sr1200 said:
I watched the video and no where did I get that he implied that a PT converter would or should sound the same as a soundblaster card from the 90's. He even admits at the end his "sins of omission" were great on this video.
Which doesn't necessarily mean any facts stated were incorrect. Only that he glossed over areas he inspected and it's a little more complicated than presented.
Adding bits adds headroom (reducing the noise floor).
In theory for a perfect simple conversion perhaps. Modern conversions are far from simple with noise floors higher than LSB so analog dynamic range as conventionally defined and measured is less that digital theoretical dynamic range. More bits does not automatically translate to lower noise floor while there is an expectation and some association to satisfy that expectation.
Adding samples allows for frequencies above the range of hearing (not necessarily perception... another fight for another day) to be recreated by the converter but it does something else he didn't mention.
Sorry if wasn't clear.. I meant perception of value not sonic perception.
Higher sample rate also minimizes those quantization "errors" in the higher frequencies which DOES change the sonic characteristic on the DtoA conversion. If you don't believe that, just record a crash cymbal at 44.1 then record it at 192 or even 48. Tell me your perceived difference.
Umm time samples affect the capability to capture and reproduce high frequency components. Quantization is generally an amplitude phenomenon.
If the cymbal crash is effectively LPF before the conversion, and you can not hear above 20kHz or whatever the limit is, there should be no audible difference. A close mic'd cymbal crash will have a lot of content tens of kHz above audible range. If you are hearing artifacts that may be a failure of pre-filtering for oversample converter or some other error. In analog paths, cymbals are notorious for causing audible LF IMD.
Its one thing to test a perfect sine wave, it's another thing to test with a complex waveform where there maybe a transient spike somewhere that doesn't get picked up by the conversion at lower sample rates. ie: the attached graphic.
If the transient is too narrow to be sampled, it is too HF to be heard.
The pic on the left would be at 44.1 for a complex wave form. The one on the right is the SAME wave form sampled at 88.2. The first 2 lines on the left would come out about the same as the first 4 on the right. But look what happens after that. That would be the detail difference. Is is audible.... depends on how critically you're listening. Is it going to be a huge drastic game stopping difference, no. But again, in the higher end of the spectrum, it IS noticeable if you're going for a higher fidelity reproduction of the sound. If you're just recording sine waves, yeah, stick with 44.1. If you're recording an orchestra or a jazz quartet... probably wanna bump up that sample rate.
Sampling theory is mature. Enough is enough, Enough said.
The biggest issue with D/A A/D reproduction was something he didn't even mention which is the clock that is running the whole thing.
I used to slave a couple of hardware recorders using MTC (the only thing i had at the time). The 2 hardware units worked well, but the PC didn't like it too much. And i wish I saved a copy of the track that I worked on when the clock really started struggling. You CAN hear jitter if its bad enough and will make a big difference in the overall sound of the file. I should note that I've only had issue with it when trying to sync multiple machines without a good word clock. But it's real and its the reason I use my Apogee as master clock to sync all the digital equipment i use, there IS a sonic difference. (which is why people get their 002/003's modded by black lion... they just put a *****in clock in there (and upgrade the crap PS) but the clock makes the difference which is why a BL modded 002 can sound as good as a blue 192) If ya ever wanna hear the real world difference between the two you're more than welcome, and bring your apogee along too. we can send the same signal into 2 or 3 converters and see what happens. (can do different sample rates too... then hear whats going on when you try to phase cancel them)
Bad clocking can cause issues, but this too is often overstated as a problem in practice, probably to sell some new hardware. There's a good white paper floating around about clocking that debunks the hysteria.
But what would I know I am not a digital expert, just been paying attention for several decades. If you go back far enough early digital really did suck. These days not so much, but still lots of magical thinking around. Digital conversion is an ideal topic to flout pseudo-science about.
JR