96Khz vs 192Khz

[par4ken]

I had a wonderful brilliant friend Russ Hamm who in the day wrote a seminal paper at the dawn of the new era, "Tubes Versus Transistors-Is There An Audible Difference" Russ always felt that 192k recording was the closest thing to tube sound that he had ever heard. Do you agree?...s

https://archive.org/details/TubesVersusTransistors-IsThereAnAudibleDifference/page/n1/mode/2up

I would have to agree with that. To me tubes have always sounded better than solid state. Todays solid state amplifiers however do seem to sound much better than that vintage seventies equipment though. I use tubes in two places the first is in my phono preamplifier. Tubes coupled with a moving coil pickup can provides amazing sound quality. I record my vinyl at 192Khz 32-bit-float. After declicking, tweeking level balance and normalising I save as 192/24 bit or sometimes 96/24 bit. I can't swear that 192K sounds any better than 96K but I do believe that the declicking process works better at the higher sample rate
.
My main system is bi-amped and I use tube amps for the mid-high frequencies. Tubes always seem to sound more natural in that function.

 

[murrysdad]

LOL my first hi-fi was a bi-amped home made Altec Voice of the theater with a 15" woofer and the big horn. I built the x-over with 6SN7's. I used two mono Dyna Kits to power it. Sweet but mono.

To me it has always boiled down to headroom and harmonics. You get a lot of headroom with 150 volts, plus.

Russ sums it up well in his paper second order harmonics versus thirds. One is musical, the other is irritating. Think fuzz tone.

A good transformer also gives 2nd order harmonics. Purests will claim that they color the sound, I say they make it richer, better.

On digital recording to me it's all about the nine-pole+ filter at the back end. The higher you can get the Nyquist away from 20K the better the sound and less ringing. ...steve

 

[jimfisheye]

I record my vinyl at 192Khz 32-bit-float.

What make and model DAC (audio interface) are you using with true 32 bit floating point output?

 

[murrysdad]

The best pro a/d, d/a's I know of are Crane Song. he builds them discrete.

To me it is all about clock and jitter. jitter_1

I did work for the military recording high power firearms up to 50cal at the shooters ears. I used RME converters using I believe using the National chip sets, at 192k and I have software where I can see and measure the impulse wave form. Pretty good recordings...s

 

[par4ken]

What make and model DAC (audio interface) are you using with true 32 bit floating point output?

I think that I already answered this, my sound card only provides 24-bit resolution but recording with floating point still allows recording above zero, without any clipping. The sound card is a Digigram VX822HR. 32-bit processing allows less degradation when using additional processing effects.

 

[par4ken]

The best pro a/d, d/a's I know of are Crane Song. he builds them discrete.

To me it is all about clock and jitter. jitter_1

I did work for the military recording high power firearms up to 50cal at the shooters ears. I used RME converters using I believe using the National chip sets, at 192k and I have software where I can see and measure the impulse wave form. Pretty good recordings...s

I remember seeing devices that went between your CD transport and DAC to reduce or eliminate jitter. I haven't read "Stereophile" or any other high end audio press in years, so haven't given jitter a second thought until this post. Digital audio seems very good (even with jitter). When I have a chance I'll download and listen to the files, I am rather interested if it will be audible or if it's just another high end fantasy.

 

[jimfisheye]

I think that I already answered this, my sound card only provides 24-bit resolution but recording with floating point still allows recording above zero, without any clipping. The sound card is a Digigram VX822HR. 32-bit processing allows less degradation when using additional processing effects.

You have been very misinformed with that.
Recording a 24 bit digital signal into a 32 bit container only pads it with zeros. It's literally the same data stream and it will null with the 24 bit copy exactly. Any clipping you do will be permanent.

Any modern DAW will have a 64 bit floating point mix engine to preserve the full resolution of source audio and after the fact processing. The original 24 bit recording (regardless of what bigger container you put it in) is a done deal and the analog signal as digitized cannot be gain staged after the fact. Any clipping or other distortion is permanent.

You might want to entertain the idea that you aren't being misled by all the official publications on this digital technology. This is all spelled out and you can experiment and test this for yourself. Line level analog audio is made to be easy on the preamp circuits to handle with at least semi pro gear. 24 bit digital has true 96db dynamic range with the lowest signal still having 8 bits of resolution. (You could call it 144db dynamic range if you think 1 or 2 bit signals are still meaningful. They aren't.) Some of the comments make it sound like someone is struggling with a dodgy old 8-track tape deck or something. You owe yourself a sanity check here!

 

[par4ken]

You have been very misinformed with that.
Recording a 24 bit digital signal into a 32 bit container only pads it with zeros. It's literally the same data stream and it will null with the 24 bit copy exactly. Any clipping you do will be permanent.

Any modern DAW will have a 64 bit floating point mix engine to preserve the full resolution of source audio and after the fact processing. The original 24 bit recording (regardless of what bigger container you put it in) is a done deal and the analog signal as digitized cannot be gain staged after the fact. Any clipping or other distortion is permanent.

You might want to entertain the idea that you aren't being misled by all the official publications on this digital technology. This is all spelled out and you can experiment and test this for yourself. Line level analog audio is made to be easy on the preamp circuits to handle with at least semi pro gear. 24 bit digital has true 96db dynamic range with the lowest signal still having 8 bits of resolution. (You could call it 144db dynamic range if you think 1 or 2 bit signals are still meaningful. They aren't.) Some of the comments make it sound like someone is struggling with a dodgy old 8-track tape deck or something. You owe yourself a sanity check here!

As I said before multiple times there is no clipping!!!!! The recording is done at 32-bit even if actually 24 bit padded with zeros! I try to stay below zero while recording but in cases where the level exceeded zero the sound was recorded without clipping! When normalised to zero the signal was reduced in level! Obviously the zero point (maximum output) of the actual sound card and the zero point of the 32 bit recording are not the same, or as you say the signal would be clipped!

 

[jimfisheye]

Sorry but something else is going on there! You can't undo a digitized clipped analog signal by putting that 24 bit data stream into a 32 bit container. This stuff isn't magical and the published spec of 24 bit PCM audio is not falsified. One of those would have to be the case for your claim to be true.

 

[par4ken]

Sorry but something else is going on there! You can't undo a digitized clipped analog signal by putting that 24 bit data stream into a 32 bit container. This stuff isn't magical and the published spec of 24 bit PCM audio is not falsified. One of those would have to be the case for your claim to be true.

JIm why do you keep talking about clipped audio. There is no clipping, that is the beauty of using 32-bit float. At no point does the audio clip, if I was to overdrive the input of the sound card I'm sure that I could force it to clip, but that is not what I'm talking about. When I record I try to keep levels below zero, as indicated by the audio programs meter/display. If I happen to go above zero (while using 32-bit float) the sound is still recorded without clipping. There is no magic there at all why can't you understand that? If I record in 16-bit or 24-bit fixed point then zero is zero anything above zero will be clipped.

 

[jimfisheye]

These digital formats and associated gear and analog devices have always worked exactly as the makers claim for me. I stand by my statements. You're the one presenting extraordinary claims! ie. That an analog signal can enter a preamp stage and into an AD converter with an above 0dbfs signal and that portion of the signal above zero can somehow be shuttled into a 32 bit file. Do you think this is happening wirelessly? Or do you think the 24 bit format as specified is falsified and actually digitizing at and passing 32 bits?

Post an example of one of your raw 32 bit recordings. Pull up the Bitter plugin (or your favorite bit depth meter if other) and show a screen shot of a 32 bit signal being displayed.

Mind you, I suspect there's miscommunication of some kind at the root of this! I'm not sure how else to jump in though. You can't get more than 24 bits out of a 24 bit AD converter. Something that clips a 24 bit converter... is permanently clipped. Maybe you're not clipping your analog input? And some other confusion is going on to corrupt the 24 bit capture (that in this example wasn't actually clipped)?

iZotopeRX has a pretty useful de-clipper tool. It's not "restoring" to an original unclipped state or anything magical. It's doing interpolation and guessing. Works really well for mild clipping. Saved my ass a couple times with live multitracks. Things happen with live sound sometimes. They're called screw-ups!

 

[par4ken]

That an analog signal can enter a preamp stage and into an AD converter with an above 0dbfs signal and that portion of the signal above zero can somehow be shuttled into a 32 bit file.

Jim, I've repeatedly said that is not at all what I have been saying! obviously if the AD converter reaches maximum 0dbfs it will clip. Zero as indicated by the audio program (while recording in 32-bit float) is not actually 0dbfs but would be something lower. I don't pretend to fully understand floating point but it can represent much larger and smaller numbers than fixed point can. While some say that it's unnecessary for recordings with limited dynamic range, I've come to swear by it. The only drawback is a larger file size but when completed you can save as a 24-bit file.
Furthermore if I was to save a 32-bit float file that went above zero as a 24 bit file without first reducing the level below zero the signal would in fact clip.

 

[par4ken]

Mind you, I suspect there's miscommunication of some kind at the root of this! I'm not sure how else to jump in though. You can't get more than 24 bits out of a 24 bit AD converter. Something that clips a 24 bit converter... is permanently clipped. Maybe you're not clipping your analog input? And some other confusion is going on to corrupt the 24 bit capture (that in this example wasn't actually clipped)?

Jim, I hate to elaborate too much on this as I am no expert. I don't want to communicate any misinformation. But this is what I understand. In 32-bit float the sound card drivers are controlled by the audio program. The extra 8 bits are used to scale the value of the other 24 bits from the souncards DAC. If a low level signal is present the 24 bits would be scaled higher to better record the low level signal. If the signal is higher the the scale would be lowered to prevent clipping. This type scheme would be very useful for live recording. That very wide dynamic range would have to be reduced to something more manageable while mastering.

And yes I'm not clipping the analog input.

 

[MidiMagic]

Yeah, looking again I see I'm completely wrong. The layout is EXTREMELY similar to what V-M did for years, though.

Apparently Collaro, like V-M, also made changers that others put their name on, so I had the entire thing backwards.

Collaro was based in Barking England.

Actually, Magnavox bought Collaro in 1960. Then all Collaro products had Magnavox on them until the mid 1970s, when Collaro sold a precision turntable/changer under its own name. And the last record changers made before record production ceased in 1993 were Collaro.

Collaro is actually known for several advances in audio:

- The first pickup cartridge sold for playing the Columbia microgroove LP.
- The first ceramic cartridge with the RIAA equalization built into it.
- The first drop record changer that could play odd record sizes automatically.
- Their changers from the 1950s work with magnetic pickups from the 1970s.

I think we are a wee bit off topic.

 

[MidiMagic]

When the input exceeds the A/D max, the signal will clip. This was the purpose of my experiment: find settings where clipping will not occur.

 

[ssully]

No. That's absurd. (196kHz most resembling 'tube sound')

 

[ssully]

JIm why do you keep talking about clipped audio. There is no clipping, that is the beauty of using 32-bit float. At no point does the audio clip, if I was to overdrive the input of the sound card I'm sure that I could force it to clip, but that is not what I'm talking about. When I record I try to keep levels below zero, as indicated by the audio programs meter/display. If I happen to go above zero (while using 32-bit float) the sound is still recorded without clipping. There is no magic there at all why can't you understand that? If I record in 16-bit or 24-bit fixed point then zero is zero anything above zero will be clipped.

What 'audio program' is showing you a meter going beyond 0dBFS when you are recording in 24 or 32 bit, but the recording isn't actually clipping? Are you sure that meter isn't just showing you what the level is with reference to 16bit? Because obviously , levels that don't clip in 24 can clip in 16. That's the main point of recording in >16bit (for live performance....for vinyl to digital, it's unnecessary).

 

[par4ken]

What 'audio program' is showing you a meter going beyond 0dBFS when you are recording in 24 or 32 bit, but the recording isn't actually clipping? Are you sure that meter isn't just showing you what the level is with reference to 16bit? Because obviously , levels that don't clip in 24 can clip in 16. That's the main point of recording in >16bit (for live performance....for vinyl to digital, it's unnecessary).

IZotope, obviously the zero level that is shown can't be 0dBFS at least not while recording or you couldn't go above zero. It would seem that floating bit work on sort of a sliding scale. When processing you can mix channels so that the signal exceeds 0 level as well but when you normalise the level returns to normal without any clipping. Likewise if you were to add equalisation (ie. boost up the bass) the signal will show above zero, but if you normalise the level is brought back down without clipping.

If I however after recording I boost the level up beyond zero the audio does become clipped and can't be restored.
Edit, that only happens if I increase the gain above zero using normalize, I can however increase the gain above zero using the gain function. I don't know how high you can go but in my recent test the audio was peaking 18 dB above zero. Normalising or lowering the gain the signal could be brought back down without clipping.

Again recording was done with 32 bit float.

 

[Sal1950]

I had a wonderful brilliant friend Russ Hamm who in the day wrote a seminal paper at the dawn of the new era, "Tubes Versus Transistors-Is There An Audible Difference" Russ always felt that 192k recording was the closest thing to tube sound that he had ever heard. Do you agree?...s

https://archive.org/details/TubesVersusTransistors-IsThereAnAudibleDifference/page/n1/mode/2up

Rediculos.
192k recordings sound like the mic feed.
Tubes can sound like just about anything the designer likes. Except for the mic feed, they have a very hard if not impossible time being 100% transparent to the source.

 

[murrysdad]

I agree that tubes and transformers color the sound with 2nd order harmonics. 2nd order harmonics are desirable in music recording. That is why professionals seek out new and used equipment with tubes and or transformers...