HydrogenAudio

CD-R and Audio Hardware => Audio Hardware => Topic started by: SoNic67 on 2012-05-26 00:54:02

Title: 32-bit capable DACs
Post by: SoNic67 on 2012-05-26 00:54:02
I don't know why people obsess about 32 bit "support" when their device manufacturer clearly states that that format is not supported. The DAC chips included on Xonar soundcards are regular 24 bit chips...
IMO, 32 bit is a worthless waste of space, all the '32 bit capable' DAC chips existent today have actual resolution less than 24 bit. Their catalog data is equal or even less than the top-of-the-line 24 bit DAC.
It's just shamless marketing and amazingly, enough people bite the bait.
The new beta plug-in (ver 2) works fine on E-MU 1820 w/ Win7 x64 and foobar ver 1.1.12 beta 6.
Title: 32-bit capable DACs
Post by: Mr.Duck on 2012-05-26 13:05:44
IMO, 32 bit is a worthless waste of space, all the '32 bit capable' DAC chips existent today have actual resolution less than 24 bit.

Twistedpear's Buffalo III has a 32bit DAC.
Title: 32-bit capable DACs
Post by: lvqcl on 2012-05-26 13:24:24
Twistedpear's Buffalo III has a 32bit DAC.

And...? As SoNic67 said, it has actual resolution less than 24 bit.
From leonvb-Buffalo_III_DAC_Integration_Guide_V1.2.5.pdf (http://www.twistedpearaudio.com/docs/community_docs/leonvb-Buffalo_III_DAC_Integration_Guide_V1.2.5.pdf) :
THD+N (with IVY III, mono mode):  -120dB
(in mono mode "all 8 balanced DACs of the ES9018 are used for each output channel achieving the best possible dynamic range")
Title: 32-bit capable DACs
Post by: Mr.Duck on 2012-05-26 13:51:57
Maybe someone wants 32bit output to their DAC. It's hardly just "shamless marketing" and a "worthless waste of space". Maybe I misunderstood what SoNic67's post was trying to convey.
Title: 32-bit capable DACs
Post by: phofman on 2012-05-26 14:58:04
They are obsessed because they want to be obsessed. Obsession and proper understanding are often in the opposite corner.

32bit DACs make sense for designers since they do not have to care about bit decimation from 32 to 24 (i.e. most PCI sound interfaces support 32bit DMA samples only, yet they have to output 24bits on their I2S lines to codecs.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-05-26 19:23:38
32bit DACs make sense for designers since they do not have to care about bit decimation from 32 to 24 (i.e. most PCI sound interfaces support 32bit DMA samples only, yet they have to output 24bits on their I2S lines to codecs.

Padding zeroes is the default DMA mechanism. Why store those zeroes on HDD? You don't have to worry about 32 bit decimation - unless you are a recording studio with 32 bit ADC. Let me know what mixing console has that ADC's...
Maybe someone wants 32bit output to their DAC. It's hardly just "shamless marketing" and a "worthless waste of space". Maybe I misunderstood what SoNic67's post was trying to convey.

Let me try it one more time
For real audio resolution, the number to look for is the THD+N. That includes the noise and distortions on output (both are not part of original music). Most of the top DAC's are in the -105...-110dB range here. Some manufacturers use % to hide it, for example a THD+N of 0.0003% means -110dB.
The best performing 32 bit DAC on market is ESS Sabre 9012/9018 (http://www.esstech.com/index.php?p=products_DAC). Listed by manufacturer with a THD+N of -120dB.

Now... a 22 bit audio signal represents a range of 132dB. Or in (%) will be 0.000024%. If your DAC doesn't have at least those numbers for THD+N, then it cannot reproduce the 22 bit correctly. Even the above Sabre32 doesn't "touch" those 22 bit...
Everything more than that is just noise and distortion added to the musical signal and sold as "32 bit magik". Wasted storage space and bandwidth.
Most of the PC soundcards don't even rise to 18 bit level of true analog quality.
24 bit was adopted because is a round number of 8 bit packets, not because it adds something to the final quality above 22 bit mark...
Title: 32-bit capable DACs
Post by: Jackal29a on 2012-05-26 19:31:17
I'm quite surprised by some people's reactions, it has been stated here that both ASIO and WASAPI provide no advantage whatsoever over DS therefore being as pointless as 32bit output so, why not ditch these plugins altogether? the same could possibly said about FLAC, APE or any other lossless codec that provides no audible advantage over well encoded 192Kbps, and so on and so forth.

I've never, not once, claimed 32bit sounded better, had any other advantage or that it was accepted by any other PCM179x DAC - other that the PCM1795 and, as Musiland discovered as an undocumented capability, the PCM1798. In fact I did write that I didn't give a bl**dy rat's a*se about it. I was interested in it  because other player I mentioned can do it thus it can indeed be done. I just wondered why Foobar couldn't. That same player can also send WASAPI @ 88.2 or 176.4 to the Xonar whereas Fooobar (and most others for that matter)  cannot reporting "not supported", just in case someone want to split this thread and create a new "88.2 and 176.4KHz capable DACs" thread, I AM NOT saying or implying in any way there is any advantage in playing such material or that it sounds any better. OK?


Title: 32-bit capable DACs
Post by: SoNic67 on 2012-05-26 19:36:27
Foobar in my opinion does the right thing telling you that that format is not supported. The Musiland driver/software lies and serves you a 24 bit claiming that is 32 bit. PCM1798 (http://www.ti.com/product/pcm1798) is a 24 bit DAC per manufacturer datasheet.
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-26 19:48:28
Foobar in my opinion does the right thing telling you that that format is not supported.


That has nothing to do with the hardware.  Its just dependent on the driver you're using.  My 16 bit onboard will happily accept 32 bit input from foobar.  You're probably using some weird driver for ASIO or whatever that doesn't support 32 bit.

Maybe someone wants 32bit output to their DAC.


Then they should  use AC97 onboard sound and foobar2000.  Works great.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-05-26 19:55:56
What is really representing the 32 bit support implemented in those WASAPI components?
My soundcard E-MU 1820m seems to "work" with WASAPI 32 bit selected (like with 24 bit). But I know that the DAC inside is a Cirrus CS4398, 24 bit one. Creative doesn't advertise 32 bit capability in their Win7 drivers.
So... I am confused. Where is this "32 bit" coming from?
I know that DirectSound will accept any bitrate/bitdepth and convert those on the fly (badly) to match the capabilities reported by the sound card driver. Is this similar?
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-05-26 20:02:54
I said that the driver is the one that "lies"... Probably uses DS to change the bitrate/bitdepth to match the real capabilities? In this case using WASAPI with those drivers would be useless...
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-26 20:03:46
For real audio resolution, the number to look for is the THD+N.

Can You defend this statement with some additional information?

Then why include Dynamic Range into specification? Just go for THD+N, right?
Title: 32-bit capable DACs
Post by: Jackal29a on 2012-05-26 20:34:39
The Musiland driver doesn't lie, the USB receiver IC accepts that bit depth just fine and then truncates it to 24bit which is what is sent to the DAC. In the 03US series the signal is not truncated because contrary to what specs say the PCM1798 does accept 32bit though, obviously, its real performance is roughly 20bit. The PCM1795 is listed at 3.30USD while the 1798 is a 2.95 part, it makes sense for TI not to admit they are that similar.
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-26 20:48:52
I said that the driver is the one that "lies"... Probably uses DS to change the bitrate/bitdepth to match the real capabilities? In this case using WASAPI with those drivers would be useless...


My point is that foobar is not doing anything special.  You're just using a driver that doesn't support 32 bit, so if you try to use 32 bit, it doesn't work.
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-26 20:55:12
For real audio resolution, the number to look for is the THD+N.

Can You defend this statement with some additional information?

Then why include Dynamic Range into specification? Just go for THD+N, right?


THD is used because in the old days it was easier to compute then dynamic range.  But now that we have computers, in some sense its unnecessary, since its usually dynamic range you actually want.  However, THD does have the advantage that its somewhat standardized in meaning, so you at least have a good idea whats actually being measured.  Dynamic range less so, since different people pick different ways to measure it, and often the one that makes them sound the best.  THD+N at 1khz for instance is harder to fudge.
Title: 32-bit capable DACs
Post by: Mr.Duck on 2012-05-26 21:06:26
For real audio resolution, the number to look for is the THD+N.

THD+N includes noise. There is only so far you can improve this number for practical reasons such as thermal noise stopping it going any better. That's not the same thing as dynamic range which is a different spec.

Besides, specs aren't everything. Different DACs might measure the almost the same but that doesn't mean they would sound almost the same.
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-26 21:52:11
THD is used because in the old days it was easier to compute then dynamic range.  But now that we have computers, in some sense its unnecessary, since its usually dynamic range you actually want.  However, THD does have the advantage that its somewhat standardized in meaning, so you at least have a good idea whats actually being measured.  Dynamic range less so, since different people pick different ways to measure it, and often the one that makes them sound the best.  THD+N at 1khz for instance is harder to fudge.


Thank You.


Then it makes me think that  THD+N better than -95 ...-100 dB has little sense. 

Untill the moment best headphones's THD I known it's HD800's one. It's <0.02% (-74 dB). Electrostatic phones don't  count as they require a special dedicated type of amplification.
If soundcard has THD+N -20 dB below of  HD800's (-74 dB), let's say  -94 dB then final THD of the soundcard + headphones will be (-94 dB) + (-74dB) = -73.96 dB. Further improvement of soundcard's THD+N is pretty pointless.
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-26 21:54:28
For real audio resolution, the number to look for is the THD+N.

THD+N includes noise. There is only so far you can improve this number for practical reasons such as thermal noise stopping it going any better. That's not the same thing as dynamic range which is a different spec.


Generally speaking, you can usually calculate dynamic range THD with reasonable accuracy, so they're not entirely different. 

Besides, specs aren't everything.


No, but measurements are.

Different DACs might measure the almost the same but that doesn't mean they would sound almost the same.


Unless you've screwed up somehow, it does mean so actually.  If you want to talk about audibility, pretty much anything with a flat frequency response and > 70-80dB will sound the same in any real world test.  Of course, you have to measure properly, which means with the actual headphones hooked up and at the level you're planning to listen to.

Specs better then that are mostly about bragging rights and about headroom if you eventually want to use the output to drive a more difficult pair of headphones.
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-26 21:58:15
Untill the moment best headphones's THD I known it's HD800's one. It's <0.02% (-74 dB). Elestrostatic phones don't  count as they require a special dedicated type of amplification.
If soundcard has THD+N -20 dB below of  HD800's (-74 dB), let's say  -94 dB then final THD of the soundcard + headphones will be (-94 dB) + (-74dB) = -73.96 dB. Futher improvement of soundcard's THD+N is pretty pointless.


Yes, in that case it is indeed pointless as you've calculated.  However, keep in mind that while a big pair of headphones like the HD800s may be hard to drive all that loud, they don't take all that much current.  If you swapped them out for a pair of very low impedance balanced armature IEMs, you might find that your THD increases a thousand fold or more.  So in principle you may still want better equipment, depending on what you listened to.
Title: 32-bit capable DACs
Post by: naturfreak on 2012-05-26 22:01:10
As far as I know there is a upper physical limit of the achivable Signal-to-noise-ratio (http://en.wikipedia.org/wiki/Signal-to-noise_ratio) in the domain of analog signals.
According to J.B. Johnson this limit is about 131 dB for a bandwith of 20 kHz.
(Johnson–Nyquist noise (http://en.wikipedia.org/wiki/Johnson%E2%80%93Nyquist_noise))
So in theory, 22 bits of sample depth should be sufficient. There is no need for 32 bit DACs, 24 bit DACs are good enough in any case.

Furthermore there are no headphones or speakers that can reproduce more than about 100 dB of dynamic – at least no affordable that I know - without noticeable distortions.
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-26 22:12:12
According to J.B. Johnson this limit is about 131 dB for a bandwith of 20 kHz.

Those are dBm. -131dBm is -161dB
Title: 32-bit capable DACs
Post by: Mr.Duck on 2012-05-26 22:12:14
No, but measurements are.

I mean it's performance is measured to find it's specs (performance specifications).


So in theory, 22 bits of sample depth should be sufficient. There is no need for 32 bit DACs, 24 bit DACs are good enough in any case.

In theory... but I am not convinced. So there is a noise floor limiting measured performance... how about we go the other way and have a DAC outputting ~20V RMS instead of the usual level of around 1-2V RMS. Wouldn't that help achieve better results?
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-26 22:13:41
As far as I know there is a upper physical limit of the achivable Signal-to-noise-ratio (http://en.wikipedia.org/wiki/Signal-to-noise_ratio) in the domain of analog signals.
According to J.B. Johnson this limit is about 131 dB for a bandwith of 20 kHz.


Assuming you're refering to the noise power table on that wikipage, keep in mind that is in units of dBm, that is power relative to 1 milliwatt.  So you are limited to an SNR of 131dB for a 1 mW signal power, but at 1 watt would be limited to 161dB.

That said, you are right that thermal effects become extremely problematic as one gets much above 20 effective bits.  And simply ramping up the power introduces its own problems (such as thermal heating of components) that in turn tend to reduce rather then improve SNR.  In practice, measuring 24 effective bits at room temperature seems to be extraordinarily difficult even for scientific equipment.
Title: 32-bit capable DACs
Post by: naturfreak on 2012-05-26 22:26:07
It ist even worse in the case microphones are involved, because the output signal of a average microphone without amplifier is only around -40 dBm, but the noise level is at around -130 dBm. Therefore 90 dB S/N is the best you can get in this case.
Amplifers produce additional noise. You simply can't lower the level of noise by using amplifiers.
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-26 22:26:39
how about we go the other way and have a DAC outputting ~20V RMS instead of the usual level of around 1-2V RMS. Wouldn't that help achieve better results?


In practice, probably not.  Most of the best devices seem to operate around a few volts.  Even those A/Ds that measure larger voltages tend to use a preamp stage to throw away extra voltage.  All that extra voltage introduces problems of its own, including the need to use much larger resistor values (which are of course much more noisy) to keep self heating down to a reasonable level.  Additionally, high voltage semiconductor logic has its own problems.
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-26 22:46:08
It ist even worse in the case microphones are involved, because the output signal of a average microphone without amplifier is only around -40 dBm, but the noise level is at around -130 dBm. Therefore 90 dB S/N is the best you can get in this case.
Amplifers produce additional noise. You simply can't lower the level of noise by using amplifiers.

Ouch You have just messed up dBm and dB again. 

The noise level in a good silent studio is around (well, it's already pretty ideally)  -120 dB, not -120 dBm. Because -120dBm is much less than -120dB (-120 dBm = -150dB -> it's too low for real-life noise).
Title: 32-bit capable DACs
Post by: naturfreak on 2012-05-26 23:11:21
I'm not talking about the level auf acoustic noise in the studio or atmosphere, Igor. I'm talking about the noise level in electrical parts, wires, circuits.
The difference between -130 dBm and -40 dBm is 90 dB, because the in the calculation of a difference of levels the reference level of 1 mW will be discarded.
So the limitation of S/N is in the electrical domain, not in the possible difference in air pressure.
You can shift the level of loudness, yes, but in the electrical domain S/N will stay the same.
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-26 23:31:26
The difference between -130 dBm and -40 dBm is 90 dB, because the in the calculation of a difference of levels the reference level of 1 mW will be discarded.
So the limitation of S/N is in the electrical domain, not in the possible difference in air pressure.
You can shift the level of loudness, yes, but in the electrical domain S/N will stay the same.

Yes,S/N will remain the same.  It's true.

But there is no ADC with  such low noise -130 dBm (-160dB). -130 dB maybe. -130 dBm - science fiction.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-05-26 23:44:51
For real audio resolution, the number to look for is the THD+N.

Can You defend this statement with some additional information?

Then why include Dynamic Range into specification? Just go for THD+N, right?

The final product is the music. Whatever is not music should be cosidered as altering the original signal. Therefore, for real resolution, you need to look at THD+N numbers. S/N is just the "N" part of that sum. Unless you really like the specific distortions... case for many tube-lovers.
Anyway I am saying that  the physically limit of the human hearing is around 21-22 dB. Even the thermal noise of the air molecules in a 25C room is at that level. There is absolutelly no point for more than 24 bit (at 96kHz) reproduction of audio signals destined to be heard by humans.

DR, as is defined, is kind of a marketing tool. Is the SNR measured at -60dB of the full scale and at the value obtained, add 60dB. Well, that tells how good is the system at low level reproduction, but the distortions are really creeping up at higher levels, close to 0dB. Is was a desperate attmpt to hide the glitching distortion in early multibit DAC's. Not usefull today when all the DAC's have a zero-detection circuit that mutes the output when "zero" PCM code is detected.
Title: 32-bit capable DACs
Post by: naturfreak on 2012-05-27 00:18:21
But there is no ADC with  such low noise -130 dBm (-160dB). -130 dB maybe. -130 dBm - science fiction.

Yes, there is no ADC with this low noise level and the consequence is there is no point designing or producing ADC oder DAC that uses more than 24 Bit of (integer) sample depth.
Are there ADC or DAC using 32 bit floating-point for conversation or internal format?

Anyway I am saying that  the physically limit of the human hearing is around 21-22 dB. Even the thermal noise of the air molecules in a 25C room is at that level. There is absolutelly no point for more than 24 bit (at 96kHz) reproduction of audio signals destined to be heard by humans.

I agree with you.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-05-27 03:11:18
As for ADC reality:
In my E-MU 1820m (http://ixbtlabs.com/articles2/creative-emu-1820/index.html) I have one of the best ADC around, same like in ProTools professional mastering gear: AK5394A. That is capable of "only" THD+N=-110dB... True 18 bits.
Higher level than that there is the DAD-AX24 (http://www.digitalaudio.dk/media/features__benefits.pdf) with THD+N at -115dB (true 19bits) for 10 times the price.

From my findings, many of the old recordings are at a very poor quality comparativelly, derived from the original master analog tapes. Even if they are sold in DVD-A, SACD or digital 24/96 formats, they are actually below the capabilities of 16 bit and with max bandwidth limited often at 17- 18kHz...
Newer recordings have bandwidth brickwalled at 20kHz and are compressed in such a manner that the true dynamic is only at 12-13 bit resolution level (at best).
Title: 32-bit capable DACs
Post by: 2Bdecided on 2012-05-28 14:21:30
I'm quite surprised by some people's reactions, it has been stated here that both ASIO and WASAPI provide no advantage whatsoever over DS therefore being as pointless as 32bit output so, why not ditch these plugins altogether? the same could possibly said about FLAC, APE or any other lossless codec that provides no audible advantage over well encoded 192Kbps, and so on and so forth.
Because that difference is sometimes audible, and always measurable. Whereas 24-bits vs more than 24-bits is, AFAIK, completely and utterly undetectable anywhere in the analogue audio world.

Cheers,
David.
Title: 32-bit capable DACs
Post by: 2Bdecided on 2012-05-28 14:25:31
The noise level in a good silent studio is around (well, it's already pretty ideally)  -120 dB
That's useful! "-120dB" means "one million times smaller than" - so "the noise level in a good silent studio is around one millions times smaller than..." what?!

FWIW I don't know of any studios that are sound proofed down to 0dB SPL across the audible range (around 20dB SPL is a common target, rarely reached across the board) - but even if they exist, you'd better hope that the musicians don't ruin it by breathing.

Cheers,
David.
Title: 32-bit capable DACs
Post by: Roseval on 2012-05-28 17:45:55
I wonder if the “32 bit DAC” has anything to do with dynamic range.
ESS Sabre is advertised as a 32 bit “DAC” simply because they do some DSP (probably ASRC to reduce jitter) and they do so in 32 bit float hence the 32 bit.

Audio editors often allows you to store the project in float.
You don’t want a quantization error each time you save your work.
Can imagine it is convenient to feed this straight into a DAC

In the computer world 24 is a weird number.
8/16/32/64/128 are the common word length.
Might be convenient to keep your audio in 32 or 64 when processing.
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-28 18:14:33
FWIW I don't know of any studios that are sound proofed down to 0dB SPL across the audible range (around 20dB SPL is a common target, rarely reached across the board) - but even if they exist, you'd better hope that the musicians don't ruin it by breathing.

Your reaction is to expect. But if You look into the context of this particular discussion and I've mentioned that those are ideal (pretty unrealistic) numbers.
And yes, those SPL numbers are available even on wiki for easy reading. I was perfectly aware of them. Thank You. 

The message was the other.
Title: 32-bit capable DACs
Post by: saratoga on 2012-05-28 19:02:33
Audio editors often allows you to store the project in float.
You don’t want a quantization error each time you save your work.
Can imagine it is convenient to feed this straight into a DAC


32 bit float has only a 23 bit mantissa, so I tend to think most stuff using floating point uses doubles.

Anyway, DACs take integer precision, not fp, so whatever you are working with is irrelevant.  Its going to be cast to whatever packed integer format the DAC specifies.  Its not like your software even knows that though, the driver keeps the DAC and your programs quite separate. 
Title: 32-bit capable DACs
Post by: db1989 on 2012-05-28 19:10:50
I'm quite surprised by some people's reactions, it has been stated here that both ASIO and WASAPI provide no advantage whatsoever over DS therefore being as pointless as 32bit output so, why not ditch these plugins altogether?
No audible advantage. (Disregard broken software or hardware.) But they exist exactly to provide their other advantages, such as low latency, exclusive access, etc. This has been gone over many times.

Quote
the same could possibly said about FLAC, APE or any other lossless codec that provides no audible advantage over well encoded 192Kbps, and so on and so forth.
Even if the above were not misunderstood, this would not follow from it. Lossless archiving has objective advantages completely separate from any discussion of perceptual transparency: archival and future-proofing are the main ones I can offer now, but that’s not all.
Title: 32-bit capable DACs
Post by: ExUser on 2012-05-28 19:19:26
(Disregard broken software or hardware.)
No need to totally disregard, either, as this is one of the other valid reasons for using these outputs. I found WASAPI output in a Windows VirtualBox to be more reliable than DS.
Title: 32-bit capable DACs
Post by: Kohlrabi on 2012-05-28 21:52:38
I found WASAPI output in a Windows VirtualBox to be more reliable than DS.

And I experienced the same in my wine+foobar2000 endeavours.
Title: 32-bit capable DACs
Post by: Rotareneg on 2012-05-29 03:23:34
But I needed a DAC that could accurately play back a recording of a mouse farting during a space shuttle launch!
Title: 32-bit capable DACs
Post by: 2Bdecided on 2012-05-29 09:35:16
The message was the other.
Oh, I thought you were telling someone they weren't using "dB" correctly, before giving an example that was incorrect by a factor of one million.
Title: 32-bit capable DACs
Post by: ExUser on 2012-05-29 16:28:10
But I needed a DAC that could accurately play back a recording of a mouse farting during a space shuttle launch!
But from precisely how many yards are you recording the mouse fart? It changes everything!
Title: 32-bit capable DACs
Post by: greynol on 2012-05-29 18:37:43
For me it didn't matter.  I had the volume turned off on my TV when I watched the launch.

Forget that the mouse was already in my cat's stomach.
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-30 01:34:08
but even if they exist, you'd better hope that the musicians don't ruin it by breathing.

Breathing can be part of recorded program.  So your treament of it like some unwanted noise is wrong.

Oh, I thought you were telling someone they weren't using "dB" correctly, before giving an example that was incorrect by a factor of one million.

Overplayed.
Show me where my statement was by a factor of one million time wrong.

I think You can apply now your criticism  to your  very own post.
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-05-30 14:17:13
But I needed a DAC that could accurately play back a recording of a mouse farting during a space shuttle launch!


BTW, got any idea what the output of the finest microphone in the world would be like under those circumstances?

Not pretty!
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-05-30 14:23:21
The difference between -130 dBm and -40 dBm is 90 dB, because the in the calculation of a difference of levels the reference level of 1 mW will be discarded.
So the limitation of S/N is in the electrical domain, not in the possible difference in air pressure.
You can shift the level of loudness, yes, but in the electrical domain S/N will stay the same.

Yes,S/N will remain the same.  It's true.

But there is no ADC with  such low noise -130 dBm (-160dB). -130 dB maybe. -130 dBm - science fiction.


Take a DAC with -130 dB noise referenced some higher level than 0 dBm, add a properly designed resistive divider with maybe an op amp buffer, and viola, -130 dBm noise levels.

I believe that there are now commercial DACs with -130 dB FS noise - ESS Sabre DACs.

There are several other DACs that come close Probably very little but money keeps you from connecting a bunch of them in parallel in such away that the noise drops -3 dB every time you double the number of DACs connected in parallel.
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-05-30 14:29:12
DR, as is defined, is kind of a marketing tool.


No, it is a good spec, but yes it is incomplete if you want to know what is going on in all relevant conditions. That's why spec sheets have more than one number on them!

Quote
Is the SNR measured at -60dB of the full scale and at the value obtained, add 60dB. Well, that tells how good is the system at low level reproduction,


Which is a good thing to know, no?

Quote
but the distortions are really creeping up at higher levels, close to 0dB.


Agreed, so we have other specs to characterize them. We expect to see more than one number!

Quote
Is was a desperate attmpt to hide the glitching distortion in early multibit DAC's.


No, it was an effective attempt to expose low level glitching which was far more common before Delta-Sigma converters became the rule.

Quote
Not usefull today when all the DAC's have a zero-detection circuit that mutes the output when "zero" PCM code is detected.


There is no such thing in general.  You need to study up on how modern converters work.
Title: 32-bit capable DACs
Post by: 2Bdecided on 2012-05-31 10:24:49
Show me where my statement was by a factor of one million time wrong.
Quoting a real world sound level as "-120dB".

No units /  1M times wrong in the implied units - take your pick - a physics teacher would fail you either way.


Which isn't the kind of thing I'd usually be so uncharitable as to criticise - but you were laughing at naturfreak for a much smaller error.


I do agree we don't need too many bits. I can hear breathing perfectly well on 16-bit recordings.

Cheers,
David.
Title: 32-bit capable DACs
Post by: IgorC on 2012-05-31 10:59:35
Quoting a real world sound level as "-120dB".

I've mention it with that reference level because if not it would just confuse more Naturefreak.  See discussion first.



No units /  1M times wrong in the implied units - take your pick - a physics teacher would fail you either way.

A physics teacher would fail you teaching that different measurements can mean the same with different reference levels.
And I've already mention why I did it this way.

You pick a  wrong person here to teach about sound pressure  levels and unit of measurements.    Wrong person, wrong topic.

but you were laughing at naturfreak for a much smaller error.

The idea was to make soft a correction. But I can see that smile icon was interpreted some aggressive way of laughing out.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-06-01 01:50:24
There is no such thing in general.  You need to study up on how modern converters work.

Now you upset me...
Want examples? AK4683 (http://www.asahi-kasei.co.jp/akm/en/product/ak4683/ak4683_f03e.pdf), page 1 "Zero Detect Function". PCM1794 (http://www.ti.com/lit/ds/symlink/pcm1794.pdf), page 1 "Zero Flag" (mute external or internal "soft mute"). CS4398 (http://www.cirrus.com/en/pubs/proDatasheet/CS4398_F1.pdf), pags 21 explains the "Mute Control": "use of the Mute Control function can enable the system designer to achieve idle channel noise/signal-to-noise ratios which are only limited by the external mute circuit". All the sigma-delta DAC's will output HF garbage if not muted, as opposed to old-school AD1955 (http://www.analog.com/static/imported-files/data_sheets/AD1955.pdf) page 1 specifies that SNR is 120dB "not muted".

THD+N is the ONLY realistic number that tells the accuracy of audio output. SNR is just a part of that number.
Title: 32-bit capable DACs
Post by: saratoga on 2012-06-01 02:14:27
There is no such thing in general.  You need to study up on how modern converters work.

Now you upset me...
Want examples? AK4683 (http://www.asahi-kasei.co.jp/akm/en/product/ak4683/ak4683_f03e.pdf), page 1 "Zero Detect Function".


From your link:

Quote
Zero Detect
When the PCM1794 detects that the audio input data in the L-channel and the R-channel is continuously zero for
1024 LRCKs in the PCM mode or that the audio input data is continuously zero for 1024 WDCKs in the external filter
mode, the PCM1794 sets ZERO (pin 13) to HIGH.


So if the source turns off, it alerts the controller so it can turn off the amp.  I don't really understand what you're trying to get at, but I don't think this is it. Its common to turn off circuits when they're not being used for power saving reasons.  Modern efficiency standards more or less require it.
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-01 12:13:46
There is no such thing in general.  You need to study up on how modern converters work.

Now you upset me...
Want examples? AK4683 (http://www.asahi-kasei.co.jp/akm/en/product/ak4683/ak4683_f03e.pdf), page 1 "Zero Detect Function".


From your link:

Quote
Zero Detect
When the PCM1794 detects that the audio input data in the L-channel and the R-channel is continuously zero for
1024 LRCKs in the PCM mode or that the audio input data is continuously zero for 1024 WDCKs in the external filter
mode, the PCM1794 sets ZERO (pin 13) to HIGH.


So if the source turns off, it alerts the controller so it can turn off the amp.  I don't really understand what you're trying to get at, but I don't think this is it. Its common to turn off circuits when they're not being used for power saving reasons.  Modern efficiency standards more or less require it.


Agreed.

This is a high-function DAC with a built-in level control. An ideal digital level control that allows level changes at arbitrary times will cause what is called "The zipper effect" which is essentially a series of clicks caused by manually operating an ideal digital level control. This effect can be avoided or minimized by allowing level changes to only occur when the signal naturally passes through zero.

Similar logic is used in light dimmers to minimize EMI generation.

As an aside in the early 1970s I devised a FM muting circuit that seemingly acted instantly and fully but without loud clicks, by waiting for a natural zero-crossing to change state.
Title: 32-bit capable DACs
Post by: knutinh on 2012-06-01 13:36:23
There are several other DACs that come close Probably very little but money keeps you from connecting a bunch of them in parallel in such away that the noise drops -3 dB every time you double the number of DACs connected in parallel.

I was under the impression that this kind of statement makes assumptions about the DAC errors that is not totally true. I.e. you cannot achieve any (large) increase in "signal to error level" by doubling the number of DAC chips a (large) number of times. They will have some signal-dependant non-linearity, clock pulse bleed-through or whatever that will add constructively, and at some point these errors will become the significant source of error (that does not improve by increasing the amount of DACs).

I don't know about using totally different designs (how many fundamentally different high-quality DAC chips are available?), or liquid Nitrogen cooling?

Of course, going from e.g. 128 to 256 audio ADC/DAC chips is mostly an academic exercise anyways. If the true gain is 1dB less error rather than the expected 3dB, very few of us will ever notice (due to the low levels in the first place, and the low probability that anyone finds it practical to use such a setup)

-k
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-01 13:52:07
There are several other DACs that come close Probably very little but money keeps you from connecting a bunch of them in parallel in such away that the noise drops -3 dB every time you double the number of DACs connected in parallel.

I was under the impression that this kind of statement makes assumptions about the DAC errors that is not totally true. I.e. you cannot achieve any (large) increase in "signal to error level" by doubling the number of DAC chips a (large) number of times. They will have some signal-dependant non-linearity, clock pulse bleed-through or whatever that will add constructively, and at some point these errors will become the significant source of error (that does not improve by increasing the amount of DACs).


Agreed. Hence my careful choice of words: "noise drops".  The Sabre DACs may be specifically designed to be free of most of the other non-noise artifacts as they recommend paralleling up to 4 DACs on the same chip. They give SNR specs for this configuration, and they do specify that the resulting nonlinear distortion is well above the noise. Ya got to be a spec sheet lawyer! ;-)

Quote
I don't know about using totally different designs (how many fundamentally different high-quality DAC chips are available?), or liquid Nitrogen cooling?

Of course, going from e.g. 128 to 256 audio ADC/DAC chips is mostly an academic exercise anyways. If the true gain is 1dB less error rather than the expected 3dB, very few of us will ever notice (due to the low levels in the first place, and the low probability that anyone finds it practical to use such a setup)


I don't' know how far people have actually gone with this game. It gets real expensive if you try to go very far. And as you say for what?  Even one DAC on the 4 section chip is vast overkill.


Title: 32-bit capable DACs
Post by: SoNic67 on 2012-06-03 19:20:47
Paralelling is done especially on passive I/V conversion. For normal OpAmp conversion, the limit is the OpAmp noise.
After a few paralleled DAC's, the noise still gets lower (more current allows for lower passive I/V resistance), but the distortion doesn't, it "settles" to the architecture minimum value. Noise is statistically uncorrelated, distortion is correlated.
For Sabre32 9018, the SNR can go as low as -135dB using paralleling of the 8 internal DAC's (129dB 8ch, 133dB stereo, 135dB mono), but THD remains in all cases at -120dB (20 bit quality). This is their data (http://www.esstech.com/PDF/SABRE32%20Reference%20PB%20Rev%200.6%20110117.pdf). At some point there is no quality return for paralleling DAC chips.
Title: 32-bit capable DACs
Post by: IgorC on 2012-06-03 19:50:33
Those are nice and probably real specifications.
The thing is that there are no headphone or speaker which will benefit from such high precision. The best headphones have THD -80 dB (best case).  The  sum of -80 dB + DAC's THD -100 dB is still the same - 80dB.  DAC's THD better than -95 ... -100 dB  doesn't improve anything in real audio experience but the specifications itself.

It makes only sense if phones have low impedance as saratoga has already mention (but still You won't get any better than -100dB)
Title: 32-bit capable DACs
Post by: saratoga on 2012-06-03 22:11:54
The thing is that there are no headphone or speaker which will benefit from such high precision. The best headphones have THD -80 dB (best case).  The  sum of -80 dB + DAC's THD -100 dB is still the same - 80dB.


THD is relative to the signal intensity though.  For dynamic music, you might have loud and quiet passages, in which case having a lower noise floor then the THD implies is still worthwhile so that you don't hit the noise floor during softer portions of a track.

I agree that having a noise floor less then 90-100dB below your typical listening level is entirely pointless. 

Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-04 13:21:57
The thing is that there are no headphone or speaker which will benefit from such high precision. The best headphones have THD -80 dB (best case).  The  sum of -80 dB + DAC's THD -100 dB is still the same - 80dB.


THD is relative to the signal intensity though.  For dynamic music, you might have loud and quiet passages, in which case having a lower noise floor then the THD implies is still worthwhile so that you don't hit the noise floor during softer portions of a track.


Do you know of anybody with a logical interest in -120 dB THD and 129 dB SNR than test equipment manufacturers?  I guess the 120dB+  SNR might make sense for a preamp-in microphone application as it eliminates the need for a remote volume control.

Quote
I agree that having a noise floor less then 90-100dB below your typical listening level is entirely pointless.


Agreed, unless you have someone who wants to hook 100 dB/w speakers to 1 Kw power amps and not hear any noise when they put their ears next to the speakers.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-06-05 01:09:57
I can hear very clear sounds at -60dB (running volume at normal level of audition) - based on spectrum analizers. I would like to have as less distortion as possible at that level. If the DAC has only -80dB THD+N, that leaves only 20dB between sounds and distortion/noise. That is perfectly audible, it will influence the sound... Any PC speakers with a 'normal' THD+N of -85dB will prove easily that fact.
If the DAC has -110dB THD+N, that leaves -50dB between sound and noise/distortion. I cannot hear that and that means that the DAC will not "colour" the sound.

Now... headphones/speakers have listed -80...90dB THD, but that is at nominal levels of power. At lower levels, because of different mechanism of producing those distortions (mechanical histerezis of suspension and limited membrane rigidity), distortion levels drop significantly. Mind also that thermal noise is generated in speakers/headphones is very, very low due to low impedance. A good pair of headphones probably can achieve -100...105dB of THD when are 'run' at -20dB levels - sufficient for a normal hearing.

Anyway, that means only 20 bit of real resolution!!! So yes, 24 bit is more than enough for audio.
Title: 32-bit capable DACs
Post by: saratoga on 2012-06-05 01:22:23
I can hear very clear sounds at -60dB (running volume at normal level of audition) - based on spectrum analizers.


-60dB relative to what? 

I would like to have as less distortion as possible at that level. If the DAC has only -80dB THD+N, that leaves only 20dB between sounds and distortion/noise.


I'm not sure I understand what you're trying to say, but I don't think your logic here is correct.  At least not the 20dB part.
Title: 32-bit capable DACs
Post by: IgorC on 2012-06-05 03:52:06
Now... headphones/speakers have listed -80...90dB THD ...

I'd  be grateful if You could provide a link to specifications for headphones or speakers which have THD better than -80 dB?  Or at least -80 dB.
Example, Sennheiser HD800 has THD less than -74 dB. Maybe a bit less, but not that much. It hardly hits -80 dB. The problem is that it's expensive  and presents only less than 1% of all market of headphones.

A good pair of headphones probably can achieve -100...105dB of THD when are 'run' at -20dB levels - sufficient for a normal hearing.

THD of headphones doesn't improve with a lower levels.  Example http://www.innerfidelity.com/images/SennheiserHD650.pdf (http://www.innerfidelity.com/images/SennheiserHD650.pdf)
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-05 13:00:02
I can hear very clear sounds at -60dB (running volume at normal level of audition) - based on spectrum analizers.


This situation probably ignores the effects of masking which are huge.


Quote
I would like to have as less distortion as possible at that level. If the DAC has only -80dB THD+N, that leaves only 20dB between sounds and distortion/noise. That is perfectly audible, it will influence the sound... Any PC speakers with a 'normal' THD+N of -85dB will prove easily that fact.


A DAC or headphone amp with only -80 dB (0.01%) spurious responses is by modern standards, a serious POS among SOTA devices.  100 dB shouldn't be rocket science. But, its overkill.

Quote
If the DAC has -110dB THD+N, that leaves -50dB between sound and noise/distortion. I cannot hear that and that means that the DAC will not "colour" the sound.


Just as people can drown in 3 inches of water and dead is dead no matter how shallow the water is, 10 dB below audibility is for all practical purposes as good as 50 dB or 100 dB.

Quote
Now... headphones/speakers have listed -80...90dB THD, but that is at nominal levels of power. At lower levels, because of different mechanism of producing those distortions (mechanical histerezis of suspension and limited membrane rigidity), distortion levels drop significantly. Mind also that thermal noise is generated in speakers/headphones is very, very low due to low impedance. A good pair of headphones probably can achieve -100...105dB of THD when are 'run' at -20dB levels - sufficient for a normal hearing.


You badly need to find out what humans can actually hear in the way of nonlinear distortion. I feel like I'm listening to a runner who thinks he can run 10 second miles!

Quote
Anyway, that means only 20 bit of real resolution!!! So yes, 24 bit is more than enough for audio.


So is 16 bits in distribution formats.
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-05 13:06:02
THD of headphones doesn't improve with a lower levels.  Example http://www.innerfidelity.com/images/SennheiserHD650.pdf (http://www.innerfidelity.com/images/SennheiserHD650.pdf)


That's probably only because you are looking at THD+N, not the results of a proper spectrum analysis.  At lower levels, the noise may make the results look higher than the nonlinear distoriton actually is.

Typically, nonlinear distortion goes down with level, unless it is due to a discontinuity, such as crossover distortion. Crossover distortion is essentially impossible in a sigma-delta DAC (that covers just about every modern DAC for audio), and exceedingly rare or absent in properly designed amplifiers. Transducers can't have sharp discontinuities in their transfer curves unless there is something like sticking.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-06-06 01:11:55
Now... headphones/speakers have listed -80...90dB THD ...

I'd  be grateful if You could provide a link to specifications for headphones or speakers which have THD better than -80 dB?  Or at least -80 dB.

You didn't read my whole statement... at nominal levels (that can damage hearing) THD+N is at -80dB...-90dB. Lower that at norma listening and we are talking a different story. As I said, that's because of hysterezis of friction (suspension), membrane finite rigidity (for bass)...
http://www.headphone.com/headphones/grado-gs-1000i.php#tabs (http://www.headphone.com/headphones/grado-gs-1000i.php#tabs)
Title: 32-bit capable DACs
Post by: saratoga on 2012-06-06 02:39:12
Now... headphones/speakers have listed -80...90dB THD ...

I'd  be grateful if You could provide a link to specifications for headphones or speakers which have THD better than -80 dB?  Or at least -80 dB.

You didn't read my whole statement... at nominal levels (that can damage hearing) THD+N is at -80dB...-90dB. Lower that at norma listening and we are talking a different story. As I said, that's because of hysterezis of friction (suspension), membrane finite rigidity (for bass)...
http://www.headphone.com/headphones/grado-gs-1000i.php#tabs (http://www.headphone.com/headphones/grado-gs-1000i.php#tabs)


That link says -60dB . . .
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-06 14:10:04
Now... headphones/speakers have listed -80...90dB THD ...

I'd  be grateful if You could provide a link to specifications for headphones or speakers which have THD better than -80 dB?  Or at least -80 dB.

You didn't read my whole statement... at nominal levels (that can damage hearing) THD+N is at -80dB...-90dB. Lower that at norma listening and we are talking a different story. As I said, that's because of hysterezis of friction (suspension), membrane finite rigidity (for bass)...
http://www.headphone.com/headphones/grado-gs-1000i.php#tabs (http://www.headphone.com/headphones/grado-gs-1000i.php#tabs)


That link says -60dB . . .


Yes, and the frequency scale seems to be misleading.  The first spurious response appears to be second harmonic.
Title: 32-bit capable DACs
Post by: knutinh on 2012-06-06 21:18:41
I don't' know how far people have actually gone with this game. It gets real expensive if you try to go very far. And as you say for what?  Even one DAC on the 4 section chip is vast overkill.

I guess it might make sense for some measurement/lab applications. I certainly think it is interesting to see how far things can be taken.

Distortion that is signal-correlated (predictable?) might also be reduced through a feed-forward or feedback-loop?

I am also depressingly aware of how some people would misguide and be misguided into believing that their records would sound audibly better for using one.

-k
Title: 32-bit capable DACs
Post by: IgorC on 2012-06-06 21:38:25
You didn't read my whole statement... at nominal levels (that can damage hearing) THD+N is at -80dB...-90dB. Lower that at norma listening and we are talking a different story. As I said, that's because of hysterezis of friction (suspension), membrane finite rigidity (for bass)...
http://www.headphone.com/headphones/grado-gs-1000i.php#tabs (http://www.headphone.com/headphones/grado-gs-1000i.php#tabs)

You're too far to get such good THD(+N) with these headphones on listenable levels.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-06-07 00:20:47
Sooo... In conclusion, we need 32 bit audio?
Title: 32-bit capable DACs
Post by: stephan_g on 2012-06-07 00:46:22
Typically, nonlinear distortion goes down with level, unless it is due to a discontinuity, such as crossover distortion. Crossover distortion is essentially impossible in a sigma-delta DAC (that covers just about every modern DAC for audio), and exceedingly rare or absent in properly designed amplifiers. Transducers can't have sharp discontinuities in their transfer curves unless there is something like sticking.

I wouldn't say crossover distortion is absent... people do tend to go to considerable pains to keep it in check in low-quiescent-current applications where it easily dominates high-frequency distortion at higher levels.
Borrowing a graph with horribly messed-up labeling from NwAvGuy:
(http://lh3.ggpht.com/-Pibk5F7iukI/TkxEfQqNWxI/AAAAAAAABLo/Rqjc0YoryjQ/s1600/O2%2520V11%2520Batt%2520Pwr%2520Gain%2520Stage%2520Op%2520Amps%25202%2520V%2520RMS%2520THD%252BN%2520vs%2520Frequency%2520BW%253D22%2520Khz%2520See%2520Legend%255B2%255D.png)
(real labels should be, from bottom to top at 9 kHz: NJM2068 2.5x, NJM2068 7x, OPA2277 2.5x, OPA2277 7x)

The OPA2277 already was the best-performing out of several low-power parts he tried. And yes, of course the decay in distortion above ~10 kHz is only an artifact due to a 20 kHz LPF that filters out the harmonics.

The different characteristics and according distortion spectra of electronics vs. transducers mean that comparing distortion specs has to be carried out with care. Transducers usually have distortion dominated by low-order (2nd/3rd) products which still show considerable masking, while with electronics it is easy to generate a near-flat "zoo" of harmonics that decays very slowly, resulting in much tighter (numerical) requirements since now high-order harmonics have to be kept below the threshold of audibility (usually <=-80 dBr) at all times. Discontinuities + feedback at work there.

Incidentally, since headphone drivers also tend to be numerically optimized these days, I'd expect a rising discrepancy between objective and subjective evaluation of level handling. Same as when introducing feedback, basically - less low-order, more higher-order products. I'd rather not be trading 0.2% less 2nd harmonic for 0.03% more 7th or 8th (since it is quite possible to hear the latter, while -54 dB of 2nd should be a little below the audible minimum found in studies).

Re: why it is useful to have DR > THD+N, one only needs to consider that the required instantaneous dynamic range in any given playback or recording situation is usually smaller than the total dynamic range for all of them. For example, you may want your 80-odd dB of instantaneous DR regardless of whether you're using speakers with a sensitivity of 85 dB SPL / 2.83 V / m or others with 105 dB. (Or headphones with 95 dB / V vs. others with 130 dB / V.) It's mostly a matter of convenience and not having to worry about stuff.
Distortion that is signal-correlated (predictable?) might also be reduced through a feed-forward or feedback-loop?

The classic predistortion? That would seem worth a shot. Each batch of chips may need a different set of coefficients (internal EEPROM?), but it should be doable. Obviously that only applies if the analog part is the dominant source of distortion.

Premium DACs are already using balanced output so they're largely getting rid of even-order distortion, but what did one have to do to get rid of odd order again? Think it involved a Hilbert xform for a 90° phase shift? Sounds kinda impractical but you never know.

As far as 32-bit DACs are concerned... seems like you're losing about 4 bits with undithered output, so you're only getting 20 bits from a 24 bit converter. At 192 kHz we're already having 4X oversampling though, for an effective 22 bits in the audio band. So if anything, it would be of concern for 44/48 kHz and more or less "synthetic" material. A very, VERY small fraction of cases.
Title: 32-bit capable DACs
Post by: nevermind on 2012-06-07 01:56:42
I wouldn't say crossover distortion is absent... people do tend to go to considerable pains to keep it in check in low-quiescent-current applications where it easily dominates high-frequency distortion at higher levels. ...  The OPA2277 already was the best-performing out of several low-power parts he tried.

You do realise that the increase in high frequency distortion is due to the reduction in open loop gain as frequency increases, it has nothing to do with crossover distortion. The opa2277 is the best of all because it has 20db extra loop gain, sort of cheating, really. Oh, and both legends are wrong.

Edit:oops I was thinking of the opa2227.


Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-07 12:39:42
Typically, nonlinear distortion goes down with level, unless it is due to a discontinuity, such as crossover distortion. Crossover distortion is essentially impossible in a sigma-delta DAC (that covers just about every modern DAC for audio), and exceedingly rare or absent in properly designed amplifiers. Transducers can't have sharp discontinuities in their transfer curves unless there is something like sticking.

I wouldn't say crossover distortion is absent... people do tend to go to considerable pains to keep it in check in low-quiescent-current applications where it easily dominates high-frequency distortion at higher levels.
Borrowing a graph with horribly messed-up labeling from NwAvGuy:

(http://lh3.ggpht.com/-Pibk5F7iukI/TkxEfQqNWxI/AAAAAAAABLo/Rqjc0YoryjQ/s1600/O2%20V11%20Batt%20Pwr%20Gain%20Stage%20Op%20Amps%202%20V%20RMS%20THD%20N%20vs%20Frequency%20BW=22%20Khz%20See%20Legend[2].png)
(real labels should be, from bottom to top at 9 kHz: NJM2068 2.5x, NJM2068 7x, OPA2277 2.5x, OPA2277 7x)

The OPA2277 already was the best-performing out of several low-power parts he tried.


Key words: "He tried". Just two of how many 100s of similar parts.

You seem to be oblivious to the fact that the above plot is totally inappropriate for the purpose of showing the existence or absence of crossover distortion.

The OPA 2277 is not designed or recommended as a headphone amplfier by its manufacturer: OPA 2277 Manufacturer's spec sheet and application information (http://www.ti.com/product/opa2277&lpos=Middle_Container&lid=Alternative_Devices) Neither is the NJM 2068. NJM 2068 Manufacturer's spec sheet (http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CG0QFjAA&url=http%3A%2F%2Fsemicon.njr.co.jp%2Fnjr%2Fhp%2FfileDownloadMedia.do%3F_mediaId%3D9349&ei=apLQT-GxIYWI6QHozul6&usg=AFQjCNHGPP_lwgHVzyfF_RPpRuzFRwPEqg)

Letsee: Irrelevant data about irrelevant parts.

<Shaking head>
Title: 32-bit capable DACs
Post by: Arnold B. Krueger on 2012-06-07 12:48:08
Sooo... In conclusion, we need 32 bit audio?


We live in a world with -60 dB distortion ears, -80 dB transducers, and -100 dB or more electronics. Obviously, we need better transducers and electronics. ;-)

The problem with the realism of our recordings has nothing to do with the vanishingly low nonlinear distortion in the equipment that we have been discussing.

The frequency response curves of the headphones we've been discussion would be hilarious, if we knew what their frequency response is supposed to be. The way that the FR of these headphones is being measured is controversial to say the least.

Then there is the problem of how to mic a live performance or mix studio recordings for realism. I see no convincing arguments that anybody really knows how to do this.

So, lets continue to obsess about the parts of the chain that are already way into diminishing returns and ignore the real problem, because the real problem must to too hard.

<shaking head>
Title: 32-bit capable DACs
Post by: stephan_g on 2012-06-08 17:29:37
The real elephant in the room is acoustics, no doubt about that. I'm sure there are plenty of people out there with kilobuck amps running broadly radiating bookshelf speakers in rooms with all the wonderful acoustic properties of a gymnasium. (Translation: Ouch!)

I grew up on headphones, which is why I gravitated to a nearfield setup quite naturally. A bit of obsession with amplifiers isn't for nought in this world, since then at least you can guess why your speakers are hissing or buzzing at you (neither being uncommon among less expensive powered speakers, and even integrated amps/receivers still show noticeable hiss at times).
Key words: "He tried". Just two of how many 100s of similar parts.

Then you try finding a better-performing DIP-8 part for a line-level gain stage (Rf = 1k5, Rg according to gain, load = 10k) with low quiescent current, run on about +/-8.5 or +/- 12 V, respectively. Preferably one that doesn't cost an arm and a leg.
You seem to be oblivious to the fact that the above plot is totally inappropriate for the purpose of showing the existence or absence of crossover distortion.

It's hard to say for sure without having examined the opamp Samuel Groner style, that's true. Could be ordinary VAS common mode distortion, too. But my educated guess would be crossover distortion, since the output stages on low-power parts are always starved for current (as they tend to contribute a large portion of idle current draw) and output loading isn't exactly negligible in this particular application. Output loading was the main reason why e.g. a TL072 wasn't even in the game.
The OPA 2277 is not designed or recommended as a headphone amplfier by its manufacturer: OPA 2277 Manufacturer's spec sheet and application information (http://www.ti.com/product/opa2277&lpos=Middle_Container&lid=Alternative_Devices) Neither is the NJM 2068. NJM 2068 Manufacturer's spec sheet (http://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=1&ved=0CG0QFjAA&url=http%3A%2F%2Fsemicon.njr.co.jp%2Fnjr%2Fhp%2FfileDownloadMedia.do%3F_mediaId%3D9349&ei=apLQT-GxIYWI6QHozul6&usg=AFQjCNHGPP_lwgHVzyfF_RPpRuzFRwPEqg)

Sorry, but you're way off-base on this one. We're talking about the line-level gain stage in the O2 (http://nwavguy.blogspot.com/2011/08/o2-summary.html) amp, parameters as outlined above. It has to drive nothing but its feedback network (2k5 and 1k75, respectively) and the 10k volume pot. Driving headphones then is performed by a buffer using two NJM4556s in parallel.

NJM2068s aren't uncommon in mass-produced consumer and pro audio equipment. (Yamaha integrated amps and EMU and some Creative soundcards, to name a few.) They're about 5532 class devices, with lower voltage noise and current draw, but less good load driving. Pretty cheap, too. (And I never paid all that much attention to it, but they can also be had in the outdated SIP case, RoHS compliant even. Cool. I've been looking for a less noisy replacement for a 4558L in an old ProLogic receiver's gain stage for ages.)

I guess the most practical approach would be sacrificing a little noise performance and easing up on the feedback network values for reduced output loading. In practice, gain stage noise would commonly be attenuated significantly anyway, besides the vast majority of sources is considerably noisier than the gain stage. The design objective for the O2 was getting about as close to a wire with gain as possible without having to use any fancy boutique parts, and once in a while that involves pushing things a little.
Title: 32-bit capable DACs
Post by: pdq on 2012-06-08 17:58:26
This is getting pretty far off from the original topic, but it is actually easy to turn an op amp into a class A device simply by applying a bias current to the output. You trade off power consumption for linearity, and you eliminate crossover distortion.
Title: 32-bit capable DACs
Post by: SoNic67 on 2012-06-10 18:33:26
Original topic was "Do we need 32 bit DAC's?".
Question started from the present fashion of implementing "32 bit audio" on all the DAC's and soundcards, regardless if the inside DAC's are only 24 bit devices (and I added that almost all the present day DAC's are limited in real-life performance to 20-21 bit resolution).