I remember reading the story (perhaps here?) some time ago of the
American who made tape recordings (using captured German recorders)
of Welte piano rolls at the Welte's summer cottage deep in the Black
Forest, and going through miles of red tape(!) to get the reels home
to America to issue as phonograph records.

I believe he made some preliminary recordings by telephone, as for
some reason he could not initially take the recording gear to the Welte
cottage.  I think the Allies didn't want the precious hardware
disappearing into the dark woods.  Of course the sound was terrible,
but he could tell by the nuances that he was hearing true reproductions
of great pianists.  Later he got to take good mikes and the equipment
to the site.  I don't recall anything about the rolls themselves being
taken out of Germany, nor would there have been a need to once the war
ended.

Mr. Welte knew his factory would get bombed sooner or later, since
it was next to the Freiburg rail yards, and he spirited away many
priceless items to his cottage, including the rolls.  But not some
great orchestrions, alas.

About the mercury recording technology, there are several questions:

1. Could the recording playing be played back within a few minutes?

2. How was expression (key velocity) recorded?

3. Could the instant playback include the expression?

My speculation is that (1) could be yes, but (3) is no.  Instant
recording can be done by wiring the keyboard (using any form of
contacts) to a roll perforator, located several rooms (and masonry
walls) away from the piano!  The freshly cut roll could be brought to
the studio and played.

A quieter system could have used 88 (plus pedals) ink-pen recorders
making dots and dashes on a paper roll.  This would be read by a bank
of 88 photocells with vacuum tubes and relays -- a German just might
have been willing to build such a beast. :-)

Now for question (2), which has two parts: Could the mercury contacts
sense expression, and could Welte have recorded it?  If the contact
fingers were indeed of carbon graphite (like pencil leads), then the
farther one was pushed into the mercury, the less its resistance would
be.  If the keyboard had a "spongy" bottom felt layer, the keys could
indeed push farther into the mercury when struck harder, so loud notes
would pass more current than soft ones.

However, this analog variation in current would be hard to record at
all with 1940 engineering, let alone for instant playback.  A more
practical system, used by Ampico and still the basis of touch-sensitive
electronic keyboards that anyone can buy today, is two separate
contacts, one that "makes" early in the key's descent and another that
makes near the bottom.  The shorter the time difference between the
closure of the two contacts, the faster they key and the louder the
note.

Ampico used sparks and presumably very fast moving paper to record
these timings.  Instant playback would not be possible, with or without
expression.

Years ago I read that Welte recorded the expression separately, using
a sliding pen to record the total pressure exerted on the keyboard,
like a seismograph.  (Liszt would have liked that :-)  This graph went
down one margin of the master roll.  This was a guide to the roll
technician who later inserted the expression codings, but couldn't be
played back instantly.

So why the elaborate mercury contact system?  Reliability!  In the
earliest days of electrical pipe organ actions, tiny mercury cup
contacts were used just to get contacts that would close every time,
not wear out, and not build up oxides and dirt that blocked the flow of
current.  The early Hollerith census tabulating machines used mercury
contacts also.  Even today, with gold and platinum and silver contacts,
some pipe and electronic organ keyboards are an "iffy" proposition.

I've also heard that Duo-Art didn't record key pressures at all, but
simply edited the roll codings until the pianist and editor were
satisfied -- much like how I do my MIDI ragtime compositions.  I have
no idea what the Angelus studio did, though from the rolls I once had,
I would guess they used the Duo-Art non-system.

Mike Knudsen

 [ Mark Reinhart presented evidence that the 'seismograph' recorder was 
 [ developed by the Welte-Mignon (Licensee) Co. in America and used for 
 [ recordings (begining circa 1926), and that an "ink master" recording 
 [ machine in New York was used for the Welte-Philharmonie organ rolls, 
 [ but so far no documented evidence has been presented that similar 
 [ technology was in use at the Welte-Mignon studio in Freiburg.  One 
 [ must not confuse the American companies with the Welte firm in 
 [ Germany.  -- Robbie