Sure. Many dacs preamps and power amps have excessive distortion. Nonlinearities that show up as HD during single-tone measurements show up as IMD during music reproduction. IMD numbers usually tend to look worse than HD numbers.
IMD muddies up the midrange where human hearing is the most senstive and where the fundamental frequencies of vocals reside. Many listeners confuse sound of added odd order distortion components as increased clarity at HF. When even order distortion mostly muddies up the midrange, adding some odd order distortion can make it seem perceptually 'better' due the perception that clarity is improved. It's a common thing we see in listeners who do not have access to ultra-low distortion equipment and who therefore don't have a good way to know what it is they are hearing that they think they like is actually more distortion than they had before.
Here where Jam and I listen to dacs only ultra-low distortion amplification is used, SOA or near-SOA quality. Headphones currently in use include Audeze LCD-X, which are good enough to hear most of what is going on. Speakers include both professional mixing and mastering types. In addition, we have some very experienced listeners who know what distortion sounds like and when there is real clarity as opposed to artificially clear sounding distortion.
In that context, the very best DSD that HQ Player is capable of doing, say, ADSM-7 512+ DSD modulator, and the 40M tap closed form filter, or the newer ext2<sp?> filters in HD Player 4. They are computationally very challenging, but sound pretty decent when used for upsampling CD audio to DSD512.
They are smooth to an extent, but closer to the sound of real instruments. Still not perfect though. A few well-programmed FPGAs could probably do it better, given knowledge of the algorithms.
Anyway, once the distortion of dacs and other equipment is the reproduction chain is reduced to near-SOA levels the midrange is no longer muddy and congested, and any added odd order distortion at that point sounds gritty and fake rather than like added clarity. At that stage of development, smooth HF sounds much more natural. The music is still punchy and percussive if it was recorded that way. Cymbals sound like cymbals, and bass guitar sounds like bass guitar. Vocals sound full, round, and warm, while still clear and distinct.
Here, with 16/44 PCM the sound is more distorted and less detailed coming out of the dac than DSD512 is, but not bad. Main problem with PCM is that HF is still slightly grainy and gritty. That's with AK4499. With Sabre dacs its the same, but in addition one of the big problems with Sabre PCM are those awful PCM interpolation filters. In comparison, Sabre DSD filters sound much better and if one looks at the filter response curves in the data sheet, there is clearly a big difference. The DSD filters are much simpler. If using Sabre with PCM then it makes good sense to use an external interpolation filter like Benchmark (and Marantz, IIUC).
Might you be referring to the following:
Q&A with Thorsten Loesch of AMR/iFi | AudioStream
Q&A with Thorsten Loesch of AMR/iFi Addendum: PCM vs DSD | AudioStream
Last edited:
Its a little like ASR around here, but not quite so bad. Over there people are threatened with being banned if they talk about how something sounds without providing 'proof.' Here they just tell you that you are crazy and or don't know what you are talking about.
It’s important to note that some of the AKM DACs (others also) have a direct DSD mode that bypasses the SDM stages of the DAC. As Thorsten says, you lose volume control and most commercial DACs don’t employ the DACs in that way. However, you still get filters on the new AK DACs, though, and boy do these chips sound good when driven at DSD 512 in direct mode. There are plenty of DIY solutions out there allowing you to sample this for yourself.
You have no way of ever knowing how it sounded while being recorded. Not to mention the usual steps taken between recording and release.
And keep tweaking the DAC until the replay of the five recordings of various school bands you managed to capture on your Zoom recorder match how you remember them?
I used to record the amateur chorus my wife sang in using a Fostex FR2-LE field memory recorder, two AKG C900 microphones in an ORTF set-up (acoustic presence filters removed) and a home-made microphone preamplifier. It always sounded far more realistic than almost any commercial recording, although the reverberation of the room always seems somewhat stronger than it was in reality.
The single biggest improvement in my DIY DAC is the use of transformers at the output.
I've probably worn the story out, but, with my amp designed with no attenuation on the input, every imperfection is exposed. Single ended outputs produce various sounds, as minute as they may be, that kept my system from being considered professional. She is completely dark in contrast now. Got some drum corps international rocking the trix ATM
Don't forget to enjoy the music
I've probably worn the story out, but, with my amp designed with no attenuation on the input, every imperfection is exposed. Single ended outputs produce various sounds, as minute as they may be, that kept my system from being considered professional. She is completely dark in contrast now. Got some drum corps international rocking the trix ATM
Don't forget to enjoy the music
The math, as explained by Lipshitz and Vanderkooy is clear - the 1 bit truncation to DSD cannot be adequately dithered. Inadequate dither leads to noise modulation, subjectively perceived as 'softness'.
http://www.audiodesignguide.com/DSC1/SACD.pdf
Or a Schitt MultiBit. The new Bifrost has their new generation USB hardware/software and now the same ladder-dac chip as the next model up (th eone i have). Made in the USA.
dave
On top of that, it leads to distortion and to idle tones that are frequency-modulated by the signal.
At high bit rates, the distortion and tones can be solved at the expense of oversampling rate by embedding a pulse width modulator with randomly rotated pattern in a sigma-delta loop and using a multibit quantizer with proper dithering. At low bit rates, you can use experimentally-determined inadequate dithering or chaos, which helps a lot but not perfectly. (Another trick is to apply a small DC offset to shift the idle tones out of the audible range and hope that no-one will play a 1 Hz sine wave through the converter, although that trick is more applicable to ADCs, as DAC idle tones are usually at 0 Hz anyway when playing silence.)
Attachments
Last edited:
This sounds quite similar to what Mola Mola are doing in their Tambaqui DAC, minus the transversal filter at the output.
Mola Mola
Indeed. In any case, post #34 only relates to the conversion from PCM to a single-bit signal, whether you use a single DAC or a FIRDAC for the subsequent conversion to analogue is a whole different matter.
Combining PWM and sigma-delta modulation is a very old trick. Philips already used it in their first bitstream converters, although their pulse width modulator was outside the sigma-delta loop. In an old Peter Craven patent, the variant with a pulse width modulator inside the loop is already described as prior art.
Combining PWM and sigma-delta modulation is a very old trick. Philips already used it in their first bitstream converters, although their pulse width modulator was outside the sigma-delta loop. In an old Peter Craven patent, the variant with a pulse width modulator inside the loop is already described as prior art.
Would that be the one where he coined the phrase 'procrastinated noise shaping' (if my memory serves correctly) ?
Ah perhaps he used that phrase when talking about the same work in the AES paper. It could have been this one : AES E-Library >> Toward the 24-bit DAC: Novel Noise-Shaping Topologies Incorporating Correction for the Nonlinearity in a PWM Output Stage
- Status
- Not open for further replies.