Is there an audible difference between 24/192 and 16/44 in your CAS system?

Is there an audible difference between 24/192 and 16/44 in your CAS system?
Can you tell the difference in a blind test? I can't even in a non blind test, in my system. This can only mean 3 things.
1. My system, esp the CAS system is not up to the mark.
2. My hearing skill and hearing capability.
3. That there is actually no audible difference.

I must confess, I did not compare the same track in 24/192 or 24/96 with 16/44. I listened to many different tracks and in many occasions and I know my system well enough to make this conclusion. Am I alone with this conclusion?

Can tell us more of your system configuration? Especially the digital player/transport.

I am using PC, Foobar karnell streaming, HiFace, Coax out. It able to play until 24bits/192kHz

I listen to the same song that I download from net in 3 formats, 16/44, 24/96 and 24/192. I hear the difference. It could be the way the guy compile the different format, but I like 24/192 more.

The Music Server is made by a reputable Chinese DIY company. HERE.
It reads music file from 3 inputs, USB, SD Card and HDD and is 24/192 capable. This player is connected to my Weiss DAC202 via a SPDIF coax. It has a built in DAC but I have removed the DAC board, thus it is a pure digital device. I hear high resolution reproduction even in 16/44 format but the other formats, 192 or 96kHz are not better.

16/44 wld sound like from a good Cd playback.

24/96 wld be quite a drastic difference as compared to 16/44.
24/192 compared to 24/96 wld not be so distinct but in a good system, it is much better.

Some hirez files are upsampled version and are inferior as compared to those from master tapes.

From the computer guy's point of view, A LOT depends on the source of your digital music. In theory, higher sampling rate will have a very big advantage and almost always give more 'accurate' sound, i.e. a sound that is closer to the original. But that original is the live sound. When we go into the digital realm, then the way we have to fit the live sound into our technology sets the limits.

For e.g., what is the master from which the music comes from ?
1. Original analog master tape ? In this case, provided a very good ADC is used, then the higher resolution music possesses a huge advantage.
2. Original DIGITAL master tape ? In this case, EVERYTHING depends on the FORMAT in which the recording was made.

For e.g. I once read in a forum elsewhere where someone made the statement "24/192kHz will ALWAYS be the closest to the original recording". Is that true ? It is not true that 24/192kHz will always be the closest because at one time, many digital recordings were done using a different format. For e.g. many of the Telarc recordings were done using a Soundstream recorder which records at 16bit, 48kHz. The much sought after M&K 'real-time' recordings were done using a SONY PCM recorder which is also 16bit by 48kHz.

So what happens now ? In this case, the RBCD format becomes the CLOSEST to the original recording we interpolate the least. For every other higher resolution format, we have to interpolate a lot more ! 24/192kHz is NOT closer to a PCM 16/48 than 16/44.1. It is in fact a lot more inaccurate since we have to fit the 16bit signal into a 24bit signal, and then recreate 4 extra samples for each sample in the original recording.

For example, which is more accurate to the original master for the recently issued Tom Petty/Mojo album ? In this example, the original was recorded in 24/96 so the most accurate possible is an exact bit for bit transfer of the original master into a 24/96 file.

So when you talk about the sonic difference between a 16/44.1 and a higher res format, you first need to know what the original recording master is in. Then only we can make firmer conclusions. It is not always that higher resolution is guaranteed better. Unfortunately life is always not this straightforward.

i think the only way to find out if 16/44.1 sounds different from 24/192 is to play the same song/album in both formats but the 24/192 version must not be an upsampled/'upbitted' version of the 16/44.1 version i.e. it has to be a native 24/192 version.

it can be quite difficult to find such examples.

i believe digital music files must be played in their native sampling frequencies.

all the upsampling and 'upbitting' tricks can cause timing errors, i think.

btw, ripped cd files sound better than the cd played on cd player even though both are 16/44.1 files.

THe issue is that if we are dealing with digital recordings, then we will almost NEVER have a fair comparison. There are only that many recording formats - 16/48, 24/96, and 24/192 being the most popular. And these same formats or roughly the same format are available for playback. So whatever formats we compare, ONE will be the closest to the original recording and all others are DERIVED from that recording. And obviously the one closest in format will have the biggest advantage (other implementations being equal).

I suppose the best way to compare fairly is when it is from an analog master tape. But that begs the question, if the record is available, why bother with the digital version ?

sflam wrote:
btw, ripped cd files sound better than the cd played on cd player even though both are 16/44.1 files.

I suppose this would depend on the device capability to resolve the data. In my system the transporter sounds much better than the music server.


@WongKN

Your comments make sense.

wong kn wrote:


I suppose the best way to compare fairly is when it is from an analog master tape. But that begs the question, if the record is available, why bother with the digital version ?

if the master copy of the song is digital to begin with - and most new songs are recorded in digital 24/96 (using protools, i think) and if the song is then downsampled to 16/44.1 for cd release, the cd would sound inferior to the digital 24/96 master.

u can try this with a hires downloaded file for e.g. linn's studio master flac files at 24/88.2. select a few songs and burn to a cd-r (the 24/88.2 would be reduced to 16/44.1) and then play the cd-r on the cd player and test against the original 24/88.2 streamed from laptop.

the cd-r will sound inferior to the 24/88.2 studio master flac file.

Yes sflam, that is exactly what I was saying, i.e. we play at whatever is the original format or nearest to the original format; 16/48, 24/48, 24/96, and so forth and we should get the best possible sound comparing to a different format, again assuming everything else is the same. So it is not always that the highest specification will always be the best. Of course if the interpolation is done extremely well, then a 24/192 -approximation- (it will ALWAYS be an approximation if it is not equal to the original format) can potentially sound better but that is because it allows the rest of the DAC process to work better (i.e. a less sonically degrading filter after the conversion, etc).

But in terms of what is closest to the master, then it is the format closest. So in terms of best potential for sound, a highest resolution recording played back at the same resolution would theoretically give the best sound for digital, i.e. 24/192 recording master stored as 24/192 and played at 24/192. But a 16/44.1 played back at 24/192 may or may not be better than 16/44.1 played back at 16/44.1. It depends a lot on how we derived that extra samples. I think someone also pointed this out above also.

DrWho, glad you agree. Hope I didn't introduce too much confusion into your digital music selection.

wongkn,

looks like we hv answered DrWho's question: Is there an audible difference between 24/192 and 16/44 in your cas system?

the answer is "yes" (but in my case, its 24/88.2 vs 16/44.1)

Thank you guys for your contributions. The key word here is audible. If in a blind test with 16/44 and 24/192 tracks been played, are you able to distinguish which tracks are 16/44 and which are 24/192. If you could segregate them correctly then there is an audible difference. If not, then it is all theoretical and it is placebo.

This topic is pretty interesting...should we be planning a H4S gathering to try out some configurations?

I have needledropped quite a few of my LPs and I daresay that the 24bit certainly sounds better than the 16bit version. On my system anyway.

I would venture to say that the difference between 24bit and 16bit will be more pronounced when you move up the chain in terms of your playback system's quality.

I just came across this article, an interview with Daniel Weiss. It is a lengthy article. Here is an extract of what is relevant with our discussions.

Q. So anyone claiming to hear a difference between 96 and 192 would either be hearing placebo or in how it was converted to that sampling rate?

A. That is a point of course. I know of professional people having done tests with these conversions and they can hardly hear a difference, even if it’s up to 192 and down to 44,1 again.

If if want to tell if there is a difference, the easiest is to listen to a track of the same song in 24/96 and 16/44.1 via a headphones.

It's hard to find similar tracks in different sampling rates. what you can do is to get a geniune 24/96 track and down-sample it using software.

The track must be genuinely recorded in 24/96 or higher, and not an upsampled one.

Some types of music are easier to make out than others. Exeriment with drum tracks and piano, preferable well recorded and with very few instruments.

There is definitely a significant difference between 24/96 and 16/44.1
Between 24/96 and 24/192, the difference is harder to discern.

DrWho,
HiFi is a voodoo, the argument is the same as capacitor or cable or snake oil etc. Everyone hearing is different, there is not much scientific data to support it and too many variables. What you can't listen to doesn't mean it is not there. If you go to a HiFi shop and say give me a 90% accurate system and the thing send to you is exactly 90% no matter how you place or tune it. What is the fun? If I tell you different pc to play the same music file also different sound you will think I am crazy, cause it is 1s and 0s we are talking about.
Don't worry, if you can't hear the different so be it, you should be happy as 16/44 music file is small in size, easier to download and easily direct extract from CD. As long as you are happy, it doesn't matter what other people say because you are paying and you listen to it not him.

carz wrote:If if want to tell if there is a difference, the easiest is to listen to a track of the same song in 24/96 and 16/44.1 via a headphones.

It's hard to find similar tracks in different sampling rates. what you can do is to get a geniune 24/96 track and down-sample it using software.

The track must be genuinely recorded in 24/96 or higher, and not an upsampled one.

Some types of music are easier to make out than others. Exeriment with drum tracks and piano, preferable well recorded and with very few instruments.

There is definitely a significant difference between 24/96 and 16/44.1
Between 24/96 and 24/192, the difference is harder to discern.

It is a known fact that using domestic device/software to upsample/downsample will degrade the sound. Only Pro Audio equipment could do the job correctly. It is therefore not surprising that you are hearing inferior sound.

@kwwong,
Thank you for your kind advice.

here's a good site to download 16/44.1 and 24/96 music for testing. The original recordings has been done on 24/96 and then downsampled to 16/44.1
(the downsampling is done at the studio)

So you have exactly the same music with diff sampling/resolution for testing. If you can't hear the difference on your system, try to listen to them on your headphones. If you then can hear it on your headphones, then you know that somewhere in the chain, your equipment resolution is not detailed enough.

http://www.lessloss.com/high-resolution-audiophile-recordings-c-68.html

Thank you carz, I will give this a try a report back.

DrWho wrote:It is a known fact that using domestic device/software to upsample/downsample will degrade the sound. Only Pro Audio equipment could do the job correctly. It is therefore not surprising that you are hearing inferior sound.

@kwwong,
Thank you for your kind advice.

Yes, i read that upsampler/downsamplers are all not equal, some better than others.

But apparently those done on the PC with good professional software is much better than those done "on-the-fly" by DAC/CD Players. On the PC, the software may take many hours to do an album, so maybe they do a more thorough job.

Weiss Saracon is one of them. I think this is under their professional website. Consumer products is under www.weiss-highend.ch
http://www.weiss.ch/p2d/p2d.html

I read iZotope SRC UPSAMPLER is better than Weiss Saracon and measures better. I have no personal experience with any of these though.

Here is a website that test and compare various different products. May be of interest to you.
http://src.infinitewave.ca/

KORG AudioGate is a Pro Audio converter software which was once exclusively meant for their Pro Audio device. It is now a freeware.

Well, OK, one more detailed explanation from the computer guy, for the benefit of those interested to read/know about it.

When we UP-sample, we basically create new data out of existing data. New data means they did not exist before. So if we want to be downright rude, we can also call them 'fake' data. Anyway the gist is this, say we work with a 48kHz recording (to make the maths a bit simpler), this means we have 48,000 samples per second. And each sample is roughly 0.000021seconds apart. Now we want to go to 96KHz but we have only this 48kHz to work with. So are we basically doing ? We will have to -manufacture-, CREATE a new sample IN-BETWEEN the samples of the original recording. I.e. 48kHz has one sample every 0.000021 seconds. 96kHz has one sample every 0.0000105 seconds (approx). So at each 0.000021 seconds, we use the data from the original recording. But at those 0.0000105seconds points, we don't have any data from the original recording so we need to create one ourselves, literally out of thin air.

In the computer world, there are several ways to do this. The most common way is what you guys would have read in reviews, the so-called 'interpolation'. Strictly speaking all means of creating those extra data are interpolation. However, the gist is that there are many different ways to do interpolation. The most basic way is the so-called 'straight-line' interpolation. We need a new value in between two existing value. So we draw a straight line from one value to the other and we take the value EXACTLY in the middle of the line. That is the new value. This is basically what many basic implementation uses. The computing power to do interpolation is very demanding and as we go higher and higher speeds, the computing power increases exponentially. So because music is real-time, it means we NEED those new data within a strict time limit. So the computing power needs to be high enough to deliver. As we go higher in sampling rate, e.g. 192kHz, then we find that we need to create 3 extra samples in between two existing samples. So again the straight line interpolation is most used.

Now, for those people interested in the share market, this will strike very close to your heart. If you take any sine wave curve and you try to do straight line interpolation like the above, you will get a jagged curve. Straight line interpolation takes into account only the immediate adjacent two values. What we need to do is to do what is called 'curve fitting'. What this does is that it looks BEYOND the immediately two adjacent values, to values beside them. For e.g. beside the immediate two value, we also take one value next to each of them. So there's four values now. And we do 'curve fitting' over those four values. When we do this, we very likely end up with values that are different from what we get from straight line interpolation. But these values will give us a more smooth flowing, more realistic looking curve.

The bad news is curve fitting increases computing power demands tremendously. One way to mitigate this is to do the up-sampling 'off-line'. Off-line means not dynamic and that means not constrainted by time. This means we DO NOT have to process 192,000 samples per second, we can take longer time. We do the up-sampling and we store the result in a file, like a flac file for e.g. Then we upload the file to the music server software. This way, we can afford to do the upsampling with a lower computer power but it takes more time lar, e.g. hours instead of minutes.

This then is the basics of upsampling.

WongKN,
Does that mean that most CD-Player/DAC on-the-fly upsampling uses "straight line" interpolation ?

Are there any that uses very powerful DSPs and can do 'curve fitting' on-the-fly ? Are they better than software based offline upsampling or as good ? I suppose it depends on the implementation too, as even the software solution varies in results

Ah, this part is going to be a bit confusing/contentous.

Basically my understanding of the basic over-sampling techniques used in the standard CD players and DACs are that they are not upsampling. Let me explain.

In over-sampling, what we basically do is when we want to go to 96kHz, that represents 2X over-sampling. As we now need an extra data for each data value of the original recording, we go the simple route of simply reading a data value twice. I.e. for each data from the CD, we simply read it two time. And space the 2nd value halfway to the next. So to get the 'value in between', it is simply a duplicate of the previous value. For 4X over-sampling, we read a data value 4X. That represents 192kHz. Its the same for higher and higher over-sampling schemes. Again, note that I stated 'represents', because it is NOT equal to 192kHz. We don't process in 192kHz mode, we simply work with a volume of data equal to 192kHz for 4X oversampling.

The basic advantage, again from my understanding, is that our current DSP technology creates a lot of artifacts when we do DAC. It seems that for every frequency in the original, an alias frequency is produced somewhere above the 20kHz frequency zone. So for standard 44.1kHz DAC, we get the recreation of the original signal from the 44.1kHz sample -and- we get some extra, unwanted signals above 20kHz (sound frequency). This creates unwanted interaction with the original wanted signal. So DACs always incorporates a filter after the DAC stage. This filter is to remove those extra unwanted signals. And because they appear just above 20kHz and our original signal can go to 20kHz, we need a very sharp filter that allows everything up to 20kHz and then sharply cuts off everything from 20.1kHz onwards. This is the 'brickwall' filter we read about often. The term brickwall is to signify the very sharp cut-off, since a brickwall is tall and narrow.

Why can we work with identical extra samples in over-sampling modes ? Again it falls back to what I have been harping on frequency, our understanding of the human body. Again remember that our science tells us that we are supposed to hear only up to 20kHz and nothing beyond that. To be honest, I challenge everyone here to subject themselves to a pure tone test, using a test CD or LP and chances are everyone will find that, like me, we have a hearing acquity of very much lower than 20kHz. You will very likely be extremely dissapointed of the highest frequency you can hear.

And so as the theory goes, so what if that extra data are identical. Even if they are different also can we hear them or not ? Because at 96kHz sampling rate, that extra data in between compared to 48kHz sampling rate represents a 48kHz sound. And if we have problem hearing 20kHz, then scientists says we should not bother harping about sounds in the 48kHz zone.

BUT by processing the signal in 96kHz or 192kHz (for 2X and 4X oversampling respectively) mode, those extra unwanted signals above are now produced in the 48kHz+ or even 96kHz+ sound frequency zone. So when we want to filter them out, the filter needs to work at 48 or 96kHz PLUS we DON'T need a brickwall type filter. So what if the filter also removes frequencies below 48kHz. Even if it removes some frequencies down to 30kHz, in the first place, our science tells us we won't hear the difference.

This then is my understanding of the whole technology. Now, the contentous part I mention is that a number of reviews of CD players etc that implements over-sampling describes the process of over-sampling AS IF it is straight line interpolation from my earlier post. So now you have to decide if you think I am right or those reviews are right.

The best way to understand the different ways of doing 'curve fitting' is to open up an EXCEL spreadsheet. All of you would have one. There is a function called FORECAST. A much easier function to understand is the TRENDLINE in the graphical function. You try to create a graph like bar chart of line chart from a colunn of data. Then select the data source, right click and select 'Trendline'. The pop dialog gives you several options to create that trendline. What 'trendine' does is to fit a smooth curve (curve fitting) over the data in the column you select. And the pop-up dialog box allows you at least 8 different methods of fitting that trendline.

In up-sampling, the algorithm used are a lot more complicated and so they are extremely computing intensive. Usually if a software based offline up-sampling method gives the best results because no matter what computing power we have, if we constraint the processing time to 3 minutes (the typical duration of a song) and we do not constraint the duration and lets it run for hours, then definitely we get a more accurate result if we let it run for hours. Another consideration applies from the angle of efficiency. If we do up-sampling 'on the go', we basically do it everytime we play the song. That is a waste because we actually just need to do up-sampling once because if we do it properly the extra data never changes.

Again, you should go read up other sources, get other opinions etc as I have not really studied DAC and DSP in deep detail, just looking at it from the I/T perspective. The basic principles never change but the implementation details do varies.

WongKN,
A concise and precise explanation! For those who needs further explanation, you can get it from Weiss DAC202 Owner Manual, here.

DrWho,
Is the toslink connection any good on the DAC202 compared to rca spdif ?

I have not tried serious listening with Toslink in DAC202. I have connected it to my AV system and the sound reproduction is better than the Sunfire TGP V Processor but only in stereo. Some high multi strands glass Toslink cables are better than SPDIF coax.

Is there an audible difference between 24/192 and 16/44?
After trying the same piano track in 24/96 and 16/44 kHz from lessloss as recomended by carz, my answer is YES. Thanks carz!
The 24/96 has better dynamics and extension. It sounded more refine as well.
This was my first time I have compared the same track but in different resolution.
Thank you all for your contributions. Time to close this thread

Hi-Fi 4 Sale wrote:This topic is pretty interesting...should we be planning a H4S gathering to try out some configurations?

OK, before everyone runs off, the admin has proposed the above in an earlier post. So I think he would hope that some of us will give our feedback. Anyone ?

DrWho,
Did you try on your speakers/system or headphones ?
What speaker and amps are you using, may i ask ?

I suggest you try the drum tracks too
http://www.lessloss.com/drums-drums-drums-p-203.html

WongKN wrote:

Hi-Fi 4 Sale wrote:This topic is pretty interesting...should we be planning a H4S gathering to try out some configurations?

OK, before everyone runs off, the admin has proposed the above in an earlier post. So I think he would hope that some of us will give our feedback. Anyone ?

I am game since I am responsible for stirring things up;)

carz wrote:DrWho,
Did you try on your speakers/system or headphones ?
What speaker and amps are you using, may i ask ?

I suggest you try the drum tracks too
http://www.lessloss.com/drums-drums-drums-p-203.html

I don't have a headphone. I have PM you regarding my system.

while we are at it DrWho, why don't we try a little experiment.

Take the lessloss 16/44.1 file and use a good upsampler software to upsample it to 24/96 and 24/192.

Then compare it with the original lessloss 24/96 file and see which is better.

Also compare the 24/192 with the original lessloss 24/96 and see if it is better

My opinion is that the lessloss is very well recorded and very clean, with a very low noise floor. So i think it is a very good recording to be used for testing, please correct me if i am wrong since i have limited experience. (What is your opinion on this DrWho ?)

If the software upsampled file can equal or almost equal the original 24/96, then all we need to do is upsample our favourite album instead of hunting for hirez version.

carz,
What you are suggesting is like taking a mp3 file and converting it to a wav file and hope it would sound as good as an original wav file. We just cannot create additional info from thin air. Unless you know something that I don't, the upsampled file should not sound better,in fact it shoud sound worse, than the original file. This is audio alchemy at work

DrWho,
I think you are right that an upsampled file cannot be better than a native file of the same resolution.

But a 16/44.1 file upsampled to higher resolution will most likely be better if done with a good software. (if you have a DVD player and switch on upsampling, you will notice how the picture looks much better and less jaggered)

How much better and how close to a original native file of the higher resolution, is something which is interesting for us to find out.

As WongKN has explained you can fill up/predict the missing info by interpolation and curve smoothing algorithm . Yes it is possible, for instance, to fill up the gaps between a sine wave to make it look less jaggered. Also playing a 24/96 file allow the use of a gentler filter which help alleviate the problem caused by the brickwall filter used by 16/44.1 file that cause the music to sound a little peaky.

MP3 is different ballgame altogether as it is a lossy compression. If you want to upsample, then it is best to use the best native lossless file you can get.

I suggest comparing the following:-
16/44.1 file upsampled to 24/96 vs native 16/44.1
16/44.1 file upsampled to 24/96 vs native 24/96
16/44.1 file upsampled to 24/192 vs native 16/44.1
16/44.1 file upsampled to 24/192 vs native 24/96
24/96 file upsampled to 24/192 vs native 24/96

If the above can be done, then if our admin's suggestion can be brought to fruition, it would be a very interesting meeting for anyone who is interested to subject themselves to a blind test - can you differentiate between the different versions ?

WongKN wrote:If the above can be done, then if our admin's suggestion can be brought to fruition, it would be a very interesting meeting for anyone who is interested to subject themselves to a blind test - can you differentiate between the different versions ?

For the can even try different resampler softwares like SoX and the Secret Rabbit Code(SRC) if using cPlay as the playback software.

Carz,
Thank you for those recommendations. Should there be a shoot out, you should bring along all those tracks mentioned. On top of that we should try a native 16/44 track burnt to a CDR and compare the performance of a transporter against the CAS.

DrWho: I don't have any good upsampler software.....anyone here can help? Need to have a very good upsampler because we are trying to test if an upsample track can be nearly as good as a native track of the same resolution. So don't want to be constraint by the limitations of the upsampling software.

tycham: The cplay SRC and Sox you suggested, are these done on-the-fly or offline ? Offline is better i read, and for good reasons too. How good are the SRC and Sox ?

I have quite a few Upsampled albums done by Weiss Saracon Software (supposed to be the better software used by professionals in the industry)

24/192
24/96

No point 16/44.1 coz there is no comparision to hirez.

carz wrote:
tycham: The cplay SRC and Sox you suggested, are these done on-the-fly or offline ? Offline is better i read, and for good reasons too. How good are the SRC and Sox ?

These are done on the fly. I concurred that offline should be better as it would be less taxing to the processor and therefore induce less electrical activities by the processor.

Here are screen shots comparison of a 1kHz Tone(-0.1dbfs & -60dbfs) of the SRC, SoX, and Weiss Saracon 1.6.







Side band non-linear distortion is almost non-existent for the Saracon 1.6.

VS126 wrote:I have quite a few Upsampled albums done by Weiss Saracon Software (supposed to be the better software used by professionals in the industry)

24/192
24/96

No point 16/44.1 coz there is no comparision to hirez.

Weiss Saracon is one of the best around.
how good were the 16/44.1 files upsampled to 24/96 and 24/192 ?

Visibly better but not as good as native.

Better air, space, more pronounce mids and bass goes down deeper.

Mostly you will notice better separation between instruments. But it all depends on yr equipments.

Just like if you have the best gasoline on a beat up kancil.

So to some people, they really cannot tell the difference btw HIrez and red book coz their equipment does not do them justice.

VS126 wrote:Visibly better but not as good as native.
Better air, space, more pronounce mids and bass goes down deeper.
Mostly you will notice better separation between instruments.

Is it close to native file or is there still a marked difference ? What are the sonic "advantage" of the native file ?

VS126 wrote:But it all depends on yr equipments.
Just like if you have the best gasoline on a beat up kancil.

So to some people, they really cannot tell the difference btw HIrez and red book coz their equipment does not do them justice.

Agreed ! System need to be high resolution enough, or else use a decent pair of headphones

Native 24/96...192 file played on capable DAC of same resolution will be totally different experience. You will notice it immediately. It is like a thick curtain finally lifted off yr CD based system. Staging and refinement. You can actually turn yr volume up very high without distortion. Do that to a CD bass system and you will get breaking sound at that volume.

Everything is important...down to yr laptop, using SSD to better media player and all stopping services on yr laptop.

USB cable also very important....current favourite now is Cardas Clear USB.
cheers