[SI-LIST] Re: resend - Specctraquest model: mounted inductance
- From: "istvan novak" <istvan.novak@xxxxxxxxxxxxxxxx>
- To: <bart.bouma@xxxxxxxxx>
- Date: Sun, 8 Jun 2003 22:34:46 -0400
Bart,
A few comments:
- the reference impedance of the measuring circuit should not make
a big difference. The 0.1-ohm reference impedance you quote from
the paper was in a specific cicuit: when we choose to use a
transformer isolation to open up the ground loop of cable braid.
This matters only at low frequencies AND if you have transformer
isolation. With no transformer isolation (e.g., with differential-amplifier
isolation at low frequencies) the reference impedance should not
matter over a wide range. The simplest "through" calibration sets the
impedance reference to 25 ohms, though you can equally elect to use
a calibration piece of say 10 ohms, 1 ohm, 0.1 ohm or any other value,
which will set the top reading of the VNA to the selected reference
value. Note also that in the referenced paper the 0.1 ohm was only
a calibration reference, it does not in any way change the inherent
impedance of the VNA connections, in other words by selecting a
calibration reference impedance other than 50 ohms does not change
the impedance of the connection environment presented to the DUT.
- the problem Larry referred to with the example of 100A current:
if it needs to be distributed on a PCB, it requires planes or
plane shapes as opposed to traces. If you have a trace to feed a device,
chances are better that a capacitor such as the X2Y part can be hooked
up properly to provide good transfer isolation. In case of planes, however,
the signal 'can go around' the capacitor, so all what matters is what is the
impedance that is being provided by the capacitor to the planes. This
involves all the vertical connections from the capacitor to the planes.
Once we realize this fundamental requirement, we next have to answer
the question: how the X2Y capacitor would be different in this vertical
connection compared to a two or multi-terminal conventional capacitor.
So far I have been unable to convince myself that the X2Y capacitor
would behave differently in applications where we have to hook up the
part to inner plane layers: what is presented to the planes at high
frequencies is determined by the current loop formed by the planes,
vertical vias and bottom structure of the capacitor. Whatever is above
the lowest layers of the capacitor has little influence on the behavior.
Note that this by no means wants to imply that the X2Y capacitor is
not great in applications where we can hook it up to traces or shapes
that will make full use of its internal construction.
- regarding the preferred model of the X2Y capacitor: probably
more than with conventional capacitors, the model should be
application and geometry specific. For high-power bypass applications,
this is the two-terminal impedance presented to the planes. In applications
where the three terminals are in fact can be connected to three different
electrical networks, a PI network may be suitable.
Best regards,
Istvan Novak
SUN Microsystems
> > ----- Original Message -----
> > From: "Bart Bouma" <bart.bouma@xxxxxxxxx>
> > To: <larry.smith@xxxxxxx>
> > Cc: <si-list@xxxxxxxxxxxxx>; <si-list-bounce@xxxxxxxxxxxxx>
> > Sent: Friday, June 06, 2003 9:40 AM
> > Subject: [SI-LIST] Re: resend - Specctraquest model: mounted inductance
> >
> >
> > > Larry,
> > > I'm involved in X2Y for almost 2 years now.
> > > My impression of X2Y is that they were and are very willing - maybe
> even
> > > eager - to share every technical detail with everybody.
> > > Patents did go through and still run. (As you maybe know,
Yageo/Phycomp
> is
> > > a licensee of X2Y).
> > > Back to the discussion:
> > > => " .. one of the responses is low and flat over a large frequency
> range,
> > > indicating a lot of insertion loss. .... .... , but no DC current will
> > > flow."
> > > You're absolutely right, this measurement shows that there will be
> minimum
> > > crosstalk between two powerplanes. We noticed the same high-frequency
> > > performance for this attachment mode and the A&B shunt mode. In the
> latter
> > > both Y-caps are tied together, i.e. connected in parallel.
> > >
> > > => The intrinsic ESL of a 0306 is indeed about 1/3 that of a standard
> 0603
> > > cap. We measured aprox 180-200 pH for our 0306-220nF and 0306-470nF.
> > > I noticed a lower ESR too, but not 1/3 that of a normal geometry
> > > capacitor. I tend to say about 2/3 of it. But it strongly depends on
> > > internal structure etc.
> > > As you mentioned, mounting inductance is very important. It easily can
> > > dominate the total inductance.
> > > An experience from my former job (dev. TV and Satellite tuners): for
> some
> > > types of tuners we were using pads in vias, until our PCB-supplier
> forced
> > > us to stop using this. We encountered big problems after positioning
the
> > > vias "far" away from the capacitor pads, due to the increased mounting
> > > inductance.
> > > Sometimes we had to use 2 or even more vias.
> > > The use of multiple vias is also valid for the X2Y ground terminals.
One
> > > via at each gnd terminal, thus one at each side of the X2Y, improves
the
> > > insertion loss performance by more than 15 dB, i.e. much more than can
> be
> > > expected from just paralleling two vias! One would expect a
> theoretically
> > > improvement of 6dB.
> > > X2Y describes it as follows: the X2Y changes from a series device
(like
> a
> > > standard MLCC) into a device that is parallel to the powerplanes!
> > >
> > > Please look also at following App.Notes at : www.x2y.com / The
> Technology
> > > / Application Notes.
> > > App. Note #1001: this note shows the above mentioned effect of
> connecting
> > > one or two grounds.
> > > App. Note #1006: discusses Circuit Configurations.
> > >
> > > When decoupling capacitors are used in a 1 Ohm or lower environment
like
> a
> > > PDS, is it useful then to measure these caps in a circuit with a 1
Ohm
> > > reference impedance? Like the way Istvan Novak did ("Measuring
> milliOhms
> > > and PicoHenries in Power Distribution Networks" presented at
> > > DesignCon2000). In this paper he is a.o. describing a method to lower
> the
> > > reference impedance to 0.1 Ohm by doing a 0.1 Ohm through connect
> > > calibration i.s.o. using the system's 50 Ohm. Is there anything that
> > > speaks against measuring in a 50 Ohm system, and calculate the
Insertion
> > > Loss back for a 1 Ohm system? Assuming the VNA's dynamic range is
> > > sufficient.
> > >
> > > I read the ECTC_2001 and ECTC_2002 papers some time ago, and looked
> > > through them again.
> > > ESR increasing past series resonance, I can understand this. I assume
> this
> > > is caused by the skin-effect.
> > > What I don't see, how does the inductance decrease? What's the
> mechanism?
> > > But this phenomenon is indeed ideal to suppress the "anti" resonances.
> > > Sometimes parasitics are useful!
> > > Hi Q-caps (low ESR caps) do show sharp and high parallel resonance
> peaks,
> > > which is unwanted. On the other hand high ESR caps are unwanted too.
> > > Lower inductance is of course the best to avoid, or at least to
> > > minimize,such resonances. For this reason a lot of people are using
> > > capacitor arrays which exhibit lower loop inductance - including
> mounting
> > > inductance - when paralleling two capacitors.
> > > I noticed you mentioning the mounting impedance in the transmission
line
> > > model in fig.3 of the ECTC_2002_caps.pdf document (R-mount and
L-mount).
> > >
> > > This brings me back to my original post: is there a method to extract
> the
> > > mounting inductance for this X2Y?
> > > Can the following method be justified?
> > > We are able to measure the intrinsic inductance of the X2Y in our
> > > 3-terminal ICM-fixture, this fixture can be de-embedded by a
> > > TRL-calibration. Say we find a value "ESL-intrinsic".
> > > When we subtract this intrinsic value from the inductance value we
find
> on
> > > the pcboard (ESL-pcb) as described in the document mentioned in the
> former
> > > mail (X2YMLCC.pdf), do we then get the inductance associated with
> > > mounting?
> > > For the PCB value will show the part's own inductance together with
the
> > > pcb's (mounting) inductance.
> > > Thus: ESL-mount = ESL-pcb minus ESL-intrinsic .
> > > This must then be done for all attachment modes, I suppose.
> > >
> > > best regards,
> > > Bart Bouma
> > > Appl. Eng.
> > > www.yageo.com
> > >
------------------------------------------------------------------
To unsubscribe from si-list:
si-list-request@xxxxxxxxxxxxx with 'unsubscribe' in the Subject field
or to administer your membership from a web page, go to:
//www.freelists.org/webpage/si-list
For help:
si-list-request@xxxxxxxxxxxxx with 'help' in the Subject field
List archives are viewable at:
//www.freelists.org/archives/si-list
or at our remote archives:
http://groups.yahoo.com/group/si-list/messages
Old (prior to June 6, 2001) list archives are viewable at:
http://www.qsl.net/wb6tpu
Other related posts:
