On following reference designs, that is what I have had to literally fire hardware engineers over that worked in my group! This is what I meant by digital engineers designing analog circuits. They take the cookbook circuit and PCB layout from the chip company and run with them, not understanding important nuances that they did not know to follow or followed for the wrong reason (just moving a trace may make a big difference in the achieved performance in analog designs). The result was products than then failed to meet the same specifications as the chip company provided in their testing and in our case, failed customer qualification tests requiring expensive redesigns. When I confront such engineers with these major problems, their answer is that: they followed the reference design so it is the chip/sample design that is at fault. So I bring the app (application) engineer who put those reference designs together and the first thing out of their mouth is that these are examples and you have to know what you are doing and not follow them blindly. They are put together for people who are not worried about the best performance and lack the competence to design such circuits. Providing such designs enlargers the market for their chip and provide practical documentation on some of the design aspects. They are not provided as a guaranteed way to get good performance.
For great performance, you have to understand what you are doing rather than following someone else's instructions. A DAC that sits inside a noisy AVR with a ton of other circuits around it is not what the DAC chip app engineer is giving you. That is a more complicated problem that is yours to solve. You can use their reference as a guide but you better know more than that app engineer to build excellence. You can look at the example measurements I gave you for this situation. Notice how the DAC had tons more distortion when driven by HDMI vs S/PDIF in that AVR. The DAC is a constant there. The upstream circuits are not, nor where they part of any reference design from the DAC company. When HDMI is running, a ton of additional circuits start to "whale" because HDMI requires video to work. We can see then so much more distortion is induced in the DAC clock circuit that was not there in S/PDIF version. The second designer clearly knew how to deal with this building a clock circuit that was almost completely insensitive to this factor. BTW, the same measurement exist for other AVRs with much higher distortion in their HDMI vs S/PDIF.
I had a discussion with a company in this forum that kept quoting the specs for the DAC chip as opposed to their processor (multi thousand dollar unit). After much pushing, they provide the measurements. This is what (TI) the chip company provided for the DAC chip:
This what the company measured for that processor with that chip in it:
The peak distortion spikes are at -90 db now vs -110 (as I indicated by the red line in the first graph). So we lost 20 db of performance. And this is from a company that said they hired the best DAC designers and provide this unit to professionals (it is used as a digital cinema audio processor).
Reference designs are most useful for digital systems where such variations do not exist. A DAC and associated clock recovery circuits is not one. Again I say this in the context of getting excellent performance. If the standard is that it spits out audio and degrades the performance of DAC chip by 10 to 30 db, sure, go right ahead. That is not how I aspire to buy equipment that way or managed hardware projects that do the same.
As I said, it is not rocket science but you have to know what they are doing. Read the EDN article I post and it goes over all the things you need to know which sadly, typical engineers don't. This day and age, the name of the game for hardware designers is digital. Good analog designers are much harder to find and fewer still are going to hang around in mass market audio companies working on designs that have to be as cheap as possible to serve a money-losing or barely breaking even industry. Open up a $300 AVR and you tell me how anyone can remotely make money putting so much electronics in there, assemble, test and ship that to you. The two things that matter more than anything else there is how many logos are on the box and how cheaply you managed to produce it!
Fortunately digital systems even when poorly implemented have very good performance. And as listeners, we are not that sensitive to non-linear distortions or else, compressed music would have never taken off. So mistakes and all, you still get sound that has full frequency response and enough dynamics (linear aspects we easily perceive). It is therefore not surprising that people tend to think these factors don't exist. There are many instances where I don't care either. For a system in a living room I don't care either. I put an AVR there and call it the day. Just don't say this stuff is easy and/or you can follow cookbook reference design and we will be friends