Where I’m at
Sitting in the front room of my aunt’s house I contemplated how and if it would be possible to record my beloved old Challen baby grand piano.
With limited time and without the desired equipment to properly record a piano it was going to be a challenge to achieve anything useful.
The piano sits in the corner of a typical suburban front room; shiny wood floors and bare walls with windows overlooking the quiet road and very little in the way of soft furnishings.
No one really uses this piano anymore and as I sat down and played, the detuned notes rattled off the assorted knick knacks, the clunk of the occasional hammer misstrike blending with the harsh room resonance.
In terms of professional standards this set-up was awful, the resonance and rattle of the room, and the reflections from the different surfaces destroy any real clarity in the sound, even if the piano was perfectly tuned and positioned.
I suppose I could have tried to take everything out of the room or something but I don’t imagine my family would have been best pleased to come home to me rearranging the furniture.
On top of this I only had some basic equipment with me not typically suited to recording a piano in any environment, so if it was going to work I was going to have to compromise….a lot!
Download the Free Kontakt Instrument and see the results for yourself.
So what did I have with me?
For microphones I had the choice of a Zoom H4n or a Superlux E523/D Stereo Condenser Microphone. I instantly decided against the Zoom as it had developed some sort of fault that meant it recorded everything with a terrible electrical hum.
The Superlux is a reasonable stereo mic but at £50 ($64) it only has at best a frequency response of 30-20,000 Hz (Fig.1) which means that it’s not great for recording pianos that can have a frequency range from below 30 Hz and extending above 20 kHz.
I generally use it for situations where I don’t mind it getting damaged (out in the rain, chucking it down wells, that sort of thing). As a condenser mic though at least the diaphragm is better suited for piano sounds than a dynamic one. The mic is comprised of a pair of small diaphragm capsule mics in an X/Y configuration which gives a fairly reasonable stereo field.
Fig. 1 Superlux E523/D Frequency Response Curve
I only had a small desk sized mic stand not a full size studio version so getting any sort of height or distance from the piano was impossible. the recording was done through a Presonus AudioBox 22VSL into my laptop. This is a simple interface with fairly transparent preamps that I had with me due to its small size and weight.
In the studio I have a number of DAWs to choose from but on my laptop I was only running Propellerheads Reason 8, which in this instance was fine as it’s quick and simple to use for this sort of recording.
For the further processing I used Audacity with the Izotope RX Dialogue De-noise Plug-in.
A baby grand by its very design is going to create problems when trying to record a decent sound from the low notes. The lower notes in a baby grand suffer from their short physical length which can lead to an increase in the higher harmonics making it much harder to discern the true tuning. So faced with an out of tune and innately difficult baby grand piano, badly positioned in a noisy room, with a single stereo microphone, and no type of proper stand I was extremely limited to how I could proceed. Added to which, the family would be back in under an hour resulting in very little time to set up and record. Speed was of the essence.
I had only 3 choices:
Record at a distance and use the room ambience as an aspect of the sound’s texture. Record close up somewhere around the piano with the lid open, or close-mic centrally behind the music stand.
The room sounded terrible as soon as I played a note at any sort of decent level. Pictures and ornaments set off a cascade of rattling and some of the notes brought out wonderful ringing tones from the room itself.
Opening the lid of the piano only increased the rattling of the ornaments to the right of it.
Running out of options, I placed the stereo mic behind the music stand, facing down into the piano. This negated a lot of the problems with room noise but also drastically limiting the sound.
When placing any mic closer to the strings you’re faced with a couple of issues. Firstly there is the uneven level between the notes; the ones closet to mic are picked up better than the ones further away. Secondly, by placing a microphone higher up away from the strings you give the harmonics time to blend which gives you a smoother sound. Being so close meant that any unwanted harmonics were going to be a lot more obvious.
This, however, is not a totally disastrous close mic position and given all the restraints, produced a decent quality sound for the experiment.
I created a stereo channel in Reason recording the mic through the AudioBox, setting the input levels as high as I could without clipping. The microphone being what it is picked up quite a lot of room noise and an amount of hiss. At this point I did not worry about any filtering or gating so I could monitor exactly how the sound was being recorded.
I proceeded to play each note in order. This is where I hit my first problem. The lowest 3 notes A0, A#0 and B0 were unplayable. I could not discern the tuning between them and all I got was noise. I decided then to begin at C1 (and if necessary pitch it down later if I need it). I recorded C1, playing it at a reasonable velocity. Not so much that everything rattled but loud enough to get the most from the sound, I also let the sound die out naturally by holding the sustain pedal down.
I then did the same with each C note up the keyboard, until I got to C8 which did not work.
Once this was done I copied the recording channel and proceeded to record the rest of the notes on the piano, meaning that excluding the broken notes I recorded 84 of the 88 notes.
As expected the combination of factors described above contributed to some notes having more clarity than others.
Audio Sample 1 is the raw recording of the 7 F notes. It demonstrates the issues that I faced in this situation.
As you can hear, there is an amount of both low end noise and audible hiss. There is also a noticeable difference in the level of some of the notes. This is not down player error, I played each note with the same weight. Some of it is explained by the old mechanisms (that you can hear before some notes). It is also because of the frequency response of the microphone and the overall space within which the recording was made.
The F5 note is audibly louder and clearer than the others, due to this being the only note that created a good response from the environment. The F7 note would not give any volume no matter how hard you hit it.
To further explore what was going on with the sounds I did a preliminary spectrum analysis in Audacity before I processed and exported the rest of the sounds.
Fig. 2 F1 Spectrum Analysis
Fig. 2 shows the spectrum analysis of the frequency spread of the F1 note. The first thing it shows is that apart from a lot of extraneous noise the fundamental frequency is slightly flat (closer to E1 than F1). The second thing is that the 3rd Harmonic is much louder than it should be and is in fact exactly the same level as the 4th Harmonic. This is exactly the sort of issue that I could foresee happening due to the frequency response of the microphone, it’s positioning, as well as the inherent problems associated with baby grand pianos.
Fig. 3 F5 Spectrum Analysis
Fig.3 is the spectrum analysis of the F5 note, it clearly shows how the fundamental frequency stands out by itself with only the 2nd and 3rd Harmonic visible.
What was clear from this was that I would need to do a fair amount of processing in order to make the samples usable.
In Audacity I’ve the Izotope RX Dialogue De-noise plug-in which I decided to test out on the notes, Fig.4 shows 3 spectrum analysis of this trial. The left one is the raw F1 sample, the middle is a heavily processed version, while the right one is the version I chose to use.
Fig. 4 De-Noise Spectrum
The following Audio Samples are the 3 versions of the F1 note: Raw, Over-Processed, Final version.
Having worked out a plan of what to do with the audio I went back to my main session in Reason and set about doing some basic processing on all the rest of the sounds.
Firstly I applied a Low-Pass and High-Pass to help roll off the unwanted low end noise. I generally do this by ear as the spectrum analysis in Reason is pretty inaccurate and cannot show the detail needed to make decisions purely by eye.
Fig. 5 is the Low-Pass and High-Pass filters that I applied to the F1note. I set the High-Pass filter to around 41Hz as the Fundamental Frequency of F1 is 43.65 Hz. The High-Pass filter while cutting the lower frequency also creates a ‘bump’ where it levels out and I used this to give the fundamental a little boost. I dialled in the Low-Pass heavily and then pulled it back to where it reduced the hiss enough without losing the character of the sound.
Fig. 5 High-Pass and Low-Pass Filters
I then repeated the process for each of the notes.
I did not bother using a gate as I decided to cut down each of the notes manually in Audacity as I liked the different amount of mechanical noise before each of the notes. This can be a time consuming process but I deal with making so many individual samples on a daily basis that the processing work only takes a few seconds for me now.
I exported all the notes from Reason as 24bit 44.1kHz Wav files making sure that all the levels were below -1dB putting them into separate folders for each key.
Fig. 6 Izotope De-Noise Plugin
I then processed them in Audacity. The first thing I always do is normalise the samples to -1.7dB which gives me consistent levels for each of the notes. Next I cut the beginning as close to the start of the note as possible and cut any extra length from the end.
After this I applied the noise reduction using a preset I made from a noise sample (Fig.6) earlier. I then manually tweaked each one by ear to get it right.
Once this was done I exported them from Audacity into their group folders.
Overall I was quite happy with the results, some of the sounds were more muffled than I wanted but with the piano lid closed and the mic position I was forced to use this was not the worst outcome.
In Part 3 I’ll go through how I went about building and programming my first Kontakt instrument using these samples.