Can you hear bit rate?

I do a fair amount of listening to channels which provide a combination of both familiar and new tracks. Whenever a track sounds particularly smooth or spacious it is almost always 96X24 or better. I attribute that mostly to bit depth and the greater resolution low level signals enjoy with greater bit depth. I'm not sure greater sample rates matter as much.

One thing I do notice is people confusing bit depth and bit rate. Bit rate is the X axis and bit depth is the Y axis. I also tend to agree with Anthony Cordesman's assessment, and disagree that the early Philips '14-bit' CD players sounded better. I always found them lacking in low level detail and 'spaace'. But theirs were sins of omission, not sins of commission.

Sins of commission include things like harsh edges on acoustical instruments, brass, and female vocals. Might just be my ears, but that failures are immediately audible and annoying to me. 

One thing I do notice is people confusing bit depth and bit rate. Bit rate is the X axis and bit depth is the Y axis.

@panzrwagn  - Mostly right.  Sample rate, or frequency is the X axis. 

Sample rate may be 44.1 kHz, but the bit rate for that, assuming stereo is 44.1 x 2 x 16 = 1,411,200 bits per second.  Technically, bit rate is expressed in bits per second, which is mostly useful to compare to network bandwidth. 

So, for Redbook, the required audio bit rate is about 1.4 megabits per second, and far far less than a 100 mbit network cable can carry. 

"Can you hear bit rate?

24 bit stream decoding sounds better for original 24 (32) bit tracks! on opposite, if original tracks are 16 bit AD originated, not much. 

@panzrwagn 

disagree that the early Philips '14-bit' CD players sounded better

I did not say they sounded good, just better than their contemporaneous competition!

Philips realised that they simply could not trim a resistor array precisely enough for those two least significant bits to be monotonic.

On the positive side, they used quadruple oversampling from the get-go, eliminating the so-called brick-wall filter with its nasty phase artifacts.

Anthony Cordesman wrote in Audio in 1986 "(French magazine) Son tested nine different Compact Disc players and found significantly different performance with regard to separation, distortion, and phase, pulse and square-wave response. The machines with the worst measurements, all of which use the 16-bit process developed by Sony, had well over 80° of phase shift, poor 1-kHz square wave performance and a considerable amount of ringing"

So, apologies, I meant to ask: Can you hear differences as the sample rate or bit depth changes?  

I was just about to throw darts at it, but now it makes sense. I think it’s indeed possible to hear the difference in bit depth—especially with more complex passages like symphonic orchestras—because of the much higher dynamic range (around 146 dB for 24-bit compared to 98 dB for 16-bit).

As for sampling rate, yes, we can often perceive improved sound quality—typically a smoother top end or more natural transients—when we upsample or oversample the signal. This is mainly because upsampling shifts artifacts such as phase distortion and pre/post-ringing to higher frequencies, beyond the audible range.  My Harmony Micro DAC offers both OS and NOS modes, and I did notice a slightly smoother, airier top end in OS mode, though with a bit of loss in detail. Personally, I prefer NOS mode.

As is known, upsampling can cause clipping, so I do reduce the overall signal gain a little.  Roon converts everything to 64 bits before this so I can do this without further loss. 

What you did (reducing the gain) is perfectly fine, but I just want to clarify that upsampling itself does not cause clipping. The 64-bit floating-point processing within Roon’s DSP engine, where upsampling occurs, provides virtually unlimited headroom, making clipping literally impossible at that stage. Clipping can only occur when Roon converts the signal back to fixed-point formats such as 16-bit or 24-bit PCM, as supported by your DAC. You can either apply a limiter to prevent this from happening or, as you did, reduce the gain to ensure that the converted signal never exceeds 0 dBFS.