After a recent posting, I was asked off line to explain why harpsichords were so difficult to record in the '50s. After I answered this person, he/she said I should post the explanation on line, (I won't reveal who it was but they'll be reading this and they can if they like).
First, let me emphasize that this comment was NOT a criticism of Mercury records, the Puyana recordings are as fine as any harpsichord recordings of their day, (in fact, the tonal range of those recordings is astounding), the criticism was of the state of the art of harpsichord recordings in those days and earlier. As I think most people know, unlike a piano where the string is hit with a felt hammer, and hence it's name, (piano is short for pianoforte, soft-loud), the harpsichord's strings are plucked with a plectrum and is capable of very little dynamic range - depressing the key very slowly or hitting it very hard produces a note of virtually the same amplitude.
Unlike any other instrument, when I first heard a real harpsichord, my first impression was how much it DIDN'T sound like a recorded harpsichord. Until the '70s, the standard level measuring device was the volume-unit, (VU), meter. The behavior of this meter was designed to simulate the response of the human ear. A "VU" is identical in intensity change to a "dB". In the '70s the Peak Program Meter, (PPM) was introduced which measured the peak value of an audio signal - peaks that a VU meter missed. The sound of every note on a harpsichord is preceded with a peak at the point of string plucking which can be as much as 20 dB higher than the body of the note. If you mike a harpsichord and measure the sound simultaneously with a VU meter and a PPM, you can set the volume of the sound so that it's reaching reference level on the PPM but the VU meter is barely moving. These peaks are very important in recreating the sound of a harpsichord but if you
raise the level until the VU meter is responding at reference level, you'll be clipping off these high amplitude peaks, hence the usual sound of recorded harpsichords from the '50s and '60s and earlier. If you record a harpsichord using only a PPM as your level setting device, you'll be amazed how "real" the instrument sounds on the recording, (this assuming that you are listening to it at a level that simulates the level of an actual harpsichord and not at the level you would listen to Mahler's 3rd Symphony).
There were some early exceptions, of course; in my experience, the earliest recording that did a good job of capturing the actual sound of a harpsichord was Bach's "Brandenburg Concerto No. 5", I've forgotten who the artists were, (and I'm currently in Florida away from my records), but this was a DGG/Archive recording in the old cream coloured jackets where each disc was initialed by the engineer and/or the producer. Each disc was also accompanied by an index card with all the info about the recording.