New vs. old DACs - opinions?

The main thing that has inarguably improved over the past couple decades is the fabrication processes by which monolithic ICs are made - and since the overwhelming majority of audio DACs are based on monolithic products . . . the palete of resources available to the IC design engineer has changed considerably.

That being said . . . just as everything else in audio (and life in general), the success of a design is very closely connected to how the designers/engineers work to reach their goals with the resources they have available. And in the late-1980s/early-1990s, the goals of most IC designers were probably more similar those of audiophile products than they are now -- there used to be many $10-and-up "statement" mono DAC chips available, that cared nothing for power consumption or single-supply operation, and assumed that the circuit designer had no problem spending extra money on separate circuitry to do the processing/upsampling.

Nowadays, a $4 audio DAC is VERY expensive, they're all at least stereo (if not multichannel), and have some sort of oversampling or sample-rate-converter built in - making the circuit designer's job much easier, if they want "off the shelf" performance. Most of them are also much more flexible in terms of software control, and place low power consumption as a priority. Translation: these days, the perfectionist audio business has very little part in paying the bills at Texas Instruments et. al.

It's true, the chips HAVE gotten "better" . . . especially if you're paying attention to the same parameters that most IC designers have. But all circuit architectures have their stregnths and weaknesses - and they're all ultimately dependent on the designers and engineers of the final product whether the parts fulfill the promise of what the chips themselves can do.

Kirkus - I have more experience with components than audio but components such as op-amp for instance made long journey in last 20 years. Currently they cost more than before but if you calculate price in dollars per MHz of bandwidth or offset drift or noise then new amps are at least 10 times cheaper. There were only few (unfortunately) intended for high en audio like very old NE5532 or very new LM4562 - not enough market for them perhaps.

I agree with you on responsibilities of designer. Designer can screw-up design of power supply today as easily as 20 years ago. Certain components like clock oscillators with extremely small jitter might be very expensive but it's money well spent. Designers often cut corners in places where they shouldn't. It is very sad to see piece of great equipment with one or two inferior components that somebody else (modding people) replaces to dramatically improve performance.

there is a lot of talk about clocking and re-clocking. in a normal transport/computer to dac connection, there is no clocking or re-clocking. except if you own certain esoteric or dcs components. to do clocking, these devices would have to work on a peer to peer basis and if the dac receives invalid data, it forces the source to retransmit or reread the data (esoteric calls that flow control). esoteric and others sell external clocks which are very expensive and they only work on certain sources.
here is a quote from Thorsten Loesch:
"In reality, it is relatively easy (i.e. low cost) to make a clock crystal with high precision (<0.5ppm). However, making a clock crystal with low phase noise (jitter) is much more difficult (i.e. expensive). Perhaps that's why the aftermarket so-called 'super-clocks' often quote the former and omit this latter and more relevant specification. Just as importantly, the use of more than one clock (even a 'super clock') if it is not synchronized to generate a master clock signal, will mean that the issue of all clocks being out-of-sync with one another has still not been addressed. Again, this leaves the other major source of jitter unresolved."

I think the basic issue is a clock is needed to convert the bits to analog at the right time otherwise the result is jitter and this typically occurs at the other end of the wire from the DAC, back at the source which feeds the DAC. Both clock and connection then come into play in regards to converting the right bits at the right time. If the DAC resamples and reclocks, then the the clocking function and DAC are more tightly integrated within the DAC and there is less to go wrong.

Kijanki, I'm right there with you on opamps being so much improved . . . thinking back to the "bad old days" of slow all-NPN types like the 741 just about makes my stomach churn. And I'm pretty amazed about the quality of diverse types of devices - i.e. high-quality JFETS and complimentary bipolars sharing the same wafer? Something we wouldn't have thought possible a few decades ago.

But an opamp example that would fit the example I gave above (for DACs) are those modern CMOS devices with an onboard DAC for trimming? Yes, their offset specs are impressive, especially so for inexpensive devices . . . but they're irrevelant for perfectionist audio. Ditto for rail-to-rail input and output capability, ultra-low quescient current, and single-supply "precision" operation.

Fortunately, in the world of opamps, it does seem that the bouquet of ultra-high-quality products available is more extensive than ever . . . but the fact that the attributes of most concern to most IC manufacturers (or at least the ones they spill the most ink over) aren't ones that matter most to high-end audio . . . is worth taking notice of.