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

[JohnRoberts]

My studio monitors, made by Wayne Jones Audio, now come with carbon fiber boxes and the new model does sound different than my model without the carbon fiber. Not a huge difference, but noticeable.

of course the sound not leaking out the back, comes out the front.

Cabinet rigidity was an issue mostly for bass response in molded plastic cabinets (back last century when I was over all product management for Peavey). Adding a wood internal strut to stiffen up the plastic cabinet helped, but the extra weight worked against the light weight selling feature.

JR

 

[kags]

of course the sound not leaking out the back, comes out the front.

Cabinet rigidity was an issue mostly for bass response in molded plastic cabinets (back last century when I was over all product management for Peavey). Adding a wood internal strut to stiffen up the plastic cabinet helped, but the extra weight worked against the light weight selling feature.

JR

These speakers, even without the carbon fiber, have heavy, rigid solid wood cabinets. The monitors weigh 45 pounds, so I assume any rearward sound is diffraction/baffle step below 250 Hz or so. The builder says the carbon fiber reduces cabinet resonances, which would be internal and would not affect leakage.

Plastic speaker cabinets sometimes benefit from a layer of dynamat or mass loaded vinyl, too.

 

[Cqwet Dbdfte]

Rauna speakers used concrete for enclosures.
Not really related to pre-amps though...

 

[RoadrunnerOZ]

Rauna speakers used concrete for enclosures.
Not really related to pre-amps though...

Maybe granite could be a contender? Little heavy though but would look good - a lot denser than concrete.

 

[Cqwet Dbdfte]

The speaker enclosure material should not be resonant, have great damping, have strength, easy to work with, looks not relevant to sound output. Added mass is a bonus. Concrete is probably overkill. Granite should work fine, not the easiest to cut. Fred Flintstone would approve. I'll stick to MDF.

 

[CLASS-A]

Hi,

I took the opportunity to measure different preamps (new SSL ones, Focusrite Scarlett, Warm Audio TB12), and investigate a bit on the topic of "audible differences between preamps".

If we run the preamps too hot, I came to the following conclusions :

• The preamps are usually rather flat in terms of frequency. At least nothing that couldn't be fixed with some basic digital EQing. Example : Measurement of an API312 here on SoundOnSound
  So it seems unlikely that frequency response is really what differs between preamps.
• The harmonics are of course different, in the case of transformer-balanced preamps. But they are usually at a rather subtle level (unless the preamp is pushed).
• The slew rate of the preamp usually seems to allow for a correct reproduction of all the audible frequencies.

So now I am wondering :

1. Why would a preamp like the API be called "punchy" ? What would "punch" be ?
2. If it's a variation of the transients (we hear about "slow" vs "fast" preamps), then where does it come from if the slew limit is > 20000 Hz ? The transformer ? If so, aren't the transformers supposed to be rather transparent, harmonics put aside for the lower frequencies, in the frequency range ?
3. Would the slew rate, if a bit too low, be able to influence the audible frequencies (even if the amplifier would, on paper, be able to reproduce those frequencies without any problem) ? It seems so : https://hifisonix.com/wp-content/uploads/2018/03/SID_and_TIM_W_Jung_77-79.pdf
4. Did I miss any phenomenon ?

I've searched all around the Internet but I can't find a clear answer.
Everyone seems to hear punch and smoothness differences, and I feel I hear them too, but where does it actually come from ? Has anyone been able to 'demonstrate' that ?

Thanks
Adrien

Hi Adrien
I made the same observation and during my various projects I tried to understand what could be happening. My (provisional) conclusion is that these differences are attributable to dynamic and not static aspects because measuring harmonic distortion in steady state on a sinusoidal signal is (I think) not very representative. When the components are well chosen and well used, the effects of DHT, slew rate and noise at these infinitesimal levels do not seem to me to be detectable by the human ear. The level differences and therefore the "punchy" aspect that you speak of are not integrated into this measurement protocol. The load impedance in relation to the current capacity must have a probable influence on this parameter. Finally, for line stages or microphone stages integrating an audio transformer, the characteristics of this component are essential for the sound signature. If we talk about legendary analog consoles like the Neve or SSL, they are all made from NE5532 or 5534 type circuits and even some with TL07n. However, these circuits are very inferior to what we can have today and yet the sound of these consoles is very appreciated and often considered "better". Is this a collective illusion linked to psychoacoustics. I don't have the definitive answer but I'm not entirely convinced that this is the case. What I noticed is that nowadays OPAMPS are often designed with consumption constraints which perhaps limit dynamic performance. In addition, I deliberately ignore the fact that the open loop bandwidth is very limited and the gain is very high which thanks to the feedback allows a high bandwidth and very low distortion but be careful, this is measured in steady state and not in impulse mode. However, under these conditions, the internal amplifier stages can go into saturation for a very short time and cause audible differences.Daniel

 

[scott2000]

Sorry for the veer, but which DAWs did you test? I've long heard differences in DAWs, which many claim don't exist, but never really quantified it. One thing I've noticed is many don't dither properly.

I haven't read through this entire thread yet. Was just searching "dither" for a question I have and came across this. Will read later today when I get back.

My question, and apologies if it's already been mentioned, is...

If playing a 24bit file from the DAW, no internal processing (volume changes,etc) through the DAC, through an external effects device, and back through an ADC recording at 24bits, there wouldn't be a need to apply dither?
Only if there were any processing done in the DAW would there need to have 24bit dither applied before it headed out to the DAC?
I hope this makes sense?

 

[RoadrunnerOZ]

I haven't read through this entire thread yet. Was just searching "dither" for a question I have and came across this. Will read later today when I get back.

My question, and apologies if it's already been mentioned, is...

If playing a 24bit file from the DAW, no internal processing (volume changes,etc) through the DAC, through an external effects device, and back through an ADC recording at 24bits, there wouldn't be a need to apply dither?
Only if there were any processing done in the DAW would there need to have 24bit dither applied before it headed out to the DAC?
I hope this makes sense?

No dither is required if not reducing the bit depth - it’s normally only used when going from 32 or 24 down to 16 bit to reduce quantisation errors by introducing random noise. If you’re not changing the bit depth, don’t use dither.

 

[scott2000]

No dither is required if not reducing the bit depth - it’s normally only used when going from 32 or 24

Yes I understand. Just wanted to make sure . To be clear, if I were to make a simple gain change or any other type of moves or processing in the DAW, I should have 24bit dither activated on the master out if wanting to record back in 24bit?

Is it common practice to have dither on at all times when working on a mix project where obvious higher than 24bit processing will be happening in a project? Basically monitoring with the dither on since interfaces and converters are usually 24bits?

 

[s wave]

I think the most important thing to find out first is to really know 'what internal processing' is being used in your DAW. DAWS can be different... I suggest to read up on quantization. Below is a GRAPHIC image (not sound) to give you a good idea. Bit depth kind of has to do with the clarity/resolution. When you downsample bit depth you probably want to smear the transitions (or dither). You get better smoothness. Dither masks the harsh Quantization Distortion. Now when you are just talking about increasing just the volume level of a track at a specific bit depth that you are using AND the DAW is using the exact bit depth - I see no need to dither. If it sounds good - there should be no need to 'smooth' it out. On the flip side - sometimes? quantization artifacts may be welcome in certain songs and genres. After all they are all actual artistic tools.

Quantization distortion becomes more audible; especially during quiet passages or fade outs. General rule of thumb (produce with a mindset of using the least amount of dithering - maybe just once when you - master for down sampling 32 bit float to 16bit for a specific format - such as mastering to just a CD or for an MP3) I also believe there is much confusion because there are different types/kinds of dithering algorithms. MOST are based on the 'POW-r algorithm' from Dolby this link has some good info: https://www.masteringthemix.com/blo...JuZZW3YuGadSNqNoQQYj4i_KCmU47bcotKvambqpBdX6w

As a tangent thought IMHO this can become a very important learning point in actual recording levels of different level sound sources. Overhead sound source capture can really add ambience or what have you vs a closer used mic that is less hot. Both have pros and cons. And reducing/increasing the DAW volume a close mic'ed snare drum vs a far away mic'ed snare drum sound very different. Also to remember ~ what is quantized and/or dithered on a track stays there if you just continue on processing that specific signal. Hope it helps a bit (no pun intended)

 

[john12ax7]

Yes I understand. Just wanted to make sure . To be clear, if I were to make a simple gain change or any other type of moves or processing in the DAW, I should have 24bit dither activated on the master out if wanting to record back in 24bit?

Yes you should. Any reduction in bit depth should be dithered. This includes 32 to 24 bit.

Is it common practice to have dither on at all times when working on a mix project where obvious higher than 24bit processing will be happening in a project? Basically monitoring with the dither on since interfaces and converters are usually 24bits?

Not sure if it's common practice, but it should be. With Pro Tools I use 24 bit dither plugin as the last in the chain before all the D/A outs, which includes monitoring and inserts. It does make an audible difference, at least with typical mix projects.

Some DAWs will give you this option "under the hood" without resorting to a plugin.

 

[kags]

I have heard arguments that dither is not necessary when going from 32-bit to 24-bit, with the reason being that quantization errors happen at the lowest bit(s), near -144 dBFS. Quant noise is cumulative, though, so if you are reducing more than once, the noise may increase more significantly. Dither noise, OTOH, is random, so will not increase in volume much if added more than once. My rules of thumb have always been 1. Do not add dither when exporting 32-bit float, 2. Dither is really unnecessary when going from 32-bit float to 24-bit (though some may prefer to add it anyway). 3. Definitely dither when you go down to 16-bit. 4. MP3/AAC conversions will provide their own dithering (or similar processing), so for best practices, use the highest bit-depth available to convert into mp3/AAC. 5. When in doubt, dither. You will likely never hear or be bothered by dither, but you may be bothered by quant noise.

I trust Pro Tools, Wavelab, and a very few others to dither when needed and not to dither when the DAW doesn't know if the user intends to or not (ie: export vs. bounce). I can't speak for many other DAWs, so I only use the DAWs I trust for critical work.

Scientific experiments and measurements aside, we are discussing music production, where noise (dither or quant) is rarely ever significant.

I'm sure some disagree—and I love to hear other POVs, but I've done plenty of critical listening and years of study to come to my real-world conclusions.

 

[RoadrunnerOZ]

Depending on the age of your DAW as earlier versions of Pro-Tools (before 10 IIRC) and other DAWs were not internally 32 bit float. Most people these days are using DAWs that have internal 32 bit float file format or higher as the base system, over which you have no control - this is not altered by your project settings.
If you record at 24 bit then bounce out at 24 bit you are reducing the bit depth from 32 float to 24 so technically you should dither.
Some say that as the Mantissa is less than 24 bit in 32 bit processes you won’t get truncated files on converting to 24 - this is not technically correct as the 32 bit file uses a part of the file to indicate the power of the 10 multiplier - this saves space - the 32 bits are not totally devoted to audio values but a polarity bit for + or -, an 8 bit exponent and 23 bit Mantissa
If you have a 24 bit audio interface then dither is obviously required. Dither must be the last plug-in of your DAW chain - post fader, post all other FX.
Can you hear the difference? Depends on the dynamic range of your audio - fades to silence or pauses, reverb tails into the aforementioned the distortion is marginally apparent but if you’re routing out and back in though FX or sending and returning at low levels you may well be advised to put a dither plug on the sends.
A lot of engineers don’t bother with this as the sum effect is marginal, at best inaudible. They only use dither for final mixdown bounces.
If you’re doing a project in 32 bit and have a 32 bit interface no need for dither until export at a lower bit depth like 24 bit or 16 bit.

 

[scott2000]

Depending on the age of your DAW as earlier versions of Pro-Tools (before 10 IIRC) and other DAWs were not internally 32 bit float.

Loading up the Bitter vst I can see that my 24bit project is showing 24 bits until I change fader gain .....then it goes to 32bit...actually a bit more .....so I'm assuming the internal processing is done in 32bits.

Thanks for all the comments. Feeling pretty good about my understanding of it now and see I have been doing some things technically wrong..lol

That Dan Worrall dither video was pretty good too and echos a lot that has been said....

 

[john12ax7]

I trust Pro Tools, Wavelab, and a very few others to dither when needed and not to dither when the DAW doesn't know if the user intends to or not (ie: export vs. bounce). I can't speak for many other DAWs, so I only use the DAWs I trust for critical work.

Historically Pro Tools has not handled this correctly (can't speak for the latest version). Samplitude/Sequoia has been one of the few that does.

One can debate audibility and sound preference , but from a DSP perspective dither is the correct way. Unfortunately there are coders, who know c/c++, but have a very poor grasp of DSP. Some plugins are particularly bad in this regard.

 

[JohnRoberts]

Loading up the Bitter vst I can see that my 24bit project is showing 24 bits until I change fader gain .....then it goes to 32bit...actually a bit more .....so I'm assuming the internal processing is done in 32bits.

A fader move is probably performed by a multiply, so if you multiply two digital words the result has more bits. Not necessarily better resolution but more bits... or something like that.

JR

Thanks for all the comments. Feeling pretty good about my understanding of it now and see I have been doing some things technically wrong..lol

That Dan Worrall dither video was pretty good too and echos a lot that has been said....

 

[scott2000]

A fader move is probably performed by a multiply, so if you multiply two digital words the result has more bits. Not necessarily better resolution but more bits... or something like that.

My Cubase is pretty old...6.5 ...kinda like my windows 7computer.
One thing I just found out is that, if I do just a straight playback, which shows 24bit, and record back in, it shows 24bit on the original and the recorded sample. But if I move the recording out of the way and try to run the original file again to record, same exact way, it shows it's playing back at 32bit ish...
There's obviously something buggy happening.
Hard to explain but the bit meter is definitely showing playback at a higher bit depth for no reason other than using the program. Like it activates some kind of ghost processing after using it for a task that shouldn't apply internal processing .....Very strange.

I can close and reopen the program and it doesn't do this. Shows 24bit. Even the file I recorded earlier shows 24bit. But if I save the project as it bugged out, it keeps the same behavior by showing there's 32bit processing happening....

dang... a bit frustrating..

 

[JohnRoberts]

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

 

[RoadrunnerOZ]

My Cubase is pretty old...6.5 ...kinda like my windows 7computer.
One thing I just found out is that, if I do just a straight playback, which shows 24bit, and record back in, it shows 24bit on the original and the recorded sample. But if I move the recording out of the way and try to run the original file again to record, same exact way, it shows it's playing back at 32bit ish...
There's obviously something buggy happening.
Hard to explain but the bit meter is definitely showing playback at a higher bit depth for no reason other than using the program. Like it activates some kind of ghost processing after using it for a task that shouldn't apply internal processing .....Very strange.

I can close and reopen the program and it doesn't do this. Shows 24bit. Even the file I recorded earlier shows 24bit. But if I save the project as it bugged out, it keeps the same behavior by showing there's 32bit processing happening....

dang... a bit frustrating..

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. Every file in the project window works at 32 bit - there are no 24 bit files running directly if being modified - the operation works on conversion of 24 bit files to 32 to make changes, a newly recorded file is 32 bit until saved. Operational changes are held in memory until you save or exit. When you apply any changes to a file - even a fader level change you have a new file at 32 bit float - this change is what shows in your edit history window - you may have noticed that your edit history disappears when you save - once you save, the temporary 32 bit file that is created when you modify is saved to a new 24 bit file.
There is a constant file output conversion running on changed files to provide intelligible data to run your audio interface and provide monitoring and audio outs for sends etc. When from a saved 24 bit file you do a check it will say the file is 24 bit, the moment you do anything to that file the modified and unsaved file is 32 bit until you do a save or exit the program and start it up again.
I use Cubase 6.5.5 (for a legacy suite of plugins) as well as Cubase 12 and Cubase 13 in different studios - running inserts as audio sends/returns to effects like compressors etc via an Apogee Dante interface system I have not noticed the need to dither on the way out.
This is pretty seamless and if you run your projects at 24 bit there is very little apparent conversion artifact distortion - it would only become apparent with very low level signal coming close to the noise floor. With a fade out I usually do a graduated fade with a drop fade at a musically timed point to end and as such the level is way above where artifacts would be noticeable.

 

[kags]

Pro Tools (at least for the last several years also processes at 32-bit float regardless of the session file bit depth. I have been told that Avid hardware handles bit reduction when analog inserts are used. For bounces, one must manually apply either, but for file exports and when re-saving files to a different format, PT automatically dithers as needed (though the algorithm is not user-selectable).

I imagine with the Carbon and other new interfaces the audio stream is maintained at 32-bit float until DA.