Not at all, and I said so several times. Dynamic range is not my concern. My concern is with the upper bound off the dynamic range (0dBFS), because thatās a hard limit which cannot be exceeded in integer space. This limit is the same for all bit depths.
1 Like
I wrote earlier:
No, the additional dynamic range doesnāt help, since the maximum bit values in 16bit, 32bit and 64bit integer formats are used to represent 0dBFS and map to each other.
Thinking about it a bit further, I believe the above isnāt quite correct. As far as I know, conversion to higher bit depths is done by left-shifting, so that the original bits are left as is and no rounding occurs. So, for example, the maximum positive value in a 16bit signal 0x7FFF will be left-shifted to 0x7FFF00 in 24bit. This is slightly less than 24bit max (0x7FFFFF). However, the gained headroom is minimal (0.000246dB) and therefore not useful in addressing inter-sample overs. The gain is similarly small for going to 32bit or even 64bit.
Hence, padding 16bit to 24/32/64bit gains resolution for processing, but no meaningful headroom. It still wonāt touch inter-sample overs (typically +1 to +3dB). For safety, ā1 to ā3dB pre-rate-conversion attenuation is required.
1 Like
Iām not sure the UI does accurately reflect the order of operations. (Someone from Audirvana would have to say definitively.) Meanwhile, I would suggest trying software volume reduction and seeing whether that works.
One other thought about your situation is that I donāt see how one gets intersample overs when the sample rate is reduced. Increasing the sample rate, sure; but if all 88.2k samples per second are below 0dB, then how would removing almost half of them reveal new samples that exceed 0dB?
Edit: Since new samples are being created by the conversion, I suppose intersample overs are theoretically possible. I am curious, though, if the chance of them is any less when doing decimation vs. interpolation. Anyway, as noted above, in the meantime before a definitive answer from Audirvana, Iād suggest trying software volume reduction and seeing what happens.
The screenshot isnāt representative or an example of inter-sample overs occurring. It was a random track playing ā I took the screenshot merely to show that rate conversion appears to occur before any chance of level reduction.
That said, unless resampling from 88.2 to 44.1, which would be simple decimation, going from 88.2 to 48 produces an entirely new sequence of samples, with the potential to hit inter-sample overs. Admittedly, and like I said above, this is far less likely than with upsampling.
1 Like
Sorry to keep quoting myself, rereading the above ā is there anything that would keep me from using the DSP (āprocessingā) stage for volume reduction? I could use any of the available EQ plugins to drop the level by 3dB or so, right? Are there any downsides to doing this?
1 Like
You have not yet answered the questions aboveā¦
Again for reference from the Wikipedia article linked above āAudio bit depthā:
Dynamic range and headroom
Dynamic range is the difference between the largest and smallest signal a system can record or reproduce. Without dither, the dynamic range correlates to the quantization noise floor. For example, 16-bit integer resolution allows for a dynamic range of about 96 dB. With the proper application of dither, digital systems can reproduce signals with levels lower than their resolution would normally allow, extending the effective dynamic range beyond the limit imposed by the resolution.[30] The use of techniques such as oversampling and noise shaping can further extend the dynamic range of sampled audio by moving quantization error out of the frequency band of interest.
Sure, there are many plug-ins to support that, however, you have made a fuss about the initial conversion from the 24/88.2kHz FLAC file to a 64bit/88,2kHz file down-converted to 24/48kHz for output after applying software volume control of -6dBā¦ What else do you need? Just try and identify ISPās in your playbackā¦ More-likely-than-not your system is producing more noise (jitter) that will influence the playback sound-quality with more consequence than worrying about the DSP processing producing or conveying ISP artifacts in the playback signal. 
Because they take us even further from my original question, which was about level reduction prior to resampling, to avoid inter-sample overs. Mind you, I was asking about avoiding them in general (i.e. in any situation), not specifically about the rates shown in the screenshot.
Some quick calculation (above) shows that increasing bit depth does nothing to address inter-sample overs, hence we donāt need to polish that particular tangent any further.
1 Like
They exist in the source fileā¦ which has a fixed bit-depth/dynamic-rangeā¦ of course they still existā¦ But there is little consequence unless your DAC cannot handle the dynamic range of the ISPs themselves.
Only potential limitation I can think of is the quality of the plugin. Audirvana Studio itself has EQ, which does make me think itās doing that (frequency-dependent amplitude adjustment) before conversion. So I donāt see a reason why it wouldnāt be doing a simple across the board volume reduction before sample rate conversion.
Also note that for folks like me who convert to DSD, thereās a volume reduction setting that takes effect before conversion, so it seems to me to make little sense to do this completely differently for PCM.
1 Like
No, the issue I pointed out was that the file was down-converted before applying software volume control. I was asking for a way of controlling the volume before conversion.
During playback I want to listen, not look for issues. Thatās why I want this problem solved for any situation and any combination of source and output formats.
Thatās quite unlikely. Jitter is well below audibly thresholds with even cheap modern DACs, whereas inter-sample overs produce actual distortion at audible levels. For some source material, anyway.
In any case, I think I found the solution that was staring me in the face all this time. I enabled and configured the new Audirvana EQ, with no filters and just a level drop of -3dB.
Thanks all for your time and replies, much appreciated!
2 Likes
The rationale for not providing a facility for gain reduction before PCM up-sampling boils down to the headroom of the SDM algorithm(s), that is transparent to all relevant PCM bit-depths.
Thanks, thatās exactly what Iāve done now.
2 Likes
Why do you think this is different than applying a Volume control gain reduction of -3dB or -6dB before output? 
My guess is the effect is identical to a -3dB setting of Audirvanaās volume control, though I canāt be certain of that. 
Of course until/unless Audirvana confirms that guess, the way youāve chosen should work just fine.
Because the sample-rate conversion on an unattenuated signal thatās mastered aggressively close to 0dBFS can lead to inter-sample overs and clipping.
1 Like
It is the effects of jitter (distortion) on the digital-audio signal in the computer and DAC topologies that are audible in the audition of any given recording/production.
Not in any measurements that Iāve seen.
1 Like
Only if the DAC does not handle the dynamics of the ISPsā¦ The ISPās are captured in the dynamic-range of the source encodingā¦ increasing the sample-rate does nothing to the encoded ISPās.
For reference (attached Texas Instruments document):
Skew definition and jitter analysis
By Steve Corrigan
System Specialist, Data Transmission
Another referenceā¦
Jitter Theory
by Julian Dunn, Audio Precision
http://www.audiophilleo.com/zh_hk/docs/Dunn-AP-tn23.pdf
Yes, I think thatās possible, because the 64bit format Audirvana converts everything to in the beginning might be a floating-point format and thereby sidestep the issue altogether. But I donāt know whether or not thatās actually the case.
2 Likes
In the up-sampling/downsampling processes, the FIR interpolation filter lowpass is the control mechanism.