AMB Laboratories DIY Audio

[NwAvGuy]

Ok,

Let's try to take these issues one at a time...

D4 does nothing when running the Mini3 off the correct battery. So that had no effect. Anyone who doesn't understand that doesn't know how zener diodes work.

The OPA690 was working fine, it just can't handle enough current.

I show my "mains powered test rig" reproducing a -115 dB signal with NO 60 hz hum visible. It's a $10,000 piece of equipment. It doesn't have a hum problem.

The dScope has floating inputs. There are NO grounding issues as a few seem to suspect. It's designed to test balanced and bridged outputs. The input and output grounds are isolated from both each other and the chassis. You can find a diagram of the input scheme here:

http://www.prismsound.com/test_measure/support_subs/support_tech.php?tt=0028

The AD8397 does require careful implementation. My point is it would work much better in a classic 2 battery Cmoy design without the virtual ground/rail splitter. That's not to say it still wouldn't need a careful PCB layout, etc.

Sorry if it's insulting, but several of the published specs are wrong. For example, the Mini3 can't make anywhere close to 300 mW into 33 ohms running on the battery. It can't even do that IN THEORY let alone reality. Anyone who understands ohms law and basic circuits can confirm this. By the time you factor in the 6 ohm series output resistor, internal impedances in the op amps, etc. the numbers I measured are entirely realistic. The published numbers are not.

The volume setting for the crosstalk into 15 ohms and 150 ohms was the same. The load was the only thing changed. The virtual ground just doesn't like low impedance loads (like 16 ohm headphones). The issue is not the volume control.

I can't comment on the FiiO E11 as I've not had any experience with it. But that has little to do with what I measured with the Mini3.

I really have nothing against DIY or AMB. I'm just telling it like it is. I'm sorry if some of you would rather attack the source than believe the truth.

EDIT: The DC coupling issue is open to debate or personal preference. I just wanted to make everyone aware it exposes headphones to possible damage if you connect the Mini3 to something with a DC offset or leakage at the output. Personally I think it's a poor trade off to make when a cap doesn't hurt the sound. In a DC servo design, or with an output DC protection circuit, it would be fine to have a direct coupled input.

[nightanole]

Well lets hope this discussion doesnt get too bad. It would have been nice if the amp was made by MisterX and that would eliminate alot a variables. I can speak too much of mine since its a carrie amp with a custom 2 watt power supply and only a gain of 2.

http://www.youtube.com/watch?v=q4mb7VfJIe8

[NwAvGuy]

I'm confident the Mini3 I tested was working as well as any Mini3 built on the same v2.01 PCB from the AMB parts list can work. Most of what I found can be explained by just studying the schematic, datasheets, and doing some math. The two channels measured nearly identically, the gain and voltage swing were as expected, the virtual ground was working as it should, etc. I have close up pictures of the board if anyone is really curious about anything specific?

[amb]

D4 does nothing when running the Mini3 off the correct battery. So that had no effect. Anyone who doesn't understand that doesn't know how zener diodes work.

That is true if the amp was only ever powered by a battery. Without D4, and if there is no battery installed, the wallwart charger's voltage will get around the 7812 voltage regulator via the LM317L constant current source, and since OPA690's maximum supply voltage is 12VDC, there is a potential for it to become damaged. So D4 is not an optional part, it's required for safe operation of the amp.

The OPA690 was working fine, it just can't handle enough current.

I can't take that simply at face value without my own testing. Why have you not contacted us first before posting your "review" (and not only that, advertise it on diyaudio.com)?
http://www.diyaudio.com/forums/headphon ... blems.html
Since you publish reviews, you have the power to do damage (to a product, or to the reputation of the people behind it) if you write something disparaging. You should have a sense of journalistic responsibility to work out any issues with us first.

I show my "mains powered test rig" reproducing a -115 dB signal with NO 60 hz hum visible. It's a $10,000 piece of equipment. It doesn't have a hum problem.

You put the Mini³ directly on a power transformer and saw magnetic interference. Aluminum doesn't do a thing for such interference. What a discovery!
This is a portable amp meant to go in pockets, backpacks or somesuch. Putting it on a transformer is hardly a normal mode of use.

The dScope has floating inputs. There are NO grounding issues as a few seem to suspect. It's designed to test balanced and bridged outputs. The input and output grounds are isolated from both each other and the chassis.

OK, that's good, but there are still possible issues with ground and cabling, and whether the dScope's internal ground link is engaged or not. Without having worked through your specific test environment I also can't rule out any possible problems.

The AD8397 does require careful implementation. My point is it would work much better in a classic 2 battery Cmoy design without the virtual ground/rail splitter. That's not to say it still wouldn't need a careful PCB layout, etc.

Agree on the AD8397 being finicky, and it has no output current limiting (which is partly why it sounds so good), but it does need a lot of care in PCB layout and usage. You have not been through the whole Mini³ design process and do not know just how much work went into making the AD8397 happy.

the Mini3 can't make anywhere close to 300 mW into 33 ohms running on the battery. It can't even do that IN THEORY let alone reality. Anyone who understands ohms law and basic circuits can confirm this. By the time you factor in the 6 ohm series output resistor, internal impedances in the op amps, etc. the numbers I measured are entirely realistic. The published numbers are not.

A fully-charged 9V 300mAH NiMH battery produces well over 10V. The D3 diode and R5s do drop a little voltage, but I was still able to measure 9Vp-p (3.2Vrms) output before onset of clipping. This is not just calculation. It's real. 3.2Vrms into 33 ohms is a hair over 300mW. The current into the 33 ohm load at this point is 3.2V / 33 ohms = 97mArms or 137mA peak. This is well within the current output capabilities of the OPA690. Now that is a single channel. When both channels are driven with a mono signal then yes, we exceed the current capability of the OPA690, but low-Z headphones are usually very sensitive so no one will be listening to such headphones at 9Vp-p, so this is fairly academic.

The volume setting for the crosstalk into 15 ohms and 150 ohms was the same. The load was the only thing changed. The virtual ground just doesn't like low impedance loads (like 16 ohm headphones). The issue is not the volume control.

But was your volume pot set to maximum for all your measurements? If not, then there is a problem. It's not just on the Mini³, but all pots will cause stereo crosstalk when turned part-way down. It's just a fact of life.

Nowhere on in my specs did I test with a 15 ohm load, so your comparisons are not fair. That's less than half the load impedance of my lowest tested load, which of course doubles the output current for any given output voltage. That is a difficult load to drive at full output voltage swing, but as I said again, it's an unrealistic case. Headphones down in that kind of impedance range are typically IEMs that need no amplification at all - they need attenuation. Since the standard headphone TRS jack shares a common ground wire between two channels, if the jack contact and the wires have significant impedance (even as low as a fraction of an ohm), when you're driving high currents through them, there will be deterioration of stereo crosstalk. Take the above 97mA for example, being driven on one channel, if your cable adds as little as 0.5 ohm of shared ground impedance, then the common ground voltage "as seen at the test equipment" will wiggle 97mA * 0.5 ohms ~= 49mV. This 49mV wiggle on the "quiet" channel, relative to 3.2V on the driven channel, is a -36dB crosstalk. Now this is not a fault of the amp. It's again, life in reality of real wires having non-zero impedances. To produce measured results that reflect the performance of the actual device (rather than the test cables, etc.), is an exercise that must be taken very carefully.

As for your low battery run-time, there are a couple of issues. First, you have a 200mAH battery. I tested with 300mAH and 270mAH batteries. That's a significant difference. Also, NiMH batteries require a number of charge/discharge cycle to reach their maximum run-time. I also mentioned time and again that the actual run-time may vary depending on the type of load you put on it, and how much output current is being dumped through it.

[linux-works]

that's what I was talking about when I asked if the pot was all the way up, 'on full' or not.

the proper way to rmaa is to send in an analog source that drives the amp under test to just under clipping. to do this properly I would use a stepped atten on the output of your DAC; the DAC that you are using to send analog to the amp under test. when you calibrate and are setting the zero point, you want that to also be the point where you are -just- about to clip if you went higher than that on your DAC.

you don't calibrate this to 'eleven oclock' on the pot; you set the pot to bypass (no atten) and you atten *outside* of the circuit.

at least that's my understanding of the calibration process for rmaa.

if you are putting the pot at anything other than 'full', you are not doing the test properly.

[NwAvGuy]

That is true if the amp was only ever powered by a battery. Without D4, and if there is no battery installed, the wallwart charger's voltage will get around the 7812 voltage regulator via the LM317L constant current source, and since OPA690's maximum supply voltage is 12VDC, there is a potential for it to become damaged. So D4 is not an optional part, it's required for safe operation of the amp.

Yes, I understand its purpose. It does, however, create a small leakage current across the battery that might shorten the storage life.

Since you publish reviews, you have the power to do damage (to a product, or to the reputation of the people behind it) if you write something disparaging. You should have a sense of journalistic responsibility to work out any issues with us first.

If I thought something might be wrong with the Mini3 I tested I would have contacted you. But I'm 99.9% sure it was working as intended when I tested it. You may see my review as damaging, but I see it as helping people make informed choices about what they spend their time and money on. I personally think it's "damaging" to make claims that are not true--like the claims you make for the Mini3's virtual ground, power output, battery life, etc.

You put the Mini³ directly on a power transformer and saw magnetic interference. Aluminum doesn't do a thing for such interference. What a discovery!
This is a portable amp meant to go in pockets, backpacks or somesuch. Putting it on a transformer is hardly a normal mode of use.

No, the point is it picked up hum a good 15+ inches away from the transformer. And, while on top of the transformer, it was 20 dB higher than the E7 placed in the same spot. 20 dB is huge. In this case it's 70 dB vs 90 dB. The first is plainly audible, the second one isn't. I honestly don't know how much it's a problem in real world use, but it's not normal. So I published what I found.

OK, that's good, but there are still possible issues with ground and cabling, and whether the dScope's internal ground link is engaged or not. Without having worked through your specific test environment I also can't rule out any possible problems.

The ground link was not enabled. The tests I run on the dScope are all saved in configuration files. Many are the the same tests I've used on dozens if not hundreds of pieces of gear. One click of the mouse sets up everything the same for every test.

You have not been through the whole Mini³ design process and do not know just how much work went into making the AD8397 happy.

You're right. I can only report on the final design you ended up with. And, from what I've seen, it could still use some work.

A fully-charged 9V 300mAH NiMH battery produces well over 10V. The D3 diode and R5s do drop a little voltage, but I was still able to measure 9Vp-p (3.2Vrms) output before onset of clipping. This is not just calculation. It's real. 3.2Vrms into 33 ohms is a hair over 300mW. The current into the 33 ohm load at this point is 3.2V / 33 ohms = 97mArms or 137mA peak. This is well within the current output capabilities of the OPA690. Now that is a single channel.

Ok, let's look at all of the numbers. First, the "well over 10V" statement is misleading. Fully charged NiMh cells start out at 1.35 volts each, for a 7 cell NiMh battey, that's 9.45 volts--not "well over 10V". If you don't believe me, have a look at:

http://data.energizer.com/PDFs/nh22-175.pdf

The discharge curves, even at 17.5 mA (less than the Mini3 draws) start out barely above 9 volts and rapidly (as in after just a few minutes) fall to 9 volts. So let's use 9 volts as a battery voltage which is still only good for the first 30% or so of the battery life.

After the 1N4001, we have 8.2 volts DC. The AD8397 will drop at least another 0.2 volts from each rail so that's 7.8 V p-p out of the AD8397. The resistor divider formed by the 6.2 ohm output resistor (R5) and the 33 ohm load drops that to 6.06 V p-p. That's 2.14 V rms. P = 2.14*2.14/33 = 138 mW best case not including ANY losses in the OPA690.

So, I'm sorry, but your 300 mW number is VERY WRONG! If you want to know why I got 98 mW instead of 138 mW, that's easy. Both channels were driven and the OPA690 was in over its head--even into 33 ohms. And this is all assuming pushing the output right up to clipping. I measured, as is the accepted standard, to 1% THD. And all the various non-linearities, not the least of which is the feedback mess created by having that 6.2 ohm output resistor inside the feedback loop, conspire to push the THD to 1% before you actually get to hard clipping. So the 138 mW drops to 98 mW on a freshly charged battery with both channels stressing the OPA690.

When both channels are driven with a mono signal then yes, we exceed the current capability of the OPA690, but low-Z headphones are usually very sensitive so no one will be listening to such headphones at 9Vp-p, so this is fairly academic.

If you look at my results table, I rate the output into 15 ohms as "Excellent" so I agree ~100 mW is plenty at that impedance. My issue is the number you publish in your specs isn't a little bit wrong, it's off by a factor of three. I think that's very misleading.

But was your volume pot set to maximum for all your measurements? If not, then there is a problem. It's not just on the Mini³, but all pots will cause stereo crosstalk when turned part-way down. It's just a fact of life.
Nowhere on in my specs did I test with a 15 ohm load, so your comparisons are not fair.

I checked the crosstalk with your 33 ohm specified load. It measured -46 dB. You list it as -88 dB. That's not a small difference. With a 150 ohm load, at the same volume setting, it measured 60 dB. At 15 ohms and, the same volume setting, it measured 40 dB. The volume setting was the same for all 3 measurements. The difference is the load. I'm happy with 60 dB and if you want to blame that number on the volume control, go ahead. But what's mostly causing the drop in channel separation is the OPA690 falling down on the job as the impedance drops. Some of it might be because the two output channels are on a single op amp die. But neither have anything to do with the volume control. There's no way the Mini3 can do anything close to -88 dB at ANY volume setting into 33 ohms. Period. Again, your number is very misleading.

Since the standard headphone TRS jack shares a common ground wire between two channels, if the jack contact and the wires have significant impedance (even as low as a fraction of an ohm), when you're driving high currents through them, there will be deterioration of stereo crosstalk.

All my measurements use the same cables, levels, loads, and set up including the FiiO E7. The E7 measured 63 dB into 15 ohms sharing current as you describe above at 400 mV output. The Mini3 measured 40 dB under the same conditions. You can't blame the huge 23 dB difference on the common ground wire in the cable. Sorry. Same wire, same cable, same current, same everything. The difference is the FiiO doesn't use a shared virtual ground.

As for your low battery run-time, there are a couple of issues. First, you have a 200mAH battery. I tested with 300mAH and 270mAH batteries

Do you have links for the datasheets for these batteries? I've never seen an 8.4 v NiMh battery with a capacity that high. But even 50% more mAH rating still won't get you to 10+ hours in real world use. Face it, it's a misleading number. What good is a headphone amp with the volume turned all the way down?

I'm really not trying to pick a fight, but I stand behind every one of my measurements. And I'll defend them if someone tries to claim they're wrong unless someone can prove where I made a mistake. So far, you've not shown where I've been wrong.

[NwAvGuy]

that's what I was talking about when I asked if the pot was all the way up, 'on full' or not.

the proper way to rmaa is to send in an analog source that drives the amp under test to just under clipping. to do this properly I would use a stepped atten on the output of your DAC; the DAC that you are using to send analog to the amp under test. when you calibrate and are setting the zero point, you want that to also be the point where you are -just- about to clip if you went higher than that on your DAC.

you don't calibrate this to 'eleven oclock' on the pot; you set the pot to bypass (no atten) and you atten *outside* of the circuit.

at least that's my understanding of the calibration process for rmaa.

if you are putting the pot at anything other than 'full', you are not doing the test properly.

First I'm not using RMAA. It's prone to all sorts of problems.

Second, for reasons of consistency among products with wildly different maximum gains, I use unity gain, or sometimes 2X gain, as standards.

Third, as I explain in the previous post, I measured 60 dB of channel separation with a 150 ohm load and the same volume setting. I'm fine with 60 dB. That's a respectable number. But as the load impedance drops on the Mini3, so does the channel separation. And this has nothing to do with the volume control. It drops to 46 dB at 33 ohms and that's getting to be rather marginal. This won't be improved by using a different volume setting (for the same output level).

If you're curious about more of what to watch out for with RMAA, check out:

http://nwavguy.blogspot.com/2011/02/rightmark-audio-analyzer-rmaa.html

[linux-works]

rmaa or not, I still question why you put the volume at other than full.

each amp has its own gain and max input and output. to use one test to judge all amps seems wrong to me.

so again, for me, I'm curious what this eleven oclock stuff was all about. please tell me why you think that having a pot in the circuit is a good test strategy when its the amp you want to characterize.

[amb]

That is true if the amp was only ever powered by a battery. Without D4, and if there is no battery installed, the wallwart charger's voltage will get around the 7812 voltage regulator via the LM317L constant current source, and since OPA690's maximum supply voltage is 12VDC, there is a potential for it to become damaged. So D4 is not an optional part, it's required for safe operation of the amp.

Yes, I understand its purpose. It does, however, create a small leakage current across the battery that might shorten the storage life.

Like less than 0.1uA leakage under normal temperatures? On a real battery that leakage is insignificant.

Since you publish reviews, you have the power to do damage (to a product, or to the reputation of the people behind it) if you write something disparaging. You should have a sense of journalistic responsibility to work out any issues with us first.

If I thought something might be wrong with the Mini3 I tested I would have contacted you. But I'm 99.9% sure it was working as intended when I tested it. You may see my review as damaging, but I see it as helping people make informed choices about what they spend their time and money on. I personally think it's "damaging" to make claims that are not true--like the claims you make for the Mini3's virtual ground, power output, battery life, etc.

You should have worked with us first regardless of whether you believe the amp is working properly or not. I stand behind my published specs. They were taken under my own testing environment and methodology which differ from yours.

You put the Mini³ directly on a power transformer and saw magnetic interference. Aluminum doesn't do a thing for such interference. What a discovery!
This is a portable amp meant to go in pockets, backpacks or somesuch. Putting it on a transformer is hardly a normal mode of use.

No, the point is it picked up hum a good 15+ inches away from the transformer. And, while on top of the transformer, it was 20 dB higher than the E7 placed in the same spot. 20 dB is huge. In this case it's 70 dB vs 90 dB. The first is plainly audible, the second one isn't. I honestly don't know how much it's a problem in real world use, but it's not normal. So I published what I found.

Without looking into why there was hum? Like you said, an amp running on batteries shouldn't have hum, if it's also connected to a source that itself has no hum. My published graphs show no hum, and I am very careful with placement of test equipment, cables, etc. to ensure that external influences don't affect the measured results (except where I am looking for such things).

You have not been through the whole Mini³ design process and do not know just how much work went into making the AD8397 happy.

You're right. I can only report on the final design you ended up with. And, from what I've seen, it could still use some work.

Every aspect of the design have been carefully considered. You mention the R5 resistors. Those are a must to protect the AD8397 from blowing out, because as I mentioned these opamps have no current-limiting protection. I had a bunch of dead AD8397s to prove it. But simply placing the resistors in series with the output is not a solution, because it raises the output impedance. Keeping it wrapped within the feedback loop is the only way to accomplish the protection while maintaining low output impedance. That resistor is really no different than, say the emitter resistors in a discrete complementary push-pull BJT emitter follower pair. Those are in the feedback loop too.

Also, we have determined early on that, despite the datasheet's claim of 24V rating, the AD8397 is not stable at such voltages. It gets too hot and triggers parasitic or outright oscillations that would also cause chip-death. The original Mini³ v1 (never released) was a dual-9V battery design and it was not stable at all. Mini³ v1 showed us that AD8397 is also not unity-gain stable. Tangent also had to cancel his PINT project which was similar and had issues. For Mini³ v2 I changed to a single-battery solution which also allowed me to use the smaller and sturdier Hammond enclosure.

A fully-charged 9V 300mAH NiMH battery produces well over 10V. The D3 diode and R5s do drop a little voltage, but I was still able to measure 9Vp-p (3.2Vrms) output before onset of clipping. This is not just calculation. It's real. 3.2Vrms into 33 ohms is a hair over 300mW. The current into the 33 ohm load at this point is 3.2V / 33 ohms = 97mArms or 137mA peak. This is well within the current output capabilities of the OPA690. Now that is a single channel.

Ok, let's look at all of the numbers. First, the "well over 10V" statement is misleading. Fully charged NiMh cells start out at 1.35 volts each, for a 7 cell NiMh battey, that's 9.45 volts--not "well over 10V". If you don't believe me, have a look at:

http://data.energizer.com/PDFs/nh22-175.pdf

I don't have to believe or not believe you, because I actually measured the voltage of my freshly-charged battery. It was 10.something volts (and yes, it was a 7-cell battery). An 8-cell version has even more voltage. Since there is more battery voltage than you claim (maybe your battery isn't fully charged?), it throws your whole measurement.

But what's mostly causing the drop in channel separation is the OPA690 falling down on the job as the impedance drops.

Yes, but as I already said, your measurements are with less than half the load impedance of my measurements so it stresses the amp much more than it was designed to do, or would encounter in real life. A cmoy using the standard OPA2132 opamp could output 40mA, yet I don't see you testing them with 15 ohm load at max voltage swing and crying foul?

Some of it might be because the two output channels are on a single op amp die. But neither have anything to do with the volume control. There's no way the Mini3 can do anything close to -88 dB at ANY volume setting into 33 ohms. Period. Again, your number is very misleading.

I think it is a given that your tests are not run with the volume control wide open. Some of your test graphs even say so. Crosstalk is additive if they're in phase, so it's a contribution from both the volume pot position and other factors I mentioned. I take testing very seriously (and those whom I have worked with or met, and saw me doing these tests will concur). Even little things like cables and how they are routed, and whether a plug is tight in its jack, are not left to chance. My RMAA test results speak for themselves. If anything, the RMAA tests often make the test results worse than reality because the sound card has its own analog circuitry that become bottlenecks.

As for your low battery run-time, there are a couple of issues. First, you have a 200mAH battery. I tested with 300mAH and 270mAH batteries

Do you have links for the datasheets for these batteries? I've never seen an 8.4 v NiMh battery with a capacity that high.

I used Accupower and Maha 9V NiMHs, they come as high as 300mAH. CTA even makes a 325mAH NiMH 9V.

I'm really not trying to pick a fight, but I stand behind every one of my measurements. And I'll defend them if someone tries to claim they're wrong unless someone can prove where I made a mistake. So far, you've not shown where I've been wrong.

If you're not picking a fight, then I don't know what is.

Unfortunately, when it comes to these measurements, the testing environment, methodology, and other factors greatly affect the results. We came to very different conclusions. You seem to already have strong opinions about certain elements of Mini³'s design, and that's fine. But to concoct a set of tests to "prove" them is like putting a cart before the horse.

[NwAvGuy]

Like less than 0.1uA leakage under normal temperatures? On a real battery that leakage is insignificant.

I don't think it's that low with 9+ volts on a 12 volt zener. But I'd have to check.

I stand behind my published specs. They were taken under my own testing environment and methodology which differ from yours.

Well then I'd love to see how you get 300 mW into 33 ohms running on a 7 cell NiMh battery. I've proven, with simple math, it's literally IMPOSSIBLE. You can't even get close even under ideal conditions. You're flat wrong or mistaken.

Without looking into why there was hum? Like you said, an amp running on batteries shouldn't have hum, if it's also connected to a source that itself has no hum. My published graphs show no hum, and I am very careful with placement of test equipment, cables, etc. to ensure that external influences don't affect the measured results (except where I am looking for such things).

If you read my review, I did look into it. I saw the hum and thought it rather strange from a device running from battery. So I checked the dScope in loopback, and published the graph showing the approximately -160 dB noise floor. No hum. So I started moving the Mini3 around and found it was picking up stray hum from the dScope's transformer when it's well over a foot away. Magnetic fields are roughly proportional to the square of the distance (exponential). So the dScope's own higher gain analog circuits, being just a few inches from the same transformer, are subjected to literally 10+ times greater magnetic field. I even went to the trouble to hook up the E7 and check it in the same positions. No hum at all on the bench, and only -90 dB sitting on top of the dScope. So I spent quite a bit of time "looking into why there was hum" and came to the clear conclusion the Mini3 is unusually sensitive to EMI.

You mention the R5 resistors. Those are a must to protect the AD8397 from blowing out, because as I mentioned these opamps have no current-limiting protection. I had a bunch of dead AD8397s to prove it. But simply placing the resistors in series with the output is not a solution, because it raises the output impedance. Keeping it wrapped within the feedback loop is the only way to accomplish the protection while maintaining low output impedance. That resistor is really no different than, say the emitter resistors in a discrete complementary push-pull BJT emitter follower pair. Those are in the feedback loop too.

Emitter resistors are usually a fraction of an ohm. So they create much less of a problem. If the AD8397 blows up into a short, I agree something has to be done, or you have to use a different op amp. It's a matter of picking the best compromise. But, regardless, I was just reporting on the performance of the Mini3 as it's designed. And that includes the 6.2 ohm output resistor. If it has to be there, fine. But people should know it causes some problems.

Also, we have determined early on that, despite the datasheet's claim of 24V rating, the AD8397 is not stable at such voltages. It gets too hot and triggers parasitic or outright oscillations that would also cause chip-death.

I never suggested running it at 24 volts. I did suggest running it at 18 volts and suspect that can be made to work just fine if you know what you're doing.

The original Mini³ v1 (never released) was a dual-9V battery design and it was not stable at all. Mini³ v1 showed us that AD8397 is also not unity-gain stable.

First, this isn't a unity gain application. The chip is set for nearly 15 dB gain. Second, no offense, but your comments, design choices, etc. demonstrate you can't even properly calculate maximum output power. So I suspect you're even less able to calculate loop stability, phase margin, or other things required to make the AD8397 stable.

I don't have to believe or not believe you, because I actually measured the voltage of my freshly-charged battery. It was 10.something volts (and yes, it was a 7-cell battery). An 8-cell version has even more voltage. Since there is more battery voltage than you claim (maybe your battery isn't fully charged?), it throws your whole measurement.

Even if you do the math with "10.something" you still don't get 300 mW. Sorry. Yes a fresh off the charger battery might be over 10 volts, but that's called "surface charge" and it goes away as soon as you put even a light load on it for more than a few seconds. The battery datasheet proves my case. If you have a datasheet that shows otherwise, please post the link?

We're not off by a little bit here. Your power rating is off by a FACTOR OF THREE. There's no way the Mini3 can produce anywhere close to 300 mW into 33 ohms. If you believe it can, you don't understand audio engineering enough to be doing this. I showed the numbers. Even best case, you get around 130 mW. If my numbers are somehow wrong, please show me exactly where they're wrong? Even better, please show me the math that leads to 300 mW?

Yes, but as I already said, your measurements are with less than half the load impedance of my measurements so it stresses the amp much more than it was designed to do, or would encounter in real life. A cmoy using the standard OPA2132 opamp could output 40mA, yet I don't see you testing them with 15 ohm load at max voltage swing and crying foul?

There are some big differences here. I haven't seen any Cmoy's claiming high power levels into 33 ohms like you claim for the Mini3. I also haven't seen any claims like "it can drive low and high impedance headphones with authority" made for Cmoy amps. The information you provide leads one to believe the Mini3 handles low impedance loads just fine. If you don't believe the Mini3 does, you should change your website to be more accurate.

Crosstalk is additive if they're in phase, so it's a contribution from both the volume pot position and other factors I mentioned.

The crosstalk is not sufficiently "additive" in this case. At best, it might amount to a few dB. The dB scale, as you hopefully know, is logarithmic. The crosstalk contribution of the volume control is a tiny fraction of the contribution of the OPA690. By comparison, the volume control's crosstalk is a drop in the bucket. Again, if you don't understand this, that's not a good sign.

But to concoct a set of tests to "prove" them is like putting a cart before the horse.

I haven't "concocted" anything. I subjected the Mini3 to the exact same tests as the other gear I've tested. Same levels, same loads, etc. The FiiO E7 tests, in particular, are an extremely close match. I'll use the same tests for the next headphone amp. I'm not trying to put the Mini3 at any unfair disadvantage. It's apples-to-apples. That's as fair as I can be.