The worse crime however is for those deploying such uber-chips to 1) Use the chip manufacturers specs not actual measurements and 2) not measure the complete circuit which is often rather worse measuring - there is a manufacturer (popular round here) that does this - and achieves the feat of taking a 32 bit chip and creating an (approximately) 16 bit circuit - inspired engineering at it's best ! - granted analog stages incur more noise but even so losing 1/2 the bits is a tad careless
Oh, this is worse...this is commercial fraud, many companies did it when the AK4396 hit the market. All of a sudden, tons of 120dB SNR soundcards popped up, like the Audiotrak Prodigy HD2 that barely reached 100 dB in real world measurements. This said, it sounded better and far less shrill than the Essence STX...and that's what both many friends of mine and I personally witnessed on very different gear. Again, THD measurements will never be the warrant of anything as far as subjective human perceptions are concerned. And the HD2 has 2 discrete clocks for 44.1/48kHz multiples when the STX only has one 24.576Mhz(48kHz*512) crystal.
This said, it was merely happening in the $100 internal PC soundcards market...I wasn't aware that some self-proclaimed audiophile companies were still doing it these days?
I was talking about this Sabre powered unit: http://www.calyxaudio.com.au/index.php/calyx-dac
Not a single word on its output stage, simply innocent killer specs:
THD+N: 0.0005% @ 1KHz, 0dBFS
SNR: 125 dB, A-weighted, 2.2 Vrms
Channel Separation: 140 dB @ 1KHz, 130 dB @ 20KHz
Or maybe they're being vague on purpose considering that they gave the "DAC Specifications"? "DAC device" or "DAC chip"? haha.
If a company is not clear about something, it usually means that they have something to hide. All I can see is a shiny box w/ a Sabre ASIC that does it all at once: S/PDIF input, DAC, I/V opamp based output stage. This is merely a deluxe edition of the Realtek ASIC's. The markup on those chips must be stellar, I can understand why ESS ask to sign NDA's in order to access their datasheets material. Golden goose at work.
The Calyx DAC is USB powered...but sure, it's "cutting edge" blablabla, thanks to its very noisy 5V PSU probably =)
Quote:
Cheap-ish audio opamps are pretty much perfect for many applications, such as in DACs and headphone amps.
"Perfect"? Please define, I didn't realize that perfection existed. Do you realize that some ppl spent their lives looking for it? You might actually be onto something, so don't give up just yet
I'll bet you'll come back with a snip about how he wouldn't risk his reputation on such an assertion - I'll retort in advance with another reminder that he is an electrical engineer (and a very, very good one, yes), not a pscho-acoustics expert. His claims as to subjective sound aren't directly related to his field of expertise and thus do not threaten his reputation as such - for example, I'm sure he could design a massive negative feedback design amplifier more than competently if he so desired.
My main reply will be that I didn't spend hours looking for "no feedback" audio designers...I'm sure there are many more, and the most notorious on head-fi has to be Kingwa. He has taken the DAC market by storm with his DAC-19, and he's been so successful that he decided to call his all newest units "NFB" just to make things very clear in ppl's minds. I read ppl raving about his NFB-2 all the time, and not just fanboys. There are many discrete believers that also got into NFB, and once you've heard the NFB gear the usual "by the books" PCM179x datasheet topology blabla goes up in smoke SQ-wise.
Personally, I have no problem with either design, as long as it is properly implemented. And that's the key. But obviously that doesn't matter because if something anecdotally sounds bad then it is "evil", right?
Well, the K.I.S.S. PCM179x implementation cannot possibly sound good IMO..even worse when it's a voltage output DAC chip with a built-in I/V stage: you can slap any top of the range opamp behind a PCM1793, it will never allow you to extract all the details of the source. And using 4x single opamps for I/V and two more to buffer a PCM1794 won't do miracles either. OTOH, I have to admit that Asus decided to use a "discreter than discrete'" topology in their new "Asus One" USB DAC, and I have to say that I've been impressed:
They used a toroidal transformer(that provides galvanic isolation if I got it right), a very smart S/PDIF receiver that resyncs silently when switching sample rates, a Sharc DSP that reclocks all inputs and a whole bunch of 22 single opamps in the signal path using a fully dual-mono design..yay! And Asus aren't exactly newbies when it comes to PCB design either. This DAC sounds very promising, too bad they had to reach for as low of a street price as possible so they skimped on the opamps..they're 5532's and 4562's =/
But I'll soon be hearing it w/ an army of AD797 opamps, and I think it will deliver. Yes, OK, when using good chips within a never seen before dual mono topology, opamps might be a good option...but it doesn't come cheap either, because good opamps cost around 10 bucks a pop.
So which opamps would you suggest I get that will make a very noticeable difference?
Personally, I would use either 6*OPA1641 or 4*OPA1641 + 2*OPA602BP. OPA1641 is very cheap and is one of the latest designs from TI(circa 2010), they advise to use it for high-end bluray players and I can see how it would be greatly successful in this role. This is not a "rock your world" discrete design, but its price/SQ ratio is stellar. The NJM1114D's that can be found on the STX are merely NE5532 clones with slightly better specs, the 5532 is like +30 yo AFAIK. And dual opamps don't hold a candle to single opamps either, even TI acknowledged the problem: http://www.ti.com/product/opa1612
"The dual version features completely independent circuitry for lowest crosstalk and freedom from interactions between channels"
Cheap can sound good, OPA1641 is a perfect example of this...but again, don't expect a $1.15 IC to sound as good as a top of the range discrete design...not gonna happen.