I’ve just started a new project to produce a virtual tour of our local fully restored water mill. I’m starting with the location sound recordings and will move on to the images and video clips in due course. My Zoom H4 digital recorder has done its usual magnificent job of capturing all the relevant sounds on each floor of the mill (a 4-floor building). There’s just one problem (isn’t there always!). When the mill was restored in the 1970s, the land owner had already made changes such that the mill pond and leat were no longer available. The restoration team therefore installed a heavy duty electric pump to lift the water from the wheel pit to the header race that feeds the water onto the wheel. Thus the mill is still water-powered, but the water is electrically assisted.
The mill is somewhat unusual in that the wheel chamber has been built over (in historic times) and is now totally enclosed. The pump is installed in the wheel chamber. When recording the sound of the wheel turning under water power, I also pick up a horrible “whine-cum-roar” from the pump. It is possible to record the wheel after the pump has been switched off, but the wheel begins to slow down after just few seconds, as the water supply runs out.
My questions is this: is there anything I can do to reduce the aural intrusiveness of the pump sound?
Did you get those right? I didn’t hear any difference and I have a good sound system. The obvious tool is Noise Reduction. All you need is to get a clip of motor by itself and you’re golden. No, I don’t think that’s likely, either. If you don’t do that, anything else you do is going to destroy the water sound. I am in a particularly advantageous position to help you here. You could add a little Hollywood.
Where is it written that the sounds have to come from the mill? Take your H4 out and record sounds that taken together could have come from the mill when it was properly operating. People here win awards for doing that convincingly.
Actually, the getting of the clip is entirely possible. I could capture the electric motor running for about 4-5 seconds all on its own, before the water cycles around to the wheel if I go to the mill before they open to the public. But is a recording made on a different day going to provide an adequate Noise Profile? My own, uneducated assessment was that the motor “whine” was within the same frequency band as the sounds of the rotating wheel and splashing water - and that Noise Reduction may not prove effective.
I’m expecting a watermill to sound like vigorously splashing water possibly pulsing slightly as the paddles come around. Maybe not. Add in wood-on-wood groaning and squeaking sounds and possibly stone-on-stone when the scene shifts to the actual wheels doing the work. The trick is to divorce your head from the associations between the picture and sound – and then put them back together manually in a way they would logically be. Nobody mics a guy walking on a gravel road. That’s all done in post production or Foley. Sometimes it doesn’t really make any sound at all, but everybody is expecting it to.
There’s another industry secret. All you’re really worried about are the three people that have ever been inside a water mill and remember the experience. Everybody else will take your word for it. You let your closest friends in on the secret of how you shot the whole thing at 1300h on Tuesday in the car park with a bucket of water, two wet bread carving boards and a rock.
This water mill is just a 10 minute drive from my home and they were open to the public again yesterday. So I went back to record the electric pump as it started up. I’ve tried using this recording to give the noise profile and then applying that to the previously recorded track via Noise Removal. As I rather suspected, the result is far from usable. However, I have played around with the recording from the previous day when I captured the few seconds of wheel noise after the pump was switched off, and before the wheel starting to slow down. Using a combination of Duplicate, Timeshift and occasional Reverse (to break up the pattern a little) I have created a suitable background track which, although it does repeat just six seconds of sound over and over again, will probably do for what I want.
Thanks for all the advice. Even though I have eventually gone down a different route, your thoughts on the subject were most welcome in helping me arrive at my decision.
Thanks for that tip. My next question was going to be: How did you identify the frequencies to remove? As I thought about it, I realised you must have done a Frequency Spectrum analysis. So I ran one and noticed the two sharp spikes at the frequencies you quoted. I then discovered that, with both my recordings imported into Audacity and with the Spectrum plot window open I could, in the main Audacity window, make another selection on the other track and run the Spectrum analysis a second time. Unfortunately, the same Spectrum window was re-used to show the second plot. Wouldn’t it be nice to have the plots use unique, dedicated windows? Then one could resize and reposition the two Spectrum plot windows so that one could visually compare the plots “side by side”. I feel sure that this feature would be reasonably easy for the programmers to code. Does it already appear among the requests on the Wiki? If so, please add my vote to it
I’m not sure if it’s on the wiki, but I remember a similar suggestion which was to show multiple plots superimposed on top of each other. Cool Edit Pro (now Adobe Audition) had a feature like this and it worked well. This sort of thing:
I’d welcome anything that allowed the Spectrum plots of two tracks to be displayed at once so that they could be visually compared. I would have thought it would be an extremely useful tool when trying to do Noise Reduction.
Multiple spectra window: > (6 votes) Plot multiple spectra for different tracks/regions in the same project e.g. Audition has four buttons to hold spectra. Currently different spectra must be in different projects.