NwAvGuy wrote:Thanks Dorkvader. Anyone who's worked a lot with op amps will generally admit they usually either work or they don't. And when they're only "lightly damaged" it almost always manifests itself as excessive DC offset or other problems with the input circuitry. The problems I observed with the OPA690 are with the most rugged part of the chip--the high current output stage.
Op amps are also rather intolerant of implementation errors. They tend to either work, be marginally stable and "ring", go wild and oscillate, or most often, just slam into a supply rail and peg their output at V- or V+. I've never known an op amp to produce less output current than spec, or suffer similar odd errors that would explain the Mini3's poor performance. If an output transistor in an op amp is damaged, it slams into the supply rail.
Stereo amplifiers have two channels so you can conveniently compare them to each other. It's unlikely the AD8397 would somehow be magically damaged equally in both channels or two resistors will both symmetrically be in the wrong locations, etc. Lots of my measurements were made on both channels and there were no interchannel problems. The OPA690 is shared between the two channels, but it behaved as would be expected in my tests and from its datasheet.
Please note, the Mini3's distortion, crosstalk, frequency response, and more, into 150 ohms, all measured just fine. If the Mini3, or OPA690, had anything significant wrong with it, it wouldn't have likely shown at least some sign on at least one of these tests.
By conducting my full suite of measurements on the Mini3 I pretty much did verify it's working properly. A manufacture mass producing the Mini3, for example, would conduct similar tests during manufacturing.
The Mini3 is not some elaborate complex device. It only has a few active parts and handful of resistors and caps. Anything out of place, assembled wrong, damaged, etc. would tend to be rather obvious in testing. But even still, I did a lot of extra investigation to make sure there wasn't something "weird" going on.
And the test methodology in this case isn't that complicated. There are not many variables to control with a headphone amp. It's pretty much just the input signal (level & frequency), volume setting, and load on the output. From the analysis side, as I've mentioned, the Prism dScope uses pre-configured files for each test. That assures consistency between say the FiiO E7 tests and the Mini3 tests. For most tests I only have to set the reference level (most of the tests show the output level used--typically ~400 mV) and let the dScope do its thing. That eliminates most of the chances for human error.
The whole "you have a bad/damaged/incorrect/whatever Mini3" argument might seem like a valid possibility, but in this case, it's very remote. Anything wrong with the one I tested would have likely been obvious in my tests.
This is, in my opinion a good post. I'm not knowledgeable enough (yet) to know if it's right, but it's well worded, and easy to understand.
Here's what I think: You should design a small, 2-battery, 2-channel "portable"-size amp like the one you describe. I'm planning on building one in a year or two, and I want variety. If you take into account all the things you say AMB didn't, it'll probably perform admirably. I am
all for more designs, and if it's better, it's better. I have plenty of time to spend deciding what I'm building next, so I'm very interested in new developments.
When you say you'd spend the money saved on having only 2 channels on a second battery, do you mean you'd have one for +, one for - and the virtual ground inbetween? Since the batteries will discharge seperately, won't you end up with poor long term performance (if one battery is at, say, 9V and the other at 8V)
As far as "what ground channel amp would you use" question, it's not meant to be a loaded question, but I do see your point. The "trick answer" is that you wouldn't use a 3-channel design. But let's say there's a huge market for these newfangled 3-channel amplifiers, and your boss comes to you one day, sighing. He tells you that he knows about all the hardships, and difficulties the 3-channel designs have, apologizes, and asks you to design one that's significantly less bad than the rest. So hypothetically, what would you do?
See, then you can state your thing about how 3-channel designs being bad. You can list what aspects or properties about opamps make the 3-ch design worse, and you'll know what to minimize. I think it's an interesting question, and I'm pretty curious. If you're making a "not-as-terrible" 3-ch design, what would you look for in a good ground-channel opamp?
Hmm, if you leave the ground channel unpopulated (like what the balanced (2-board) mini^3 people do) would it perform better? I feel like that's easy to test.
Finally, I still think we can do a mini^3 survey by proxy, where people measure the output power in some sort of standardized way. (even if the measuring tools aren't exact: for example different multimeters, etc.) Even if everyone's wrong about it (which would be the "worst case scenario," I think) It'd be cool to see the data.
