|Sonic Restoration Of Cylinders|
1.1 Choice Of Transfer Path: All-Electric Versus Miked-Horn Playback
The starting point is to transfer the cylinder playback to a WAV file, for processing by computer tools. For me, the choice is between all-electric playback, using my home-made player (Figure 1 below) or using a miked-horn on my Fireside Model-A machine. My preference is always to use my home-made player, as it gives a better result with less things to go wrong (I've had the phone ring during a transfer!)
Figure 1: My Home-Made 2 & 4 Minute Cylinder Player (Revised September-2004)
But the home-made player has one crucial flaw; although it uses a tangential tracking tone-arm to play the cylinder, there is no feed-screw to force tracking. Thus, locked grooves and heavy surface damage can cause the player to skip. Also, the tangential tone-arm only tracks in one direction; it cannot "go back" beyond a few grooves.
Another problem arises with the few 2- and 4- minute US Indestructible cylinders in my collection. To the naked eye, these cylinders (the 2-minute ones in particular) appear to have very shallow grooves. Like celluloid Blue Amberol cylinders, these have a tendency to off-centre molding/insertion of the plaster/paperboard core, and this is sufficient to disrupt playback of some examples.
Half-speed mastering is not currently available to me as an option, so difficult cylinders will have to be transferred using my Fireside Model-A and studio microphone. And, I take the phone off the hook!
With the homemade player I use a Stanton 500 Mk-II cartridge. I don't have a wide variety of styli to choose from; I use:
A Stanton D5127 standard "78rpm" stylus for 4-minute Blue Amberols,
A "2 Minute Edison Doorknob Sapphire mounted on a Stanton cantilever" stylus for 2-minute hard wax cylinders.
Brown wax, white wax & 4-minute Wax Amberols all present playback stylus problems that I'm still thinking about..
1.2 Pre-Amplification Of Low-Level Signal
Path 1: All-Electric Playback Using My Home-made Player (Revised April-2004)
The output of the Stanton 500 Mk-2 cartridge is too low for direct connection to the PC-sound card input. I have constructed a basic phono & microphone pre-amplifier described recently in a local electronics magazine. Unfortunately, the phono pre-amp incorporates the RIAA(playback)* curve, and is not readily modified.
I use the Line Output from the player's flat-equalization pre-amp, and produces sufficient output for the Line Input of the sound card (SB-Live! 5.1). This eliminates any messing about with inverse RIAA curves when I process the WAV file.
Path 2: Miked-Horn Playback Using Fireside Model-A & Studio Microphone
Microphone pickup is fed to a separate microphone pre-amp. No compensation problems exist here, though the signal quality is degraded by the mechanical components of the original player.
This is not the preferred path, and will only be used to transfer difficult-to-play cylinders.
Either way, now we have a transcription of the cylinder it is time to apply processing techniques to remove unwanted stuff such as rumble, surface noise, clicks and crackle.
I have been using Diamond Cut Audio Restoration software tools for some years, but still don't always get it right! I'm always open to suggestions on how to improve my efforts...
2.1 All-Electric Transfer
The initial step is to capture the cylinder recording with the DC-Art32 "Record" WAV file option. My homemade player produces a single channel output, so I store the recording in "mono" mode, 44.1kHz sampling. This aims to have the maximum audio components available for processing.
The Line-level signal output of the transcription player is fed to the Line Input of the Sound Card.
2.2 Miked-Horn Transfer
Most of the transfer problems here are room acoustics, machine gear noise, my neighbor's HiFi, the phone or doorbell...
...On To The Audio Software Tools... (Revised September-2004)
Step1: Using The High Pass/Rumble Filter
This filter is set to "100Hz for 2-Minute Wax, or 150Hz for Blue Amberols. ", in order to remove rumble components mostly due to cylinder off-centeredness. Usually this filter produces a dramatic change in the WAV file, most often for the better. It is arguable that these first two steps should be done in the reverse order, so that high-amplitude rumble components are removed early in the restoration. This could improve the operation of the Impulse Noise filter. I'll have a tinker with this one day...
Step 2: Using The Impulse Noise Filter
This is where I start; removing the clicks from the raw recording. I use the "Cylinder Record Start Point" setting, and leave other settings alone. Results are mixed, with some more noisy cylinders requiring individual "tweaking" of the filter.
After applying the Impulse Noise filter, I examine the recording for big transients, which can generally be removed by either
Re-running the Impulse Noise filter, using the "Big Clicks Only" option.
Tediously & carefully using the "Paste | Interpolate" function on each glitch.
Step3: Using The Low Pass Filter
This filter is set to "7Hz 18dB/octave for both 2-Minute Wax & for Blue Amberols.", in order to remove high frequency noise components.
At this time, if the recording level is low, I use the "CD-Prep | Gain Normalize" function to bring the level up to 0dB.
The sheer size of WAV files precludes uploading them onto the net...the people at "Angelcities" offer vast web space for free, but I'm not going to test their generosity! Instead, I have put the transcriptions up as RAM files, instead of the more-usual RM files. The reason is simple - I've seen net congestion produce a jumpy stop-start-stop again playback and spoil the overall enjoyment. The RAM file downloads the content to the user's PC, and starts the Real Audio player. You can readily save them for later too!
I've settled on encoding Real Audio for "28k" modem, "Music With Voice" codec, "no video". This gives a nice balance between file size and audio quality. Any less, and the sound quality suffers.
Hence, the main purpose for the Real Audio program is to reduce the 20 Meg/4-minute cylinder files and the 10 Meg/2-minute cylinder files to a size suitable for sending over the internet.