So the first suggestion was to provide some basic low pass filtering and dynamic compression of unwanted frequencies. The next two pages on noise reduction are based on using the Fast Fourier Transform (FFT) based continuous noise filter (CNF) and are a little tricky to construct in a comparable manner so you will have to apply your own judgment and accept the results for what they are.

The biggest problem with broad band noise reduction using a FFT based CNF is the balance between loss of high frequencies and the generation of echo/ringing artifacts in the final recording. These generally occur where there are sharp transient responses brought about by cymbals as well as residual impulse noise/distortions! The greater the noise attenuation, the more likely these artifacts will develop.

Is there an advantage in a stereo recording of a monaural 78rpm record?

Below is a series of simple noise reduction tests involving broadband white noise generated using the "make waves" function in DC8. The image shows a series of single FFT passes and their effect on white noise attenuation when applied to the original stereo or later monaural file. The tests were as follows:

1. A stereo white noise file 30s duration was created with amplitude as shown in the image (yellow trace).
2. A monaural noise file was created from the original stereo noise file using the "File conversion" L+R setting (6db attenuation on both L and R channels - result is same as yellow trace).
3. A single 2048 FFT pass with 18dB attenuation was applied to the above (2) original L+R monaural noise file (orange trace).
4. A single 2048 FFT pass with 9dB attenuation was applied to L and R channels of the above (1) stereo noise file independently then the result converted to monaural form (red trace).

If that isn't interesting enough, then you may want to have a listen to the final results as well!
Although these tests are using default values for attack and release as well as using white noise, I have tried this using the lead-in groove of a 78rpm record and obtained very similar results:

So there may be a benefit in running the CNF on the Left and Right channels of the recording before converting it back to the original monaural form. There is also the advantage that you can taylor the CNF threshold to the different noise thresholds of the Left and Right sound recordings. Often one will contain more surface noise than the other - using the above method, you can suppress the noise in that channel a little more by bumping up the CNF threshold and the attenuation a few dB or more.

From my experimentation I have found the following to work best if using this step:

1. Use the "Channel Blender" to blend frequencies below 1kHz to mono. - This reduces leftover low frequency rumble.
2. Run the CNF independently on the Left then Right channel, taking the noise threshold from the lead in groove of the respective channel.

I would recommend trying to reduce the surface noise 6-9dB for the best (lowest noise) channel and 9-12dB for the worst. The aim is not to remove the noise, but reduce it without impacting on the tonal qualities or introducing any artifacts. This way, when the Left and Right are combined, you get a greater noise reduction effect over just combining the channels. However, keep in mind that sometimes one channel may be so noisy it may be useless and better left out.