CoolAudio V1402 (ADAT Decoder)
Ah let's see.. it is as I feared. You have a shitty datasheet.
Everything up to section 3 seems okay, but then this:
Absolute Maximum Ratings
Supply Voltage Vcc max 4.0V
There's no Vcc pin on the stupid thing! What idiot filled out this table!
DC Parameter
Supply voltage Vdd
4.5-5.5V
Assuming Vcc == Vdd, they're overvolting the thing by their own admission!
I also don't get why Temperature should warrant being a DC Parameter. They already specified Tamb = 25C as a measurement condition. And 70C max is in the ratings. (BTW, 70C is kind of low, and you may be surprised to reach it under otherwise fairly ordinary conditions, depending on just what all goes into your project. If it starts misbehaving on a hot day, or something, there's not much you can do about it, it's simply not rated for that kind of use.)
Other basic checks:
V_IH, V_IL are kind of loose. 0.3 and 0.7 of VDD are more typical of CMOS devices. This implies they used a shitty CMOS process that's cheaper and less reliable (but hey, it only goes up to 70C, and it's presumably ancient 5V capable, so what do you expect?). These still shouldn't be hard to meet though.
C_IN says 5pF typ, 1pF max. Which is it?
V_OH, V_OL are fairly reasonable, but it's not stated what current they are measured at. Nor what voltage I_OH, I_OL are at. In general, the two pairs of specs are not measured at each others' level.
Useful info: Supply current Idd = 5-8mA, I guess depending on mode. This is probably going to be accompanied by sharp spikes of some amount greater than this, but after the local bypass cap(s), it's probably going to look pretty flat.
The AC Parameters are pretty brief. I would think to expect some basic stats about the protocol it decodes (how long does it take to process a packet, how good is the PLL, how stable is the resulting clock...). They could obviously go to considerable length documenting just what the heck it does, but sometimes datasheets are intentionally terse like that ("oh sorry, you have to buy the $10k license and NDA to get the full datasheet, this is just a teaser" -- or the Chinese ones just don't give you anything and you have to figure it out yourself!).
Conclusion: crappy chip, but if it does what you want, it's probably OK. Special purpose brainy things often do that, sadly.
ATTINY (Microcontroller for bitbanging the settings into the DAC)
Good, or good enough. Won't add much to digital noise. Could probably throw some analog functions in as well if you wanted (ADC temp reading..?! Idunno).
analog supply:
PCM1680 (DAC)
Oooh, 100mA, she's a biggun!
Beware the logic pins are TTL compatible. TTL inputs can be connected to TTL or CMOS; TTL outputs cannot be connected to normal CMOS. It seems this is only a concern for ZERO1 and ZERO2, so if you don't need them, it doesn't matter.
As a DAC, the performance specs are pretty appalling. They don't tell you this, of course: 100dB of dynamic range or SNR or whatever, and 0.002% THD+N, sounds great (and is more than sufficient for audio purposes). But 105dB is only 17.4 bits. 0.002% THD is 15.6 bits.
So be fully aware that "24 bits" is only marketing wank. You'll never be able to measure, let alone hear, the difference of those additional 8 bits. They do not give any traditional DAC measures, like INL, DNL, SINAD, etc.
A proper, professional duty, precision 24 bit DAC can get down to a few LSBs, but I don't think any are out there that claim a few LSBs at anything more than a few Hz bandwidth (i.e., pretty much DC). The SNR of space itself becomes dubious at that kind of precision (quantum fluctuations and all)!
Because ears have about 100dB of dynamic range (below the pain threshold), and far less under average ambient conditions, it should be no surprise that 16 bit audio has been as pervasive as it is, since its introduction. Yes, you can play some gimmicks with extra bits (mostly numerical, for intermediate processing), but ultimately there's very little gained.
OPA337 (Opamp)
This looks good, but mind the downsides of CMOS amps: they're noisy (26nV/rtHz). I do like the PSRR. Note that the input common mode range goes not include the positive supply.
TL974 (Opamp)
Hmm, this thing is really weird. 1.4mA source but 80mA sink? What the hell?
The equivalent circuit shows PNP inputs, but these will not function up to +V as claimed. Their circuit is patently wrong. Very likely, a complementary design is used, but they should've shown that; they should also have a Vos vs. Vicm graph, which usually gives some evidence how the NPN and PNP combo works together. (Often, non-complementary parts will have this graph too, which usually suffices to show how the device fails when you exceed Vicm -- useful info when designing a circuit.)
The GBW varies massively with supply voltage. It's advertised for its wide supply range, but if you actually used it over that range, you'd probably get an oscillator at some point! It's not hard to keep GBW stable. This thing is poorly designed.
The only thing that is reasonable is the noise level, which is low as op-amps go. But it's not painfully so; Fig. 6 shows it rising below 1kHz (1/f flicker noise), and more steeply than usual. For comparison, the TL071 claims:
http://www.ti.com/lit/ds/symlink/tl074.pdfabout 45 nV/rtHz at 10Hz, while the TL971 claims ~55. And that's an otherwise noisy JFET amp (its advantage is high impedance, so it's less noisy around high impedance sources -- noise is, in part, an input matching thing).
The input bias current is also fairly high (~1uA). Most bipolar amps are in the 200nA range. (FET and MOS amps are in the nA to pA range, but they're also better for high impedances, like I said.)
Here's a better "audio" op-amp for comparison:
http://www.ti.com/lit/ds/symlink/lm833-n.pdfSame 4nV/rtHz noise spec, but it extends down to much lower frequencies (~7 at 10Hz, meaning very much less rumble/flicker noise than the '971). This is achieved with the same input bias current. PSRR is also much better (Fig 14 and 16, CMRR and +PSRR), which means less demands for filtering those supplies.
Of course, the problem is, you can't use LM833 because it requires +/-5V supplies to really begin to be useful.
You should seriously consider this: if you are going to the trouble of making a 24 bit system, it will be well worth the added expense and difficulty of using boosted voltage supplies. 24 bits is HARD. And I mean for the experts, it's hard. You have essentially no chance of getting true 24 bit performance here (even with a DAC that was fundamentally capable); you have even less chance of being able to measure and verify how many bits it's actually doing!
TPA152 (Headphone Amp)
Hmm, kind of heavy on current draw for battery operation, if needed. Pretty noisy, but, you probably won't notice because it's a high level output.
That said, out of a +/-2.5V range, 6uV noise is 40 LSBs at 24 bits, not counting the noise of anything else. 40 is a bit more than 2^5, so it's more like 19 bits already.
I don't like how it's documented: they're only giving parameters and curves directly corresponding to its intended purpose, not its actual function -- which is a specially configured power op-amp pair, so it should really have op-ampy data. But like I said about special things, if it works it works.
I don't know what value to look for in the datasheet. Can someone explain?
At a basic enough level, you're going to have to look up the definitions in the first place. Sometimes these are provided in the datasheet (i.e., what negative current means, where to measure currents and voltages, what test circuits/jigs are required for DC and AC tests, etc.). Sometimes, it'll be typical for the family: a 74HC00 datasheet might be pretty short, but somewhere, the manufacturer will have to say that, yes, their product line meets the de-facto standard of everyone else's 74HC00s, and that everything else in that logic family meets the same input and output pin characteristics. So sometimes you can absorb it, piece by piece, after looking at an awful lot of datasheets and app notes. (No, this isn't very directed instruction...) Having a working understanding of the characteristics (using them in your own breadboarded circuits, and measuring DC parameters and AC waveforms), their range of behavior (even outside what the datasheet says), is also helpful, and reinforces that knowledge.
Tim