Tweet
Recently I had a task to digitize 193 cassette tapes.
The tapes were homemade recordings of radio (some FM and some AM) and television broadcasts, .
Most were 90 minute tapes. Various brands. A few were 120 minute tapes.
All were international - music and some comedians - not English speaking.
Quality was low due to the nature of the broadcasts and the quality of the machines used to make the recordings.
There was no need to produce a high quality digital recording, just low bit rate MP3s.
The files just needed to be listenable so the client could survey what was on the tapes quickly and then select any needing better restoration.
After recording a couple of the tapes with my Nakamichi Cassette Deck One, and using DCForensics10 to determine the frequency range of the recordings, I decided to use a high speed method.
First, I did not want to play almost 300 hours of questionable cassettes through a high end Nakamichi.
Second, I wanted to complete the task as quickly and easily as possible.
The solution: Use high speed dubbing on a Kenwood
My Kenwood KXW 4070 has a frequency response of 30 to 19,000 Hz +/- 3dB and High Speed dubbing with auto reverse.
I calculated that the resulting recording would have a frequency response of about 15 to 9500 Hz, more than adequate to capture the audio from this low quality source. (I recorded at 96 kHz/24 Bit just to be sure I kept as much High Frequency data as possible.)
First I determined the actual high speed vs normal speed ratio of the Kenwood using DCForensics10
I used the Change Speed Function in DCF10 to determine the actual speed of the Kenwood in High Speed Dubbing mode:
Published speed per the Kenwood manual is 2x normal.
Manuals are not always accurate when it comes to these type specifications.
So I checked the actual conversion speed change required:
To do this, I measured the same segments of a file recorded at high speed and the file recorded at normal speed.
By Zooming in on two easily identifiable portions of a file about 2 minutes apart, I was able to find the exact zero crossing (keyboard shortcut Q) of the same wave in 2 spots on both files and dropped markers at each zero crossing and selected the resulting segments.
I created two new files from these segments using the Edit - Paste - As New File function
Applying the Starting Pitch and Ending Pitch of 49.94% to the High Speed file gave both files the exact same length to within .01 second
This is easily determined using the Resulting Time on the Speed Change Function in DCF10.
Then I converted the High Speed sample to normal speed
Using the Spectrum Analyzer in DCF10, I could see little difference between the Normal Speed and the High Speed recording after conversion.
I made a preset for DCF10 or Kenwood High Speed reduction to convert the files to normal speed after recording:
I started recording.
The speakers were turned off so I did not have chipmunk voices and chipmunk music to listen to while I did other projects.
Every 45 minutes the Kenwood would stop (90 minute tapes, auto reverse) and I changed cassettes and started a new recording.
Each day, when I stopped recording for the day, I set the Batch Editor in DCForensics10 to change the speed of that days recordings using the preset I had determined. The files from the previous day were converted to MP3s using the Batch Editor on DC8.5 on a differend computer. By morning, these files had finished being Batch Edited.
Total time I spent each day was less than 20 minutes - changing tapes, saving and starting recordings on DCF10, and starting the batch edits.
Since I like to fast forward and rewind each cassette before recording, I placed my Onkyo on top of the Kenwood and used it for that purpose. This saved a lot of time. It only took a few seconds to change the tapes in the Onkyo (double tape deck) and hit Fast Forward. By the time I had saved the file in DCF10, cleaned the head, capstan and pinch rollers, and inserted the next tape and started the recording in the Kenwood, the Onkyo was ready to flip the tapes and hit FF again.
Project finished in just over 2 weeks instead of taking over a month as it would with normal speed recording.
The tapes were homemade recordings of radio (some FM and some AM) and television broadcasts, .
Most were 90 minute tapes. Various brands. A few were 120 minute tapes.
All were international - music and some comedians - not English speaking.
Quality was low due to the nature of the broadcasts and the quality of the machines used to make the recordings.
There was no need to produce a high quality digital recording, just low bit rate MP3s.
The files just needed to be listenable so the client could survey what was on the tapes quickly and then select any needing better restoration.
After recording a couple of the tapes with my Nakamichi Cassette Deck One, and using DCForensics10 to determine the frequency range of the recordings, I decided to use a high speed method.
First, I did not want to play almost 300 hours of questionable cassettes through a high end Nakamichi.
Second, I wanted to complete the task as quickly and easily as possible.
The solution: Use high speed dubbing on a Kenwood
My Kenwood KXW 4070 has a frequency response of 30 to 19,000 Hz +/- 3dB and High Speed dubbing with auto reverse.
I calculated that the resulting recording would have a frequency response of about 15 to 9500 Hz, more than adequate to capture the audio from this low quality source. (I recorded at 96 kHz/24 Bit just to be sure I kept as much High Frequency data as possible.)
First I determined the actual high speed vs normal speed ratio of the Kenwood using DCForensics10
I used the Change Speed Function in DCF10 to determine the actual speed of the Kenwood in High Speed Dubbing mode:
Published speed per the Kenwood manual is 2x normal.
Manuals are not always accurate when it comes to these type specifications.
So I checked the actual conversion speed change required:
To do this, I measured the same segments of a file recorded at high speed and the file recorded at normal speed.
By Zooming in on two easily identifiable portions of a file about 2 minutes apart, I was able to find the exact zero crossing (keyboard shortcut Q) of the same wave in 2 spots on both files and dropped markers at each zero crossing and selected the resulting segments.
I created two new files from these segments using the Edit - Paste - As New File function
Applying the Starting Pitch and Ending Pitch of 49.94% to the High Speed file gave both files the exact same length to within .01 second
This is easily determined using the Resulting Time on the Speed Change Function in DCF10.
Then I converted the High Speed sample to normal speed
Using the Spectrum Analyzer in DCF10, I could see little difference between the Normal Speed and the High Speed recording after conversion.
I made a preset for DCF10 or Kenwood High Speed reduction to convert the files to normal speed after recording:
I started recording.
The speakers were turned off so I did not have chipmunk voices and chipmunk music to listen to while I did other projects.
Every 45 minutes the Kenwood would stop (90 minute tapes, auto reverse) and I changed cassettes and started a new recording.
Each day, when I stopped recording for the day, I set the Batch Editor in DCForensics10 to change the speed of that days recordings using the preset I had determined. The files from the previous day were converted to MP3s using the Batch Editor on DC8.5 on a differend computer. By morning, these files had finished being Batch Edited.
Total time I spent each day was less than 20 minutes - changing tapes, saving and starting recordings on DCF10, and starting the batch edits.
Since I like to fast forward and rewind each cassette before recording, I placed my Onkyo on top of the Kenwood and used it for that purpose. This saved a lot of time. It only took a few seconds to change the tapes in the Onkyo (double tape deck) and hit Fast Forward. By the time I had saved the file in DCF10, cleaned the head, capstan and pinch rollers, and inserted the next tape and started the recording in the Kenwood, the Onkyo was ready to flip the tapes and hit FF again.
Project finished in just over 2 weeks instead of taking over a month as it would with normal speed recording.
Comment