this is the 350 megahertz one --
which comes with the rigol --
and it's got a X1 X10 switch, so we'll put it on X1, so it's operating just like a regular coax,
and I've got this little coax adapter. It's a bit loose, so please forgive me if
it's-- the connection is a bit intermittent there. I may have to hold it.
There you go, look at that! We've suddenly, with this good quality
properly shielded, high-bandwidth oscilloscope
probe, look! It's not picking up nearly as much.
So our performance has gone from, you know, pretty [ ugh ],
sort of fairly ordinary -- we're stil at the same volts per division,
2 millivolts per division -- but much cleaner with the scope
Now let's put it on times ten and then we have to compensate.
We've got to go in here and we've got to turn that to
times ten and then we're on 20 millivolts per division and we've
got t-- oh, we can't actually go down to 2 millivolts per division
because we're-- we have to be 5 millivolts per division. And it's higher.
Why is the noise higher on times ten? Well, it's
because a times ten oscilloscope probe is higher
bandwidth than it is on times one. And if you don't believe me
here's the spec sheet for it. There it is, times one DC to 8 megahertz.
Times ten, DC to 350 megahertz. This is the spec sheet
for this Rigol probe. And all probes are the same.
That's why a lot of them only come times ten, because they give you the high bandwidth.
It's due to the input capacitance, I won't go into it, that's a whole separate video but.
Times ten probes, that's why they're used, is because they
are higher bandwidth. So effectively we've gone from that 20 meg filtering
on our scope to an 8 megahertz bandwidth
filtering, and that's why our times one probe is
actually going to give us a lower noise measurement,
because it's bandwidth limited, so all that high-frequency noise, wherever
it's coming from, is being attenuated. So really, because the bandwidth
of this power supply is specified from zero to 20 megahertz, we
can't just use a scope probe on times one, because that's only giving us
for this probe, it's only giving us an 8 megahertz bandwidth, so we have to put up with the
fact that we're using a times ten probe. And here's an Agilent one, this is my
500 megahertz high-quality Agilent probe. And that is
5-- we can only go down to 5 millivolts per division because of the times ten but
there you go. That, folks-- you still get the occasional high-frequency
glitch in there, you might be able to see. In fact we can probably even trigger off that.
There we go, yep, we can actually trigger off that.
Occasional little high-frequency pulse which is coming through, but
not a big deal. So there you go, now we can
measure our noise with reasonable
performance. Excellent. So there you go, I hope you
found that interesting. That just goes to show that there's more
to a simple noise measurement than meets the eye. Common-mode noise
go look it up, go research it. It can be a real pain in the *** and a big trap
for young and old players alike. Let me tell you!
So if you like that, please give it a big thumbs up, and if you want to discuss it
jump on over to the EEVblog Forum.
Catch you next time!
[ sparking noise ] captioned by sen