[time-nuts] Data sheet for the old version of Dallas DS18S20 used in TBol...
GandalfG8 at aol.com
GandalfG8 at aol.com
Tue May 25 21:43:15 UTC 2010
In a message dated 25/05/2010 21:49:30 GMT Daylight Time,
hmurray at megapathdsl.net writes:
I have data sheets for the DS1620 from 053105 and 072099. The text on the
fancy temperature conversion looks the same.
The difference might be at the command/register level. I took a quick
and didn't see any differences.
Is 072099 old enough? Does anybody know what the critical differences is?
If I remember rightly, not always guaranteed these days:-), Mark Sims, the
programmer of Lady Heather, was the first to draw attention to the lower
temperature monitoring resolution using the later versions of the DS1620,
and would probably be best placed to answer your question.
I got the impression from previous discussions here, that all versions of
the IC, at least according to the data sheets, offer the higher precision
facility but something has changed, either in the way it's implemented or in
the command structure, such that the Thunderbolt firmware is no longer
able to switch it into the higher resolution mode.
I've just made a quick search of the archives and these two emails from
Mark, under the subject TAPR Thunderbolt Temperature Reporting Question, were
posted on 26th July 2009 and explain this in some detail.
It's not reasonable to suggest that this only affects Lady Heather's
display of data, rather it's more that Lady Heather is capable of drawing
attention to what might not otherwise be obvious!!
The Tbolt uses the temperature reading to fine tune the oscillator
disciplining. It has a major effect on holdover performance (when the GPS signal
is not available) and some effect on normal performance. You can quite
clearly see the effect of the temperature reading on the DAC setting in plots.
If the sensor produced a temperature reading that was directly proportional
to the true temperature, any error should be minor. I don't think absolute
accuracy is as important as relative accuracy since the Tbolt learns how
the oscillator responds to temperature readings, but your readings are WAY
The fault is almost certainly in the small eight pin DS1620 chip near the
RS-232 connector. There are two versions of this chip. The ones before Rev E
work much better than the later ones because they report the temperature
in a way that can be resolved by the Tbolt firmware quite finely (around
0.01 C raw values). The later chips report the temperature with a very crude 1
degree C granularity. Search the archives for the details.
I found that the chips from:
are Rev C chips and work well in the Tbolt. The chip itself is cheap at
$3.50... shipping is not... it ran me around $11. A group purchase/reshipment
program might be useful. I know of several Tbolts that had bad DS1620
chips. I had several, but wound up sending them out.
Search the archives for discussions on the DS1620 chip issue:
Google "site:febo.com DS1620"
The key to the problem is explained in:
Basically the old rev chips have a high resolution mode of around 0.01C.
The new rev chips have a resolution of 0.0625C, but the Tbolt firmware is
not compatible with some changes made in the new chips, and it winds up with
the temperature quantitized to 1C steps.
The old rev chips had two different ways of accessing the high-res temp
registers. The new chips have only one way. Of course, the Tbolt firmware uses
the method that was eliminated from the newer chips. BTW, Dallas Semi
apparently never told anybody about the changes and their current data sheet
still talks about the old way of doing things.
The rev level of the chip is found at the end of one of the lines (usually
the middle line) of text etched in the package. It will end in something
like E2 or D1, etc. Chips E and above have the crude resolution. D and below
have the fine resolution. All chips from Dallas Semi or their distributors
will be the new rev. The one place that I found older (rev C) chips was
The sensor and oscillator are in that metal box. Yes they have different
time constants, but are fairly well coupled. You are better off isolating the
Tbolt as well as possible from the environment (as long as you don't let
the internal temperature get too high). I have found rapidly diminishing
returns below around 0.1C internal temp changes over an air conditioning
cycle. Below that level, system noise and GPS effects become the limiting
Hope this helps
More information about the time-nuts