I think most of us here have read Rob Watts say that he believes his DACs are the most transparent because they have the most accurate reproduction of the original analog waveform. They are most accurate in terms of reproducing transients and they are most accurate because they have the least noise floor modulation. Because of the lack of taps, most DAC chips can't produce transients like Chord or R2R ladder DACs. But the problems with R2R ladder DACs are the noise floor modulation (which can be minimized slightly with closed form upsampling as in Schiit Yggradasil) and the low-level linearity (which can be somewhat mitigated by having multiple ladder DACs to average things out and by summing two R2R DACs for high and low signals as in Metrum Pavane). But interestingly, Rob Watts rarely comments on what the noise floor modulation sounds like (or for that matter low-level linearity problems). Romaz talks about TotalDAC sounding slightly soft. I have to admit, when I listened to MSB DAC IV, I don't find it soft. But there are two things that jump out at me. First is that the timbre of some drums sound different on the MSB than my Chord's. And even my old Benchmark DAC1 produces the drum timbre to be closer to the Chord than to the MSB. The timbre of the instruments vary from instrument to instrument. Some are very close to each other from DAC to DAC but others are quite different. This is not exclusive to drums but also to some midrange and treble instruments. Which one is more accurate? If I were to believe Rob Watts, I'd say it's Chord. The next thing I noticed with the MSB was that let's say there is an instrument that's somewhat soft and has a rhythmic sound throughout the track. When the instrument first comes in, it is very clear with the Chord & MSB, but as other instruments come in, particularly louder ones, the clarity of that softer instrument becomes somewhat lost or at least the timbre of that instrument change slightly with the MSB but not with Chord DACs. Now for most simple music, like vocals or chamber music, this is hardly noticeable. But for symphonic works, it becomes slightly more obvious. Is this what higher noise floor modulation sounds like? I don't know. Of course, I don't know if it's the chicken or the egg. I picked Chord QBD76HDSD because I like the sound and then I like Rob Watts's explanation of why I like the sound. Or is it that because Chord DACs are the closest to the original analog waveform of the music that I like Chord DACs?
Ultimately, the challenge in my mind is that we do have different preferences to sound. In theory, we should want the most accurate reproduction for music so we would want headphones (or speakers) with a flat frequency response and no dynamic compression. But we all sometimes prefer headphones/speakers with carefully designed distortions. I wonder if the same is true for DACs. Except ironically, the distortions in most DACs are not by design, it is merely the limitation of the DAC technology used to produce the analog sound.
Your comments are thoughtfully articulated and are interesting to ponder. We're all inclined to our own preferences as far as presentation but I think as audiophiles, we each seek the same musical truth that we call "realism." I'm not sure I fully understand what noise floor modulation sounds like either but I am in line with your thinking regarding tone and timbre and this is how I look at it.
As I look at what a DAC is supposed to do, simplistically, it is designed to convert a digital signal into an analog one but what has always caught my attention more than any other parameters are the bit depth and sampling rate that each digital file is tagged with. It's as if "16" and "44" are the two most important parameters a DAC needs to concern itself with when reproducing a Redbook file and as I try and make sense of what makes the DAVE special, I go back to these two fundamental parameters and what each represents in practical terms.
Bit depth is an expression of dynamic range where the formula is 1 bit = 6 dB of DR and so a 16-bit Redbook file, as an example, equates to 16 x 6 = 96 dB of DR. Similarly, a DAC that claims a 20 bit effective resolution should have a DR of 20x6 or 120 dB. Looking at it another way, in audio, a component's DR is its maximum output signal minus its noise floor and as Rob has shared with us, the noise floor of the DAVE is very low, in the order of -180dBFS. For a long time, I had assumed the first number, the maximum output signal was more important when in fact, it is the second number, the noise floor that is probably more difficult for a DAC to do a good job with. The higher the DR, the better the dynamic contrasts which is fine, but the lower the noise floor, the better the low level detail, which as many will agree, is one of the parameters that separates a great DAC from a good DAC. When the noise floor is constant, I can imagine the human mind has a way of compensating for it and even if the noise floor is fairly high, as long as it's constant, the human brain learns to adjust to it. When the noise floor is not constant, or is modulating, I can imagine that the human brain will have more difficulty blocking it out. A good analogy would be if we were flying in an airplane traveling at a constant velocity of 600 mph, we learn to adjust to it and eventually we don't realize we're even moving but if the plane is constantly speeding up and slowing down even slightly, we will probably notice it as turbulence. If my understanding of this is correct, than elimination of noise floor modulation, which Rob has claimed to have accomplished with the DAVE, is at least as important as a low noise floor. If you have the combination of a low and a non-modulating noise floor, I would expect that this would result in your best case scenario with regards to low level detail or "nuance." I would surmise that this would result in a finer ability to discern not just the timbre of an instrument such as a violin but also the timbres of two instruments that have very similar timbres, such as an upper standup bass and the lower registers of a cello or an electric bass and electric guitar. To a trained ear, you might be able to even discern between the tone of a Stradivarius and a Guarnerius.
With regards to the second number, the sampling rate (for a CD, this equals 44KHz or 44,000 samples per second), this equates to time resolution which allows us to discern where in space and time a sound is coming from. It's what gives us the dimensions of the soundstage, that the woodwinds are seated to the left of the brass section, that the percussion is behind the string section or that there are 4 rows within the string section. Because female voices generally are at a higher frequency than male voices, usually female voices will reach our ears faster and with better localization than male voices. Because certain sounds are faster than others, we get the sense not only of mass but force and impact. The higher the sampling rate, the better the time resolution and it has been said that at a sampling rate of 192 KHz, a digital file is finally capable of equalling the continuos data stream of analog playback to where a human ear cannot tell the difference. What impacts this, however, is jitter, and the more poorly timed digital playback is, the flatter and less dimensional it sounds and the less you can accurately discern where a sound is coming from and how fast it's coming. In a simplistic sense, jitter is what makes a digital file sound digital and as much as noise floor and noise floor modulation, it leads to fatigue.
Only Rob knows how he's addressed these parameters and while his claims were quite bold and even audacious when I first heard them, as I have listened to the DAVE and as I look at my understanding of the science of sound, it would appear to me that Rob is telling the truth.
