I've been a Minidisc enthusiast for over two years and am dismayed by so many negative comments about LP4 recordings. (For instance, there are two disparaging reports on Minidisc.org)
Firstly my credentials. As an "oldtimer" I've had a lifelong interest in Hi Fi having, from the 1950s, built my own amplifiers, speaker enclosures, et al. In fact I still design and build enclosures. Now retired, I was a radio & television broadcast engineer with the Australian Broadcasting Corporation in Melbourne.
I own two Sony MDS-JB940s and an MZ-G750 and have carried out a number of recording and listening tests. I think we all agree that ATRAC DSP (Type-R) at 292 kilobytes (standard stereo mode) is completely transparent, regardless of the type of input signal - a great achievement! There seems to be general consensus also, that LP2 produces a minimum of sound degradation.
Using a MDS-JB940 I conducted several A-B blindfold tests using a number of different listeners. Each listener was unaware of which source they were listening to when I switched between a high quality source (usually exemplary CDs, but sometimes LPs) and the Minidisc replay of same. I carefully matched the output levels of the two sources. (A louder signal can sometimes sound better than a softer even if from the same source.)
No listener consistently picked which source was the original and which the Minidisc replay, and general opinion was that there was no significant difference between the two in LP2.
With LP4 and carefully chosen originals the same conclusions (remarkably) as with LP2 were reached. (One understandable comment was that the stereo image was not quite as wide.) So why was I reading poor reviews of LP4?
I measured the frequency response in LP4 mode and it was ruler flat from 20Hz to 20kHz. The noise was almost 100dB down (excellent), and although I couldn't measure distortion, listening on tone I could hear no sign of harmonic distortion - it must be very low. Remarkably, in view of the "joint stereo" in this mode, there was no interchannel cross-talk when recording on one channel alone.
However, with certain types of program source, there is a difference - sometimes quite dramatic. After much experimental comparisons in LP4 , I have come to the following conclusion: LP4 does not like non-mono-compatible sound sources!=20
The replay from a pure, high quality mono source recorded in LP4 sounds exemplary. However, when I recorded a mono source from a stereo cassette deck, the LP4 replay sounded "phasey" and with a loss of certain high frequencies (comb filtering effect). The reason was that the heads on this cassette deck were out of alignment causing a phase difference, or small time delay, between left and right channels.
I found that some of my stereo CDs, when switched to "mono" mode on my amplifier, suffer a loss of highs and/or some comb filtering. Recording these in stereo in the LP4 mode created the same "phasey" sound as before. When I say 'phasey', I mean that the direction from which the sound is coming is obscure - ethereal might be a good word. If you're copying from a cassette tape and the tape is slightly buckled there will be a phase drift between left & right channels. This gives a "wandering' (truly ethereal) sound on LP4 replay.
However, in practice, non-mono-compatible sound sources are fairly rare. What I always do now is monitor the sound source in mono before recording in LP4. If there is little or no difference in sound quality change when you do this, then the sound quality of the LP4 recording will be good. To conclude: in my opinion, given a good quality mono-compatible source, LP4 is capable of producing an excellent recording.
I advised Ken that simple tone tests will not give an accurate measure of frequency response characteristics of a perceptual coder, and that listening through a pair of good headphones would give a better picture of LP4 mode. He replied:
I finally got around to making some more LP4 tests. Firstly, I agree that the tone tests would not give a true indication of how the encoder works on music, etc., but I wanted to prove that the bandwidth in LP4 was not deliberately restricted in order to reduce the number of decisions that the coder has to make at such a low bit-rate. In other words it's capable of reproducing the full frequency range.
I have just recorded, on a MDS-JB940 via the co-ax input, some white noise at -3dB in LP4 from my Denon Test CD. On replay, when I adjusted the output levels for an accurate match (important!), I could hear no real difference between source and replay both through my speakers and through headphones!
Then I recorded (at 0 level co-ax in) track 8 from the Naxos CD of the Adam Zero Ballet by Arthur Bliss. This disc is a demonstration quality recording and track 8 contains the full orchestra plus some high level percussion - a difficult piece to reproduce well. On replay, synched up with the original, I really couldn't pick any difference. Any perceived difference in sound quality was, I think, more in my imagination. LP4 is even better than I thought!
At 66 years of age my hearing is now well down at 16kHz and I can't hear higher frequencies any more, but my years of experience in broadcasting and hi fi has given me, I feel, a pretty good assessment of sound quality.