D&R on Analogue vs Digital
- Thread starter ruffrecords
- Start date
Help Support GroupDIY Audio Forum:
living sounds
Well-known member
Can we actually hear a difference between a 12 khz sinewave vs. a squarewave? I doubt it. Isn't what we are hearing merely the signal chain reacting to the out-of-band signal?
I do think it has been established that we can differentiate timing differences (spatial information) much below the minimum lenght of the waves (samples) in the audible spectrum. But this information can be encoded in lower bandwith (say 44.1 khz) audio, the bottleneck here would be the resolution in bits.
I do think it has been established that we can differentiate timing differences (spatial information) much below the minimum lenght of the waves (samples) in the audible spectrum. But this information can be encoded in lower bandwith (say 44.1 khz) audio, the bottleneck here would be the resolution in bits.
Matador
Well-known member
So is the hypothesis that a square wave's first harmonic (3rd) should be well above hearing (36 kHz), but because we can hear a difference between a 12 kHz sine and 12 kHz square, then it means that steep LPF's in digital signal chains are removing audible content?
This seems to be conflating a lot of different issues together. First off, only an ideal square wave has purely odd harmonics: real (e.g. slew rate limited) squares can generate a lot of other frequencies, even at even intervals. Why not just play back a 36 kHz sine through the chain and see if it's audible?
And if you are that worried about it, why not just record at 96 kHz sample rate?
This seems to be conflating a lot of different issues together. First off, only an ideal square wave has purely odd harmonics: real (e.g. slew rate limited) squares can generate a lot of other frequencies, even at even intervals. Why not just play back a 36 kHz sine through the chain and see if it's audible?
And if you are that worried about it, why not just record at 96 kHz sample rate?
john12ax7
Well-known member
I have generally preferred 96k to 44.1k. It's also more accurate in the time domain. So just settled on it as a default and haven't looked back.
I could probably hear the IMD depending on how dirty the path is....
This is pretty well inspected.
JR
This is pretty well inspected.
JR
living sounds
Well-known member
Counterintuitively, time domain precison is linked to bit depth rather than sample rate.
https://science-of-sound.net/2016/02/time-resolution-in-digital-audio/
As for 44.1k vs 96k - it depends a lot on the converter and there are always trade offs. Lower sample rates often are better in terms of jitter. On the other hand, good filters are easier to construct for higher sample rates...
Last edited:
john12ax7
Well-known member
Accuracy and precision are not the same thing. Let's just call it signal integrity. Either way bit depth and SR both matter. In terms of signal integrity 2.2xBW is not a good choice. I know of no competent designer who would call it optimal, or choose to use it if not forced to. SR of 5-10xBW is a general guideline for time domain signal integrity of transient signals.
Sorry but reading that article it sounds like someone who read something about the topic, maybe does some programming, but has zero real world experience in hardware design. The dismissal of "sound events" is just silly. Music is much better modeled from a transient than steady state perspective. Take these blog articles with a grain of salt.
And yes filters matter a lot, both from a signal integrity perspective, and a sound perception perspective. Iirc you are partial to R2R type DACs with no filter? There is no right or wrong when it comes to preference.
living sounds
Well-known member
No filter would be a very bad idea IMO. My DACs use two cascaded digital antialiasing filters, one high in the ultrasonic region, and the other one below Nyquest. The ones loaded into my DAC make very sure no aliasing happens, these are linear phase filters and even filter out signal far below Nyquist, ranging a little into the audible band. So they are the opposite of the steep digital filters you find in most modern DACs. They are the only filters that lack the artifical mix of midrange aggressiveness and veiled dynamics that plague most digital playback today. There is no plausible correlation of these audible differences with actual measurements, since the measureable differences are almost completely out of the audio band. Signals coming from the DACs with these filters react better to processing and produce better end results in the mixing process.
I really wouldn't call it "preference" BTW, since the signal these filters produce is audibly the closest to the (analog) source. It would be hard to sell DACs with these filters via the internet though, since people would assume that cutting higs below 20k "damages" their signal more than the effects of the steep filters do...
john12ax7
Well-known member
I would agree with the premise that there is such a thing as audible accuracy which may or may not agree with measurement accuracy (though I'm guessing many here might not). I've certainly experienced this, where what sounded closest to the original, wasn't what measured closest to the original.
john12ax7
Well-known member
Also I didn't recall correctly your filter preference, thanks for clarifying.
I did read somewhere regarding the no filter thing. The premise was the electronics and speakers themselves are a LPF, if the image is high enough they would then do all the heavy lifting for you. Seemed interesting.
I did read somewhere regarding the no filter thing. The premise was the electronics and speakers themselves are a LPF, if the image is high enough they would then do all the heavy lifting for you. Seemed interesting.
living sounds
Well-known member
You can upload NOS ("non-oversampling-filters") into the DAC, which means no filters at all. As a result you get a high end boost and mirror images potentially causing problems with stages not designed for it. Percussive signals sound very alive, but it is a deviation from the source signal. It can work as a "sound" though (I seem to recall that the Fairlight samplers worked via this principle).
