Last updated Aug 18th 2003, 10:55pm (Paris time)

Frank Lambert's Talking Clock An 1878 experimental talking machine

Version Française English Version

Contents:

Introduction
Let's discover together!
        1 - Possibly earlier spoken hours
        2 - Spoken bell experiments
        3 - Longer spoken bell experiment
        4 - Maybe "Half past five"
        5 - Hums!
        6 - Yet another One o'clock?
        7 - Another voice?
        8 - Another O'clock?
Conclusion
Contacts

Introduction:

Many people forgot or ignore what preceeded disc records (CDs, vinyl, 78s, etc): wax cylinders, tinfoil phonographs, etc. Some others remember their grand-mother having an old cylinder phonograph, but know little or nothing about it. Some others collect phonographs, 78 records, wax cylinders and other surviving items from a past that did not know computers and laser. Why? Because they represent a step in the evolution of the sound recording and reproducing technology.

I have always been interested by these old phonographs. I first saw them in a 1981 exhibit (I was only 6 years old), and they were all impressive. Large 14 inches records turning at 120rpm and shouting old popular songs, small machines using cylinders instead of records, etc. Time has passed, and I became another phonograph collector.

I first heard of the experimental Lambert talking clock machine in 1999 after doing a small research about tinfoil phonographs. I never heard of such a machine. Dating only a few months after the first Edison phonograph, this machine yet used a permanent recording method that consisted of a lead cylinder (instead of a fragile tinfoil sheet placed around a grooved mandrel). The article I read was written by Aaron Cramer and Allen Koenigsberg (read part 1 and part 2) and I strongly suggest that you read it to know more about this unique talking machine.

The recording still present on the lead cylinder was not public until Aaron Cramer decided to make it available on the internet. Sound archivist Glenn Sage made the audio transfer with Aaron. Thus, they wrote a special page about the recording. Thank you so much to both of them for letting us hear the earliest recording in the world!



Let's discover together!

If you read the page written by Glenn Sage and Aaron Cramer, you will discover that the recording contains several parts. One is clear and contains spoken hours. The rest was, as of february 2002, not yet understood. Hence why both Aaron and Glenn asked if anybody could identify anything else than the section with spoken hours. As a computer engineer, I thought about editing the audio in order to discover other sections. Using my PC computer and a sound editor, I could identify at least three other sections.


1 - Possibly earlier spoken hours

Gleen Sage's page contains several audio samples, one of which is called fourth section, reversed (MP3 format). This is the section I started to analyze as Gleen indicated that it could have been recorded in reverse.



This fourth section is located between 1:28.00 and 1:34.00 in the complete sound recording. Once extracted and zoomed, its wave form (as shown by the editor) clearly indicated that the sound contained words. I listened to this section in reverse and discovered something that sounded like words spoken periodically. I replayed this interesting section several times and instantly thought about what Frank Lambert could do back in 1878. If he recorded spoken hours, why not having done so several times as experiments ? I rapidly found what this small section contained: another set of spoken hours!

Two O'clock... Three O'clock... Four O'clock... Five O'clock

Two O'clock and Three O'clock were clearly hearable, though Three O'clock seemed to play a very little bit slower than Two O'clock.
Four O'clock was still hearable but played at a slower speed.
Finally, Five O'clock was hardly hearable, though the rythm of the words was same than Five O'clock.

Two interesting details took my attention:
a. The sound volume increases
b. The recording speed decreases

My theory is simple: Frank Lambert tried to find the accurate distance between his mouth and the recorder (thus changing the volume) and the accurate speed to make a correct recording. Which means: he recorded at an increasing speed while his mouth approached the recorder.

It is obvious that this theory may be false and I do not consider it as definite. However, there are chances that the volume and speed variations be an experiment made by Frank Lambert before recording the spoken hours that are clearly hearable on the recording.

Using the sound editor, I extracted the small section containing these earlier spoken hours, equalized it to remove low and high frequencies, and then progressively pitched it up to make it play at a nearly constant speed. I did not investigate a strong time in this first research, so the results are not the best I could get. Still, they make the recording easier to understand.

2 - Spoken bell experiments

Another history cought my attention while trying to discover other sounds. I recently acquired a copy of Tinfoil Phonographs by René Rondeau. On page 71 I read the following:

Reading that Frank Lambert thought about making the clock say something instead of raising a common alarm bell, what if the recording still contained parts of such speeches? This is what I tried to find.

I first found a small section between 1:20.70 and 1:23.30 in the whole recording. This section looks like it contains Five o'clock! Time to work! recorded in normal way. Although it is unsure that the speech is correct, the "Five o'clock" makes is an amazing coincidence with the Edison quote listed above.

Aaron Cramer gave another opinion with different words (sorry, I lost the text), but at least we both thought about the same style of speech.

3 - Longer spoken bell experiment

In the same idea, it is also possible that Frank Lambert imagined a longer type of spoken alarm. Another interesting section is located between 1:11.60 and 1:19.70 in the complete recording. When played in reverse, it sounds like a longer sentence, more agressive than Time to work, but with more humor. I also extracted this section and tried understanding its meaning. So far, here is what I can hear:

00:00:00 - 00:02.80 = Come on man! Get up! Get up!
00:02.80 - 00:04.35 = It's time to put up or It's time to put on, not clear...
00:04.35 - end = Ooooh, it's time for [indistinct words] aaaannnnd woOoOork !

The first part is a bit difficult to understand, but after listening it several times, the words can be cought. The second part is not very clear, especially for a french guy like me, so I have put what I could hear. The third part is also difficult to understand, but its last words are easier to catch than the first.

Although this section is difficult to understand, it is clear that it is another attempt to a spoken alarm. Obviously more agressive than the previous, but still with a certain humor!

4 - Maybe "Half past five"

I have found another interesting section between 0:19.45 and 0:21.61 in the complete recording. It has clearly been recorded in reverse at lower speed, so it must be pitched down by 20%. The words are difficult to distinguish because of the surface noise, but after listening several times, I could hear Half past five spoken slowly with space between words.

5 - Hums!

The 21 first seconds of the whole recording appear to contain several sections. They can be clearly separated by hearing a decreasing "hum" which marks the moment when Frank Lambert started rotating the cylinder for recording. Let's give an example and explain it...

Click to hear a "hum" example (located between 0:09.04 - 0:10.01, played in reverse)

This "decreasing sound" effect is easy to understand. On the cylinder surface, the first curves of the recorded sound take very few space since they were recorded at a low speed. In the other hand, the last curves take more space as the speed increased. Playing the recording at a fixed speed will mean playing less and less curves during a same time, hence a decreasing sound or frequency. This is what happened on this Lambert recording. The contact between the lead sleeve and the recorder generated small vibrations. When Frank Lambert started recording (i.e: rotating the cylinder by hand), these vibrations got recorded at an increasing speed until the speed got constant. When playing the recording back, its begining played with a noice (these vibrations) that appear to be decreasing.

Why talking about vibrations then? Simply because if you can catch them, you have just found where one recording started or ended! When I realized I could identify recordings this way, I thought about trying to find more of them...

Try yourself! Download the complete recording, convert it to WAV, open it with a WAV editor and see if you can find anything!

There's another hum between 00:06.90 - 00:07.30. This one is shorter. I don't know if it was recorded at the begining or the end of a recording. As a matter of fact, a recording could begin with a "decreasing hum" and end with an "increasing hum". In other words, the two hums listed above would delimit a single recording. Or they could be part of different recordings. Who knows...

6 - Yet another One o'clock?

Following my research around "hums" in the first 21 seconds of the complete recording, I think I have found another section containing "One o'clock". It is located between 0:16.44 and 0:17.40. The sound has been pitched down by 10% and the words are very difficult to catch. I am absolutely uncertain that they are correct.

7 - Another voice?

Another interesting point is the possibility that a different voice from Frank Lambert be recorded on this cylinder. As a matter of fact, Frank Lambert was not working alone, and other people might have recorded his voice accidentally or willingly between 1878 and the date when Aaron Cramer acquired the machine. Although there is no way to know this for sure, Aaron think he is the only owner beside Frank Lambert. "As the mandrel was frozen when I got it, I would guess it hadn't been rotated in more than 50 years, and the dealer had it less than a month" he says.

A small portion located between 1:33.40 and 1:34.40 in the complete recording seems to contain the words Your Attention!. Recorded in reverse at a higher speed (around 20%), it is located just before the earlier set of spoken hours (described in section 1), and seems to have over-recorded the "One O'Clock" of this earlier hours set. This may be an attempt to announce the hours (i.e: "Your Attention! Two o'clock!"). Alghouth these words may be correct, it may be impossible to determine whose voice is there... Edison?

8 - Another O'clock?

Section 01:27.30 - 01:28.00 in the complete recording seems to contain the words O'clock although it is hard to determine if the words are correct and in which way they were recorded. They sound almost same in both ways...

Conclusion:

Even though I have discovered and/or located different parts in Frank Lambert's recording, there is still a lot to discover. Not only because some of the sound I found is not clear or partial, but also because we need to understand why one sound or another has been recorded, in which context, for what purpose, etc.

The following graphics illustrate what I have found so far in the recording.
Blue areas either contain nothing or have not yet been understood.
Black areas contain no speech.
Each area has an arrow showing the way it has been recorded, as well as the number of the section where it is described on this page.

00:00.00 00:09.99

00:10.00 00:19.99

00:20.00 00:29.99

00:30.00 00:39.99

00:40.00 00:49.99

00:50.00 00:59.99

01:00.00 01:09.99

01:10.00 01:19.99

01:20.00 01:29.99

01:30.00 01:40.40

Contacts:

Aaron Cramer's web site
Glenn Sage's web site
David Winter

The pictures of the experimental talking machine are the property of Aaron Cramer.