Your Vinyl Transfer Workflow (sharing best needledrop practices)*

You use dither when you go from high bits to a lower bit.

32 to 24 DITHER BABY YEA!
32 to 16 DITHER BABY YEA!
24 to 16 DITHER BABY YEA!
18 to 16 DITHER BABY YEA! (1998 DAT tape)
20 to 16 DITHER BABY YEA! (2002 DAT tape)

24 PCM to DSD2 NOPE.
32 bit PCM to DSD NOPE.
DSD IS A ONE BIT SYSTEM.

The noise shaping is done by the DSD system itself. One bit is equal to 6 db of signal to noise ratio. The DSD system pushes the noise beyond the range of human hearing. However, done improperly it can and will cause much trouble.

 

I think you are referring to the EIN figure, which is the electrical self-generated noise of a mic input, which is referenced to a particular voltage level. Given where 0.0 dBu = 0.775 volts, or 0.0 dBV = 1.00 volts. The lowest possible figure is -133dBV - just the thermal noise of connecting a 150 ohm resistor. Understanding Microphone Preamplifier Noise - Sound Devices

Take a new Rode NT1, plug it into a SoundDevices MixPre-3 II, and you get a noise floor amazingly better than the mics on a Tascam portable. Any actual mic (or actual room) still won't achieve that audio interface's 142dB dynamic range, and nor is that figure useful for digital transcription, with only a maximum -86dB THD+N.

Here's an older comparison of digital recorder EIN to mic dynamic range Recorder Tests - Avisoft Bioacoustics

Specs this good, and what dither to use? It doesn't matter when there's 40dB of dynamic range on an LP and the vinyl noise floor is its own dither. Listen to this tape noise from a 1967 LP: orff-jochum-sample.flac

 

Good point. Many people (and even pro engineers) don't realize the limitations of the Dolby A and DBX Type 1 back in the day either. People just read the specs and did some math. (Guilty as charged!) Most here would just take the signal to noise ratio spec of your typical 2 inch 24 track running at 15 ips and add 15 db for Dolby A.

64 db 'A' weighted for MCI JH24 @ 15 ips.
Plus 10 - 15 db for Dolby A = 79 db 'A' weighted.
Push the machine hard on those meters and 80 maybe 81 db 'A' signal to noise ratio......Or use DBX Type 1 and you could be seeing a nice
94 db 'A' weighted easy. Right? Not really.

One small problem. You see the tape machine actually produces it's own self noise. And I don't mean the sound of the tape machine in the studio either. I am talking about the noise coming out of the tape machines outputs in the stop mode. The input and output preamps were not as super quiet as they are today. This is one of the reasons mixers went transformerless.

Today we are used to - 90 dba on everthing. Even in a cheap $50 mixer. And - 110 to -120 dba almost becoming a standard. Even my 10 year old Tascam DP-02 (8 track portable DAW) provides -90 dba.

But back in 1974.....Things were different. For example, the hand made (no laughing!) Stephens 2 inch 40 track tape recorder:

@ 30 ips.....Signal to noise ratio 65 db unweighted

Tape machine stopped........ 70 db unweighted.

Neither Dolby A or DBX back then was going to get you any more than 5 db of signal to noise ratio out of that tape machine. Which is a lot by the way. Of course for the typical 2 inch 24 track of the period Dolby A would get you another 8 db. I should note that John Stephens didn't put balance jacks in his tape machines. He held the audiophile belief that balanced Jacks compromised sound quality. Any other tape machine would be swamped by buzzes without balanced connections but John John Stephens was a genius.

As the years went by mixers and tape machines got quieter. The best pro mixers back in 1975 might give you - 85 dba at best. But not - 110 dba.
By the late 1980's you had 2 inch 24 tracks like the Studer A827 that were super silent. The preamps were rated at - 100 dba or better. So the new Dolby SR could really give you 100 db on your recordings.

 

Yes, you are correct.
Using dither to cover up uneven vinyl noise is another use of dither. Good post. I agree.

 

My understanding of dither has always been quantitative noise (if that's the correct term) is created from lowering sampling rates. Dither and noise shaping lets you push that into the upper ranges. So dithering a 32/96 file to 24/96 no problem. I know when I dither 32/44.1 down to 16/44.1 I can clearly see the noise above 16-17K in RX after the final scan. Somebody with well trained good ears might hear this, I'm not one of them.

I have been using Noise Shaping "Ultra" Amount "Normal" and Auto-Blanking "When Quantized"

 

There is no 32-bit audio. It's just 24-bit with an 8-bit mantissa. In other words, it has a theoretically infinite dynamic range that's great for processing and avoiding clipping. It's also great marketing.

 

I've been using flat dither for a few years since I discovered that. The pros no longer really use noise-shaping either.

However, I decided to use Noise-shaping with low dither on a needledrop I just finished. I was concerned about the bass impact because the bass is so deep. Turned out nice. I won't make a habit out of using it.

The reason noise-shaping is so disliked these days is because it does color the sound. You sacrifice more sound by using noise-shaping than not.

 

There is debate on whether dithering should be used for 32-bit > 24-bit. The noise floor is so low it shouldn't matter either way. Some pros, like John, say yes, baby! Some others I have listened to say it doesn't matter. Ah, go ahead and dither anyway. It can't hurt. besides, you can't always be sure what the software programs are doing in the background.

My listening, critical and not, gets done at 16-bit level for the needledrops I do. That is why I am so critical about how transparent the dither I use is.

 

Does that go by the name Steinburg in other countries? My searches for this product with absolutely no results. All went to Steinburg.

 

What Is Dithering in Audio?

Dither - Wikipedia

As for using such extreme dither and noise shaping settings is concerned, compare with flat dither. Concentrate on how the bass, midrange, and high frequencies sound compared to the original. I'll bet you'll end up preferring the flat dither with no noise shaping.

 

Quantisation noise is an intrinsic property of any digital system even when not downconverting. It originates at the AD stage. The lower the bit depth, the higher the quantisation error, regardless if you originally sample at lower bit depth or create it computationally from a higher one.

Not so fast The problem with "raw" quantisation error (the red curve in the graph) its is irregular shape which sounds unpleasant. You can't avoid the quantisation error as such, but you want it unrelated to the music, i.e. white noise, which then becomes digital's equivalent to analogue tape hiss. This is what adding dither achieves in its original form. Again, this is not just a necessity when downconverting computationally, but even in the ADC.

Noise shaping is a "higher form" of dither where you don't achieve white noise (i.e. uniform across the spectrum) but one focussed on high frequencies. The two are not the same. White noise dither is also fine as such.

 

That's what I do. I record needledrop with MOTU M2 at 192/24-bit PCM. I convert it to DSD128 (5.6 MHz). This is equivalent to 352.8KHz/32 bit PCM. And then I convert back to 352.8KHz/32 bit PCM. This is 2x oversampling, yeah, but the noise is thrown out by the DSD format beyond 22 KHz and below 15 Hz. After that, with an analog minimally invasive filter, eg. Bessel bandpass 8-28000 Hz order 6 (-0.7 db cutoff at 20 Hz, -0.8 db cutoff at 20.000 Hz) I wipe out all that useless noise. Finally I dither with 20-bit TPDF. I think there is nothing useful beyond 20-bit because the ADC limitations.

 

For 32 bit to 24 bit, use a standard TPDF Dither.

 

Yes. Noise-shaping is simply moving that white (dither) noise to areas, usually to the spectrum extremes, where it will be less noticeable. Look at most popular noise-shape curves. They push down and away the dither noise in the parts of the musical spectrum that is most audible, such as the midrange/upper/midrange. But, that changes the tone and nuance of the music to a degree. If one uses flat dither, the tone isn't altered at the expense of more white noise.

One can increase or lessen how many bits of dither you apply to the file, but too little, and you risk quantization distortion. Too much, and you get the equivalent of tape hiss.

 

So I thought I sorta knew what I was going to do, and then tonight, a Schiit Jil popped up locally for cheap and I jumped.
Couldn't help myself.

So much for 'planning'...

 

In my set-up I recently switched out a Focusrite Scarlett - most widely used external soundcard for ‘pro amateurs’ for a Solid State Logic SSL2+ , which is about 230 EUR. The SSL2+ has a slightly (small but noticeable) better sound reproduction and controls on top instead of at te front. I would recommend the SSL2+ above the Focusrite.

 

It is $279 USD, almost twice the cost of the Focusrite. But, it's something to look into. Thanks for the tip.

 

It's called pre-echo, and I think I know what happened. While I was working on an album by The Reddings last night and I realized that I had the Liner-Phase filter checked from the last project, and I failed to use M/S processing for the de-hummer. I think that's what happened.

 

What's the use of needledropping at 24/192 if you know you're gonna downconvert it ? More precise declicking? Better transients? I can understand capturing professional tape into professional equipment at such high resolution for archiving purposes, or even higher as John said, but for records it sounds like a waste of hard drive space. I'm following Grant's example, I'm working at 24/44.1 and it's good. Check out Prince's latest HD remasters (1999 and SOTT): 24/44.1 and they sound great.
To each his own I guess.

 

Let's play "spot the click"



This is actually inaudible, being masked by the music, but the high-frequency impulse of a dust click sticks out like a sore thumb above the background. (these can be heard by scrubbing across the section at 1/4 speed though..)

Or in this case, spot the mistracking:


Or finally, the LP's high-resolution audio (a hard glockenspiel hit)


It's also revealing to find different frequencies of noise up there, noise and ultrasonic tones from electronics, your own equipment or the mastering engineer's or the studio's. Even aliasing from old digital.

You can throw away all you can't hear by using robust downsampling, but you're also throwing away information, for example a tone the plangent process could track to remove wow-and-flutter.

Playing back native 96k+ removes any argument about what exotic digital reconstruction filters to use.

There are also caveats to be mindful of when working with higher resolution, such as the reduced low-frequency resolution of FFT-based filtering.

 

It all depends on what you can HEAR. If you can hear a difference that capturing and processing at 192k or 176k can offer, go for it. If you don't hear a benefit, don't bother.

Some people capture at 192k or 176k, or even 96k or 82.1k, sample rate convert to 44.1 for processing, then take it to redbook for the final product.

There are no set rules for this game. Do whatever works for you and don't worry about what others do. But, all of us are offering our own workflow as guides or suggestions. Nothing we say is set in stone.

 

Some would agree with you. But mastering labs use dither when we go from one a higher bit rate to a lower bit. It has been that way for 40 years. I can't speak for other engineers but we don't dither over sample rates. Others might. But then some engineers think 44.1 khz is the best sample rate.

If it is DDD sure. But adding dither to a 1958 recording where the signal to noise ratio is 60 db is pointless. The noise floor of the recording does the job already.

 

Agreed.

 

I have made 16/44.1 recordings that sound amazing with depth beyond that of analog.
And the CD I made of these files (2× duplication) sounded just like the master file. I even called up the 8 track and ran it up through the digital mixer - same identical quality.

As for what mastering does is another matter. Does something happen in the CD process? Who knows? But with a few exceptions your CD DOES NOT sound anything like the original 16/44.1 file. Do not judge 16/44.1 on the compact disk. Bizzare I know. Of course their is a big difference between a 16/44.1 recording made in 1984 and a 16/44.1 recording made in 2012. Early converters sucked and blew.

You want the original file? Check out the Australian (region free) Blu-ray of Corbra Kai with it DTS-HD soundtrack. Even with cans you can hear that you are hearing master quality of these 1980's song. I own these on CD and they DO NOT sound anywhere near as this. Not on any CD version.

 

Then, there are people like Bob Katz who would probably say to dither that anyway.

You are a pro, but for we semi and amateurs, it's fun to experiment. I have spent days of precious time trying to work out the best-sounding sample rate conversions and dither schemes. One thing, though, I am so careful to not double dither. but, if I accidentally do, and delete my original file, it's not the end of the world.