Preamp difference : if it's not the frequency, not the slew rate, and not the harmonics, what is it ?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
With Cubase the interface handling is automatic but what is sent to it is the selected output software bus which has the option of inserting a plug-in. If no dither plug is inserted the software does its own form of downscaling. Like ProTools, Cubase has a hardware bus matrix that allows the creation of named and numbered inputs annd outputs selectable from within the DAW.
For file export you can select any or all available tracks from the audio mixer whether a single track, multiple tracks, group bus or main outs.
There are external effects sends and returns with delay compensation that allow round robin ping to establish delay in send/return path and set the time advanced playback of the software bus to have zero insert delay - the delay compensation is sample accurate.
There is also a pre-record function you can set to record in advance a selectable number of seconds on any record enabled track - this ensures no late dropins or missed starts.
For using hardware inserts for exporting the render is done in real time, if no external fx are used the render can be accelerated by deselecting the real time export.
 
Cubase 6.5 is running internally at 32 bit float - if you import a 24 bit file into a project set to 24 it (the file) remains at 24 bit in the pool but if it is processed with a change and played back, as a 32 bit file it gets converted back to 24 - if you record a new recording sampling at 24 bit it gets internally converted to 32 bit file format and downconverts in playback to 24 bit fixed file format.
I think I may have figured out what was happening. When having the original 24bit file and recorded 24bit file on top of each other, with one or the other muted, the files would play back with 32bit processing. If I slide the tracks out of each other's way on the timeline, they both go back to playing at 24bit.
So the mute is a process , and it affects even the unmuted/soloed track if they share the same timeline I'm guessing? Even when muting or soloing when they are at different places on the timeline, they play back without any 32bit processing.......weird..... interesting....
 

Attachments

  • 2024-10-01 23-01-17.mp4
    1.5 MB
That is just the result of the multiply.... imagine if you multiply a tiny digital signal by 1/4 (-12dB) the LSB is now down 12 dB from where it was before. It takes more lower bits to display that new smaller result.

JR
This may be true and more to hitting the nail on the head? (maybe someone can clarify) Hypothetically; let's say we have a DAC/ADC set up whereas 0 is no sound and 10 is full sound. We record a 'simple' sound and it registers as a '3' on the scale. Then we increase the sound to twice relative loudness. It would register as a 6 (generally - and usually a +3db gain) of course ADC/DAC programs use a different inherent programs that include filtering-interpolation-shaping etc. So I would think
"gain is just a factor you multiply your finished filter design with" Of course sampling rates are much more complicated in the real world. So as long as there is no encroaching filters ~ gain would be just the increasing of numbers in a given 'bit world'. If anyone can interject that this is on the right track? And this would be done by a multiplication factor? hmmm...
 
Mathematically gain is a multiplication, yes. Is your question actually that simple?
So ~ in multiplying (binary numbers) is there any anomalies? (but how does reduction work by division?) In other words; would there be ANY differences (distortion from gain level to another?) I think this addresses the question upstream then. It seems in pure multiplication there would be NO differences BUT? if we reduce by, say 1/3 gain level... it seems the algorithm/filter would have to adjust. So the inherent algo/filter may have a very minor distortion. Cool stuff... what do you think?
 
So ~ in multiplying (binary numbers) is there any anomalies? (but how does reduction work by division?) In other words; would there be ANY differences (distortion from gain level to another?) I think this addresses the question upstream then. It seems in pure multiplication there would be NO differences BUT? if we reduce by, say 1/3 gain level... it seems the algorithm/filter would have to adjust. So the inherent algo/filter may have a very minor distortion. Cool stuff... what do you think?

Not understanding there tbh.
Re reduction - division is multiplication with a multiplier less than 1.
 
in multiplying (binary numbers) is there any anomalies?

Depending on the implementation there could be rounding considerations. Another thing to consider is word length of the multipliers. In a simple gain stage that just possibly changes gain slightly from the ideal calculated value, but in filter implementations it can result in frequency response variations.
You also have to be aware that if multiplying by a value greater than 1, the result is of course larger than the starting value, which may require more bits to represent accurately. That means if you have 24 bit audio data you need more than 24 bits to store the result (unless you are always using fractional multipliers, i.e. less than 1).

how does reduction work by division?

Well, if you divide something by 2 it is half the value of the starting data, i.e. the value is reduced. Dividing by 2 is also the same as multiplying by 0.5, so depending on whether your hardware can only do integer math, or fractional math using fixed or floating decimal point, it may be better to divide by 2, or multiply by 1/2.

would there be ANY differences (distortion from gain level to another?

Not in a competent implementation. You can always find someone to mess up the simplest thing if you look hard enough.

if we reduce by, say 1/3 gain level... it seems the algorithm/filter would have to adjust

I have trouble understanding what you mean by that. A linear operation by definition does not have changes in distortion with level, but of course an algorithm attempting to mimic a distortion producing device (e.g. guitar amplifier) would probably respond differently to a signal that is only 1/3 the amplitude of another. Filter is generally used to refer to a linear change in frequency response, which does not "have to adjust" when the signal is lower amplitude.
 
I have fairly accurate clones of the Telefunken V72 and EMI Redd47 preamps and they produce a noticeable bump in the upper mids. Was this by design?
 
Last edited:
I think I may have figured out what was happening. When having the original 24bit file and recorded 24bit file on top of each other, with one or the other muted, the files would play back with 32bit processing. If I slide the tracks out of each other's way on the timeline, they both go back to playing at 24bit.
So the mute is a process , and it affects even the unmuted/soloed track if they share the same timeline I'm guessing? Even when muting or soloing when they are at different places on the timeline, they play back without any 32bit processing.......weird..... interesting....
When two files are placed within the same time boundaries then they are both playing - whether or not one is soloed or muted - it’s the fact that two files are playing simultaneously with different sample start points requires the shift to 32 bit operation I guess.
 
This may be true and more to hitting the nail on the head? (maybe someone can clarify) Hypothetically; let's say we have a DAC/ADC set up whereas 0 is no sound and 10 is full sound. We record a 'simple' sound and it registers as a '3' on the scale. Then we increase the sound to twice relative loudness. It would register as a 6 (generally - and usually a +3db gain) of course ADC/DAC programs use a different inherent programs that include filtering-interpolation-shaping etc. So I would think
"gain is just a factor you multiply your finished filter design with" Of course sampling rates are much more complicated in the real world. So as long as there is no encroaching filters ~ gain would be just the increasing of numbers in a given 'bit world'. If anyone can interject that this is on the right track? And this would be done by a multiplication factor? hmmm...
not exactly how math works inside the digital domain.

If we multiply two 16 bit words together we get a 32 bit result. If that multiply term represents a -6dB fader move the the digital multiply would be times 10000000-00000000 the result would look like shifting the original digital word right one bit... the extra 16b remainder would be small value stuff.

Sorry I doubt this makes sense or helps much.

JR
 
not exactly how math works inside the digital domain.

If we multiply two 16 bit words together we get a 32 bit result. If that multiply term represents a -6dB fader move the the digital multiply would be times 10000000-00000000 the result would look like shifting the original digital word right one bit... the extra 16b remainder would be small value stuff.

Sorry I doubt this makes sense or helps much.

JR
It does make sense. Like when you multiply most single digits you (can) get 2 digits but - especially if that is the way it is intended - or set up for the result) So, also, I think - to lower the fader... is by multiplying by a factor of >1 you may get some minor anomalies?; or it can be not exactly a perfect reduction... (I think ) thx
 
Things are a bit more complex in the audio digital domain as there are binary numbers with ranges for both the positive and negative going voltage excursions at the input to the ADC. Chart of max excursion values and crossover point for 16 bit fixed point audio - this represents a 1V +/- swing at ADC input:
 

Attachments

  • IMG_0237.jpeg
    IMG_0237.jpeg
    95.3 KB
So it is 'in' the voltage excursions (+ or -) that actually tell the playback system what kind of power/gain to use... and this is where it gets complicated... (of course there will be theoretical and actual on top of that) ty
 
For volume rides both sets of numbers have to be adjusted and this is done with a floating calculator depending on the step position (in binary) of the virtual fader - for midi faders for either midi tracks or controllers there are only 128 steps but audio faders can be much finer - each sample needs to be modified by the same reduction or increase factor relative to where the fader sits to give a uniform scaled gain control no matter what level each individual sample is at - all of which is part of the DAW software, 32 bit as it does not work from 0dB downwards but allows levels above lets the mix bus go to very high levels without clipping , this being brought down to under 0dB for output to the audio interface by the mix faders and the master output fader. Binary calculators don’t work like decimal calcs - it’s worth having a look at one online and playing with some numbers - 4 bit or 8 bit as there are limited easy to see results are a good way to get a handle on how it all works. For example 1110 (14) / 0010 (2) = 0111 (7)
 

Latest posts

Back
Top