Alt-BEAM Archive

Message #03529



To: "'Jean auBois'" aubois@trail.com, beam@corp.sgi.com
From: Wilf Rigter Wilf.Rigter@powertech.bc.ca
Date: Sat, 22 May 1999 09:47:12 -0700
Subject: [alt-beam] Re: Tricore info question


Hello Jean,

Sorry for the chopped up thread, I have been to busy to follow up quickly on
some earlier comments by Brock and Dennison.

You are absolutely right about the effect of phase shift on negative
feedback. That is the clue!

There are two main causes of HF oscillation:

1. HF oscillation through feedback from poor layout or supply filtering,
which creates an inadvertent HF feedback paths
2. HF oscillation through the normal feedback components that also sustain
stable low frequency processes and is the deliberate feedback path.

What ever the reason, HF oscillation is caused by feedback of signals phase
shifted by 360 degrees within the gain bandwidth of the amplifiers.

HF oscillation caused by poor layout etc can occur in all Nv cores but "main
feedback path" HF oscillation is a stable and predictable process which
always occur in odd numbered Nv cores and which Tilden has (apparently and
aptly) named "supersaturation". It occurs because odd number Nv cores have
180 degree shift from the sum of all inversions which when added to the
phase shift caused by propagation delays provides the required 360 degree
shift for "in-phase" positive feedback.

Type 1 HF feedback occurs mostly in linear devices such as the HC240 will
also oscillate from local feedback of individual inverters when the inputs
are in the linear region.
This is particularly the case with plug-in prototyping breadboards which
have a fair amount of capacitance between sockets. For linear devices, this
HF oscillation is a sinusoidal waveform which can be steady state or occur
in bursts during transition through the linear region.

Type 2 HF oscillation generally occurs only in odd numbered Nv cores and is
strongly depended on the gain bandwidth and type of amplifier. The use of
linear devices in odd number cores requires very careful control of
gain/bandwidth and spurious feedback. I recommend that linear devices should
be avoided for odd number Nv applications.

I do recommend the use of 74HC14 Schmitt triggers which have hysteresis
preventing linear feedback and which are therefore not prone to spurious HF
oscillation making them the best choice for all microcores. One might say
that Schmitt triggers have a limited "phase/space" potential for feedback
oscillation.

This "supersaturation" of HF processes can be quite useful. Take the simple
case of a 3 stage Schmitt trigger ring oscillator with the output of one
stage connected directly to the input of the next. Such a device oscillates
at a characteristic frequency set by the gain bandwidth. In this case, the
frequency is limited by the propagation delays of the 3 stages. For a 74HC
device the HF frequency is around 15 Mhz at 5V and for a 74C14 the frequency
is 3 MHz at 12V. The output is a "squarewave", with each output waveform
shifted by 120 degrees. This is the basic 3 phase oscillator mentioned for
driving the Mighty Mo brushless motors. Of course the natural propagation
delay frequency of 15MHz would drive the motor at several mega RPM
(depending on the number of poles) which is too fast even for the MightyMo
=) .

To control the frequency of a ring oscillator, we have to control the gain
bandwidth of each stage and that can be done by varying the power supply
voltage (Vdd). The TI HC data book gives an example of such a circuit. The
result is a very simple HF VCO which can be used for numerous applications.

Unlike the useful application of the HF VCO, HF oscillation in Nv cores is
generally undesirable and must be suppressed.

For 3Nv cores made from 74HC14 which have the 3 possible valid process
states (0, 1 and HF supersaturation), the simple addition of a small cap (ie
1000pf) from any one Nv bias point to ground is generally enough to control
the HF loop gain for 74HC14 devices and is a simple supersaturation PNC for
Tricore/3 phase generators.

As a general solution dealing with odd number Nv core HV supersaturation as
well as the more familiar LV saturation, the conventional NuPNC or the
inline NvPNC ( as discussed in the Pentacore) are recommended establishing
a single process in the core.

As a foot note:

the number of diodes used in the Pentacore NvPNC to suppress "excess"
processes can be reduced to just one.

enjoy

Wilf Rigter mailto:wilf.rigter@powertech.bc.ca
tel: (604)590-7493
fax: (604)590-3411

> -----Original Message-----
> From: Jean auBois [SMTP:aubois@trail.com]
> Sent: Friday, May 21, 1999 10:28 PM
> To: beam@corp.sgi.com
> Subject: RE: Tricore info question
>
> At 7:10 AM -0000 5/21/99, Wilf Rigter wrote:
> > and 3 processes at very high frequency.
>
> I believe that all of the Nv loop devices can exhibit this high-frequency
> response, a condition that Tilden calls supersaturation or some such.
> Comments have been made by beginners for quite some time about a condition
> where their microcore or hexcore shows all LEDs lit, but dimly. If they
> happened to have touched the chip (presumably a '14 of some sort) they'd
> also find that it gets hot pretty quickly (at least for true TTL there is
> a
> point somewhere between turn-on and turn-off of the output where both
> transistors are mostly conducting for a moment... at high frequency, this
> results in a fair amount of power being used). Confirming what Wilf
> states, the waveform is very high frequency.
>
> This isn't terrifically surprising, however -- an Nv neuron is a
> high-frequency pass filter followed by a pretty fast inverter. The filter
> & the inverter produce just the right amount of phase shift to support
> high
> frequency oscillation.
>
> Hooray for PNCs?
>
>
> jab
>

------------------------------------------------------------------------

eGroups.com home: http://www.egroups.com/group/alt-beam
http://www.egroups.com - Simplifying group communications



Home