Upsampling/MultiDAC technology works!

Sorry I have been away from this forum for a while... but just to let you folks know that despite my initial reservations on upsampling, my Arcam CD192 upgrade (applied to a CD73) has made a considerable improvement to my listening experience... the improvement has been profound. I agree in the principle that upsampling cannot extract more data from the disc than is already present but I feel the big difference is the use of kinder filters as a result of this - there is definitely an improvement in resolution at the top and bottom end - this system can now generate truly subterranean levels of bass and exposes nuances that the previous DAC arrangement cancelled out or ignored. One of the best cd's I own for showing off the extremes in bass resolution is the Ray of Light album (Madonna) - the experince was wonderful before but I can honestly say the experince is now awesome. I eat my cynical hat.(Now where is that TCI power cable...)

 

I can only agree

 

Interesting set-up you have. I stuck with the Arcam theme somewhat: CD192 running through bi-amps A85/P90 into Monitor Gold Ref 20's... I can honestly say I have reached the limit of the finance I wish to throw at a system but it does work well. Off on a tangent - earlier recordings transferred to cd still frustrate me as they are so strangled in terms of bandwidth (I suppose that is down to the mastering - even for some recordings made well into the 80's to late 80's it seems as if some studios later set a level "limiter" on every sound that hit the cd master - definite example of where LP's win). The other annoying thing about cd is that the output volume level seems to be calculated ahead - based I suppose on the highest peaks it finds on a track/album (maybe in the redbook spec there is a signature encoding hidden on recordings that set this level... I do not know as I am no teckkie) - the result of this is that you get acustomed to listening to one album at a certain volume setting - change to another cd and sometimes it is as if someone has leaned on the volume control and turned the "perceived" volume down by as much as 50% - good example: Lou Reed - New York album (after perhaps listening to modern Peter Gabriel recording). Another example of terrible transfers is the Billy Joel Greatest Hits double cd from around the late 80's - o.k. I am not a huge Joel fan but if you bought this c.d. thinking you were going to hear something that sounded as if it had any life in it - forget it! What could they do wrong with the recording of just a male voice and a piano you may ask? Well Columbia (I think it was Columbia...) - really screwed it up - it is the flattest, dullest, compressed piece of crap you are ever likely to insert into your CDP. Do you have similar issues? Otherwise everything is cool.

 

...meant to add that it just means you can have the most expensive system available to mankind yet we are still at the mercy of the record companies. One good reason for setting a realistic budget for a system I suppose.

 

to be honest i've not heard many upsampling cd players / dacs that i've liked. the only one i can think of off the top of my head was the c.e.c dx 71. generally i prefer oversampling which means you get all the benefits of the gentler filters but without the guesswork involved in sample rate conversion. but then i like multibit more than delta / sigma so i'm a luddite - pass me clogs someone.
cheers


julian

 

Julian,

When interpolating new points between the different original samples in order to obtain an artificially higher sampling rate you are always guessing... regardless of name assigned to it.

How the interpolation is made depends uniquely on the arithmetic calculations being made, which in turn determines how "wild" the guess will be...

 

bd,
you're right, even with oversampling (if you interpolate rather than just replicate the samples) you are still guessing however the original sample data is still in there along with the guestimated data. with a transform, say from 44.1 to 96 khz, very little of the original sample data is used. maybe 1 in 96 as compared to 1 in 8 for full blown 8x interpolating oversampling? i can't be arsed looking it up. anyone?
cheers


julian

 

largest common denominator of
44100 and 96000

er...
44100 = 2*2*3*3*5*5*7*7
96000 = 2*2*2*2*2*2*2*2*3*5*5*5

2*2*3*5*5=300
zo in 1/300 seconds 44100 sample rate gives
44100/300=147 samples
and a 96000 sample rate has
96000/300 = 320 samples
only 1 in 320 samples are exactly the same as in the original 44.1 khz stream, whereas 2x oversampling has 1 in 2 samples. the rest is interpolation.

am i right?

 

seems to be right to me. of course we are talking about minute fractions of a second here but it's still sobering that only 1, 320th of your music is what is on the cd.
cheers


julian.

 

Maybe, but is the rest dithered around the correct/probable value? As you and I both hinted in the thread on CD copies, this is probably not a bad thing.

 

Just listen, then you decide, me I'm a 44.1khz man myself

 

graham,
dithering the original signal to bring out detail that you wouldn;t normaly capture at the expense of a slightly higher noise floor is pretty different to only having 1 in 320 of the original samples present. the ONLY reason i can see for upsampling to 96 or 192 khz is to enable the use of cheap a/v oriented dac's. even an 8x oversampling dac is the equivalent of a 352.8 khz sample rate and you are getting 1 in 8 samples that are accurate. this enables you to use a far gentler filter than even 96 or 192 which is the main plus point put forward bu advocates of upsampling.
at the end of the day though it's what it sounds like that matters and for my money i'll go with integer multiples of 44.1 and a multi bit design every time. and that's from listening to examples of both.
cheers


julian

 

Surely if you just like a different sound, why not just buy a different or better CD player?

 

Yes Julian, it's sobering indeed... just imagine how little of the original signal gets out of an amp that doesn't have a perfect null, or out of the speakers, for that matter...

That's why "hi-fi" is a joke without accuracy... without it, it's just a BIG WILD GUESS!

Wanna gamble? Put your money on your heart is...

RED or BLACK?!

Just make your bet... Will you regret?!

BLACK or RED?!

Maybe you'll go mad... HifiPlus you'll have!

RED and BLACK, BLACK and RED

Just another thread... and nothing is said!

Not RED nor BLACK!

Bye, bye pussycat... Zerogain is the bet!

You a 44.1kHz man?!?!

Nah... I don't think so.

At least, as long as you keep listening to bitstream UPSAMPLING digital amps...

 

Please excuse my lack of in-depth knowledge on this subject, but it isn't just random red-or-black guessing is it? You have a start point and end point, and you are interpolating a mid-point using some curve-fitting algorithm, which in my experience is usually a pretty good representation of what actually happens. Even if you are only actually using 1 in 320 EXACT samples from the data you are given, you are still using 100% of the data on the CD to obtain all of the other values.

I would be very interested to know how closely an upsampled segment of musical data from a CD matched a native higher sampled recording of it. I think would bet it would be pretty close.


The Moog

 

With oversampling you are...

To use is one thing... to use while preserving is another thing completely different...

 

i would imagine they would look pretty close if represented as a waveform however there would be deviations/ errors / distortion call it what you will. especially if the original waveform doesn't fit the interpolation scheme used.
my point is that if you use oversampling (increase of the sample rate by an integer value 2x 4x 8x etc..) rather than upsampling (increase of sample rate by a non integer rate 2.1768 (96khz) 4.3537 (192)) you can still interpolate the added samples and get the benefit of interpolated data but without the detriment of only having 1 in 320 original samples. as i said the only reason for upsampling is cost not quality (imho).
cheers


julian

 

If you listen to the music and not the numbers,or the potential errors etc,a combination of upsampling and interpolation can be very effective,and a very worthwhile addition to standard CD,if properly implemented.

 

Surely whatever approach is taken you always get to the stage of 'guessing' what happens between the samples, during the D to A conversion. I mean the final stage is effectively some sort of joining the dots exercise to generate the analogue output waveform; upsampling just gives you more dots to join. Presumably the argument is that whatever interpolation is done in the digital domain can be done 'better' than the final DAC stage?

 

i'd agree that it has improvements in certain areas over straight 44.1 non oversampling / upsampling, non interpolating dacs (although it's weaker in others imho). however to my ears an OVERsampled, interpolated dac at 8x is far better than a 96 or 192 interpolated upsampling. at 8x oversampling you have an effective sample rate of 352.8khz which pushes aliasing noise even further away from the audiable band (the whole point of up/over sampling) and allows an even gentler slope to the filter. yet with 1 in 8 samples being accurate rather than 1 in 320. it's a win / win imho whereas with upsampling at non integer values you only get a small advantage over 44.1 (relatively) with the filtering and loose the accuracy that is strived for in hi-fi.
as you say though it's your ears that matter and mine have repeatedly told me that oversampling + multibit will beat upsampling + single bit delta / sigma in most cases. if your ears tell you differently then fair enough each to their (subjective) own.
cheers


julian