Upsampling/MultiDAC technology works!

Yup, but I just don't like the use of the word 'guessing' I suppose, it sort of implies that it has been done randomly. It is very educated guessing using all the available information, and surely 'joining the dots' using some form of curve is better than using straight lines?


The Moog

 

Thats an interesting distinction...I must admit to only having tried or used upsampling and interpolating systems in this context(either dCS during a few demos,and Perpetual Technologies at home),but it would be very interesting to compare that type of system with the oversampling/multibit system you describe....for me,anyway,upsampling/interpolation does very nicely,and sounds vastly better(most of the time)than straight 16bit CD,with my system and what I listen to....but.at the end of the day,that's only my preference,and thus one of the reasons why we have so many different types of replay systems around,and people who like/dislike each.....variety being the spice of life etc!

 

Could someone please explain the fundamental differences (or post a link) to an article doing the same about up/over sampling? Is 2x oversampling (theoretically) the same as 88.2kHz upsampling? Cheers,


The Moog

 

the only difference that i can discern between upsampling and oversampling is that oversampling is multiplication of the sampling frequency by an integer amount (i.e. x2) and upsampling by an non integer amount (i.e. x3.14159)
take a look here for more guff...
http://www.audioholics.com/techtips/specsformats/upsamplingvsoversampling1.php

cheers


julian

 

i'd like to add a caveat to all this though. ime the analogue output section and it's power supplies make just as much if not more of a difference to the final sound (to my ears, imho, etc.). this is another reason i chose the dac i did as it has 3 big psu's with lots of regulation and potted, carefully designed output stages.
cheers


julian

 

Damn you Julian, you are so f**kin' right!

Just add bitstream to it... Do you know that bitstream was originally created for car stereo?!

 

bitsteam is another cost over quality thing afaik....
the way that a multi bit dac works is by the 16 bit data word being used to switch a constant voltage through a bunch of precision resistors whose outputs are summed to provide a varying level which is the raw analogue signal.
getting these resistors to provide the correct values (down to 2.2 / 2^16 or 0.000033569 of a volt if your maximum output level is 2.2v) is expensive and leads to various qualities of the same dac e.g. the bb1801 (k)??? and famous tdc 1541 double crowns etc. which are horribly expensive but have spot on perfect resistor values. (if i understand things correctly)
a one bit dac is a lot cheaper to produce as you don;t have to faff about with laser trimming 16+ resistors etc. it can just be stamped out on an i/c. once again quality sacrificed on the altar of profit.
cheers


julian

 

Julian,

In practice you have current(not voltage)-source DACs whose output is then connected to a current-to-voltage (I/V) converter.

Damn you Julian... Did you know that switching amplifiers (aka digital amps) have grown for car audio applications?!

 

bd,
but i like the sound of (the tripath class t) digital amps when i've heard them. possibly something to do with them having a higher a/d sampling rate at the start than 44.1?
cheers


julian

 

...I see the debate goes on - I still cannot say exactly WHY the 192 Multi DAC is working for me... there is also the issue of improved jitter control that I expect the new DAC is providing me - that is no doubt contributing something too. Just seems the result is a little bit more than the sum of all its parts...

 

Julian,

For red book CD resolution you would need a bitstream ADC with sampling rate equivalent to 65536 (2^16) pulses for every 44.1kHz sample. Digital amps don't have sampling rates of 2.89 GHz (65536 x 44100). That way, you are loosing precious resolution...

It's like upsampling but, this time, to a much higher sampling rate, say 500 - 1000kHz... the (analog converted) 16bit/44.1kHz digital data from the CD player is then re-converted in the amp by interpolating new samples...

Would you like the sound of a 2.89 GHz switching amp better?

 

bd,
the digital amps i've heard have all had analogue inputs (bel canto's mainly) so they sample an analogue signal at a much higher frequency than 44.1 khz.
i suppose a true digital (digital inputs) designed to work at a multiple of 44.1 khz would be great - but i'm now just guessing.
cheers


julian

 

Julian,

You didn't understand... the analogue signal that is sampled by the amp is the result of 16bit/44.1kHz signal. So, you only can have a much higher frequency than 44.1kHz if you upsample from it... In this case, that is done by the amp.

In practice, you may end up with two upsampling stages instead of just one if you use an upsampling CD player... if you don't like the first one (CD player) why add a second upsampling stage (digital amp)?

 

surely it would depend on the dac. even with a modest x2 interpolating (as opposed to zero stuffing) oversampling you end up with a max frequency of 44.1khz (as opposed to 22.05khz) with only every other sample being guesstimated. at this point the signal (being analogue) can be sampled at whatever frequency you want by the amp, dither added and all the other tricks only usually applicable at the recording stage to help out.
why would an amp upsample anyway surely they'd just sample the incomming analogue signal at whatever rate it uses natively.
cheers


julian

 

The analog signal cannot exceed the digital resolution (16bit/44.1kHz or 2^16 pulses for every 44.1kHz sample) that originated it...

If you take the digital data directly out of the transport and then by-pass the ADC stage of the amp feeding it to its DAC only, you'll end up with an UPSAMPLED signal at amp's switching frequency...

Now, just imagine you have a perfect DAC within your CD palyer and an equally perfect ADC within your digital amp... feeding the amp with the CDP's analog output surely would give you exactly the same UPSAMPLED signal as before...

Moreover, any superfluous DA or AD conversion can only further degrade the sound...

 

sorry bd you've lost me....
i'm not sure what you're problem is here.
you have a source which creates an analogue signal. this is then converted into the digital domain at a very high sample rate and this is digitally amplified and then re-converted to a louder analogue signal.

where is the upsampling in the amplifier section?

if you are talking about tact style true digital amps (which i'm not and you seem to be) then i can;t comment on how they sound as i've not heard one. this is why i qualified my above statements by stressing the tripath / class - t style of amps.

i like the sound of the bel canto and es labs amps i've heard however my favourite amps by far are the laser biased edge amps.
cheers


julian

 

As demonstrated, of course, by no less a luminary than Ivor Tiefenbrun. Wasn't it?

 

Julian,

No problem... it's really my fault because I cannot explain myself better. I will try one more time...

Like you say, I have a source which creates an analogue sound. This is then converted into the digital domain at a very high sample rate... ok, so far so good.

Now, why do I need to create an analogue signal prior to the amp?

To have a very high sample rate re-conversion into the digital domain?

Certainly not!

What do I gain then?

Nothing!

What do I loose?

(69) If I had perfect DA and then AD conversion... I'd loose nothing. At some point I'd have exactly the same digital data stream as that from the transport output.

So, where is the upsampling issue?

Taking an analogue signal which was converted from 16bit/44kHz data and then convert it into the digital domain again at a very high sample rate IS EXACTLY THE SAME THING, PROVIDING I HAVE A PERFECT (69) SITUATION, AS TAKING THE NATIVE 16bit/44kHz DATA STREAM AND THEN UPSAMPLING IT, BY INTERPOLATING NEW CALCULATED DATA (equivalent to new samples from the first case AD conversion), TO A VERY HIGH SAMPLE RATE BITSTREAM WHICH WILL DETERMINE THE SWITCHING FREQUENCY OF THE MOSFETs AT THE AMP OUTPUT.

The MOSFETs at the output will be switching at exactly the same very high sampling frequency in one case or the other... and in the same sequence.

 

hmmm,
i don;t thing sampling an analogue signal at a high sample rate IS the same thing as upsampling. sure if you are using a cd as your source then you are going to have a lot of wasted bandwidth above 22.1khz but with vinyl or dvd-a etc. then a sample rate of 400+ khz is going to be fine. i can't see the need for any kind of upsampling in the amp as upsampling (or sample rate conversion) is only necessary when you are trying to make something fit that wasn;t intended (i.e. changing a sample rate of 44.1khz so that it can be decoded by a 92khz dac). similarly if your digital amp works by manipulating a bitstream rather than words it would be sampled as a bitstream again negating the need for conversion of any sort.
cheers


julian

 

Ju,
1.2/5ghz sample freq, high and wide shite removal, before returning to the real world, something which BD has a problem living in