[time-nuts] LF power supply noise
xde-l2g3 at myamail.com
Tue Jun 23 14:16:53 EDT 2009
>>> So about the only thing left of interest is histograms of the
>>> jitter. Unfortunately, the 543310A cannot store enough samples to
>>> really make an interesting graph. What I would like to be able to
>>> do is similar to an invention I made for the disk industry long
>>> ago, called Phase Margin Analysis. There is a brief description
>>> on my web page at
>> Somewhere in my map of apps there is a HP appnote for doing the
>> same, to discs, intended for disc industry, back in the days.
> That may very well be a result of my invention, which occurred in
> 1970, was published in 1979, and was copied by IBM in the 1990's.
> But I have all the HP appnotes for disk. I don't recall any of
> them describing what I show above. Can you provide more
I found it.
AN 191-7, 8.34 MB PDF, for the 5370B Universal Time Interval Counter
There are so many thing wrong with that app note I don't have time
to list them all.
First, it is dated June, 1987.
That is 17 years after I invented Phase Margin Analysis, and 8 years
after the Katz & Campbell article was published in IEEE Transactions
By this time, the technique of Phase Margin Analysis was used in
every hard disk company all over the world. You simply could not
design, analyze, or manufacture a hard disk without it. So it was
well known and understood.
But in this case, it looks like some junior applications engineer
got a brilliant idea to advance his career and publish this app
note. He has no clue about magnetic recording and what it takes to
store data on the disk and read it back.
The article was written for floppy disks. The maximum rate of data
collection is mentioned at 8,000 samples/sec.
That is useless. In order to detect the loss in margin caused by a
single bit defect, you must collect data on each and every
transition. For example, see Figs 6 and 7 in the Katz paper:
Thus, 8,000 samples/sec is far below the data rate needed for
floppy, and hopelessly inadequate for hard disk.
The timing measurement described in the app note does not clearly
describe where the data and clock signals come from. Ideally, the
timing measurement should be made at the output of the PFD in order
to include any offsets in the phase detector, and to include the
loop dynamics in the measurement.
For example, the heads on a hard disk flutter as they move across
bumps and imperfections in the surface. This causes a phase error in
the loop that can turn a minor disk defect into a hard error, or
mask an error by shifting the phase in the opposite direction. A
similar effect occurs on floppy disk with imperfections and
variations in the coating thickness. The next time the sector is
written, the timing shifts slightly. So what was previously an
error-free disk suddenly turns into a hard error. Or vice-versa.
For example, see my micro-defect patent at
These issues are not emphasized in the app note, so it is unlikely
the author is even aware of them.
It is not clear how the 5370B is connected to the electronics. Just
blindly hooking up cables could cause all sorts of loading and
grounding problems. These issues are not emphasized in the app note,
so it is unlikely the author is even aware of them.
The Gaussian curve is a parabola when measured on log-linear plot.
Any deviation from a parabolic curve on the side of the margin plot
is an indication of problems that need to be addressed.
The plots in the Katz paper all show the parabolic drop. But the
plots in the app note have any kind of shape you may want, and the
software allows you to select various parameters for the curve fit.
This is wrong.
The plots shown in the app note are useless for making analytical
measurements of the performance of the channel.
The app note is wrong, misleading, propagates false information
about data channels and measuring Phase Margin, and should never
have been published.
I could go on, and will probably kick myself for forgetting some
hugely important issue. But I have to stop and work on several other
issues that suddenly came up.
(Nothing terribly important. My landlord needs some help with urgent
problems in the laundromat and kitchen. He is a nice guy so I try to
help whenever I can, and he counts on me when problems crop up. So
one minute I am analyzing one of the most exotic measurement
capabilities available on the planet, and the next minute I am
buried in a front load washer trying to figure why the drain valve
is stuck open, then fixing a broken tomato slicing machine:)
More information about the time-nuts