Re: Reference voltage clarification needed

Arpad and IBIS folks:

I agree with you that this needs to be clarified. I support the notion
that the [Voltage Range] or any of the Reference voltage values should
be set to the actual supply voltages that the part is to be operated at.
While it is possible to generate correct curves by using other values, the
model itself should contain values that are the operating point values.
As IBIS grows, more and more data will be related. Already, the [Rising
Waveform] and [Falling Waveform] tables are related to [Pullup] and
[Pulldown] data and are not totally independent. So, in the last example,
while the [Pullup] table could be correctly presented with respect to a 0 V
reference as might occur during an actual extraction, the safer IBIS model
should be build using the expected operating supplies as the references
So the [Pullup Reference] or [Voltage Range] should be set at 5 V and the
[Pullup] voltage data shifted by 5 V if it were captured using a 0 V reference.

Bob Ross,
Interconnectix, Inc.

> Date: Thu, 15 Feb 96 11:06:00 PST
> From: Arpad Muranyi <Arpad_Muranyi@ccm.fm.intel.com>
> Message-ID: <Thu, 15 Feb 96 11:14:57 PST_6@ccm.hf.intel.com>
> To: ibis@vhdl.org
> Subject: Reference voltage clarification needed
> Status: RO

> Hello IBIS folks,

> As I am working along on our IBIS model maker tool, I discovered another item
> that I believe needs clarification in the IBIS specification.

> We have four keywords in the IBIS spec which define the reference voltages of
> the pulldown, pullup, GNDclamp and POWERclamp. At the time these were conceived
> they might have been obvious, but reading the IBIS spec now I am confused as to
> how these must be interpreted. After talking to Bob Ross, I understand how we
> intended them to be used, but this is not clear from the spec alone.

> Examples:

> Connect a 5 volt CMOS device to a FLOATING supply with 5 volts between its (-)
> and (+) terminals.

> If I measure the pulldown I-V curves so that the GND connection of the curve
> tracer is connected to the (-) terminal of the floating supply, my Pulldown
> Reference is 0V. Similarly, if I connect the GND terminal of the curve tracer
> to the (+) terminal of the FLOATING(!) supply and measure the I-V curves of the
> pullup structure, my Pullup Reference is 5V. In other words, in this manner the
> refence voltages are derived with respect to the (-) terminal of the floating
> power supply.

> In another situation I am measuring an RS232 driver. The device is connected to
> three floating supplies so that it gets the necessary voltages of -10, 0, +5 and
> +10 volts.

> Since I know that the pulldown is associated with the -10 volt terminal, and the
> pullup is associated with the +10 volt terminal, I will connect the GND terminal
> of the curve tracer to -10V and +10 to obtain the pulldown and pullup I-V
> curves, respectively. The Pulldown Reference is -10V and the Pullup Reference
> is +10V in this case. Not that these numbers are all with respect to the
> internal ground, 0V. If I did not know what that was, I would end up using a
> different set of reference numbers (such as 0V and +20V), but the difference
> between Pullup and Pulldown References will still give me the Voltage Range
> (=Power Supply) voltage of the output stage of this driver. (This is why the
> Voltage Range keyword is somewhat redundant and is not required if all four
> Reference keywords are present).

> However, the reference keywords could be interpreted another way. Let's say
> someone wants to curve trace a pullup structure in a ground-relative (and not
> Vcc relative) way. In that case we have a device in high state, connected to a
> 5 volt supply and the GND terminal of the curve tracer is connected to 0V. So
> the Pullup Reference = 0V in this case. At the same time, the Pulldown
> Reference is also 0V, because the pulldown structure of the device was also done
> in a ground-relative manner. Even though the device was powered up with 5V, the
> difference between the two Reference voltages is 0, and value for the Voltage
> Range is lost if it is not explicitly defined.

> >From this, I conclude that 1) the probing device must always be connected to the
> same voltage to which the measured structure is connected to (this might not
> always be known, though), or 2) we cannot get rid of the Voltage Range keyword
> even if all four Reference keywords are defined.

> There are many fine details and implications associated with this subject which
> I cannot explain here. The main thing is that the spec is not clear on which
> interpretation is the correct one regarding the Reference keywords. Or did I
> miss out on something?

> Arpad
Received on Thu Feb 15 17:11:53 1996