^^That's great. It sounds better. To you. With your headphones. It still has greater impedance than the real ground. It increases distortion and crosstalk. One appears to be a function of the other.
No one is knocking you for liking the way it sounds. I happen to love the way many tube amps sound. They won't measure better than a properly implemented opamp based amplifier. Not even close.
If the idea of the ground channel is to reduce distortion and cross talk, it fails in that respect. It most certainly doesn't decrease impedance. That is all anyone has ever suggested. A better idea is to use a battery minder that will keep the charge in the cells balanced and includes a low voltage shutdown.
Honestly, the likely hood of two rechargeable 9v batteries becoming dangerously imbalanced is low. Even more so if they are charged separately. Which they usually would be. If one cell begins to take longer to charge than the other, it might be time to buy a new set. That isn't really enough insurance for an expensive set of phones. Also as stated, one could pop out.
I would rather waste the current on battery protection and kill two birds with one stone. Three if you decide to include the charging setup.
No one is knocking you for liking the way it sounds. I happen to love the way many tube amps sound. They won't measure better than a properly implemented opamp based amplifier. Not even close.
If the idea of the ground channel is to reduce distortion and cross talk, it fails in that respect. It most certainly doesn't decrease impedance. That is all anyone has ever suggested. A better idea is to use a battery minder that will keep the charge in the cells balanced and includes a low voltage shutdown.
Honestly, the likely hood of two rechargeable 9v batteries becoming dangerously imbalanced is low. Even more so if they are charged separately. Which they usually would be. If one cell begins to take longer to charge than the other, it might be time to buy a new set. That isn't really enough insurance for an expensive set of phones. Also as stated, one could pop out.
I would rather waste the current on battery protection and kill two birds with one stone. Three if you decide to include the charging setup.
Thanks Rembrant! Two 9 volt batteries with a charge monitor is exactly what I've done with the O2 project:
O2 DIY Headphone Amp Thread
As expected, it outperforms the more costly 3 channel Mini3 in literally every measurement and has significantly longer battery life as well.
I was wondering this 3 channel thing. I have been in impression that it could be beneficial but reading your blog has made think it a little bit more. How much the shared ground wire of headphones affect the crosstalk? RocketScientist, maybe the measured crosstalk does not need to be lower than like -40 dB?
I need to build a nice dual supply for my current headphone amplifier, that at the moment has paralleled NJM4556 (both channel used) buffering the resistor divider virtual ground, input and output grounds the same. It would be interesting to test, but the star grounding must be very good too. NJM4556 should be able to output more than OPA690.
Have you tried and measured AMBs M3 headphone amplifier? It has very robust ground channel, that will not get current starved for sure. Also there is lots of negative feedback added that should lower the output impedance (high GBW op amps used)
I am not quite sure why Ti chose to use OPA690 in Mini3 but that amplifier is a compromise anyway.
I need to build a nice dual supply for my current headphone amplifier, that at the moment has paralleled NJM4556 (both channel used) buffering the resistor divider virtual ground, input and output grounds the same. It would be interesting to test, but the star grounding must be very good too. NJM4556 should be able to output more than OPA690.
Have you tried and measured AMBs M3 headphone amplifier? It has very robust ground channel, that will not get current starved for sure. Also there is lots of negative feedback added that should lower the output impedance (high GBW op amps used)
I am not quite sure why Ti chose to use OPA690 in Mini3 but that amplifier is a compromise anyway.
Last edited:
The issue here isn't how robust the ground is, or even how much crosstalk is audible, the issue a virtual ground makes things worse instead of better. Even if it's flaws are made inaudible, it's still a waste of money. The M3 and 3 channel beta22 are just as flawed, in concept, as the Mini3.
If anything a small portable amp like the Mini3 has more reason to use a virtual ground so it can run from a single battery. A larger more costly amp like the M3 or beta22 don't have that excuse.
The biggest problem with the Mini3 isn't the crosstalk but the relatively severe interchannel distortion, and generally high distortion in general in many tests.
Can you post some measurement graphs of b22 2channel compared to 3 ch. active ground version and their differences? I would like to see something concrete already.
Not sure if you read my 3 Channel Article but there's plenty "concrete" in there including a direct comparison of the same amp using a real ground and a virtual ground. The eBay Cmoy and O2 are further concrete evidence. I'd love to test a beta22 head to head against the O2 and I've even offered an open challenge for anyone to put the M3 or beta22 up against the O2.
The only practical benefit to a 3 channel/virtual ground headphone amp is the ability to let it run from a single supply. Otherwise it's just a band-aid trying to correct for other design flaws that are much more easily solved just by using a proper conventional star ground. And, because the beta22 already has a dual bipolar power supply, it has zero reason to use a virtual ground.
Honestly, for the Mini3 at least, an output cap would have worked far better than the virtual ground. Output caps have their own advantages and disadvantages. But a proper bipolar supply with a conventional star ground is literally unbeatable for a headphone amp. I challenge anyone, including Ti Kan at AMB, to objectively demonstrate how another architecture can do better.So would an output coupling cap, but we won't go there, will we?
Having built and tested BUF634 open-loop buffered TLE2426 with an output impedance on the order of about 10-20 Ohms I can safely say this is not a route I'd like to go through again. There is just too much crosstalk and this seriously degrades the stereo image, and needs relatively large caps across the V-GND to offset this issue.
That said, I was wondering if using a linear regulator to regulate down to the half supply voltage would be considered a better idea than say... a buffered virtual ground wrapped within a opamp's feedback loop to decrease the output impedance?
I tend to think that linear regulators are fairly similar to an opamp-buffer topology, however they tend to burn off a lot more energy regulating to voltage. I was wondering if this could be a better idea if power consumption is not an issue.
(Too many unused laptop switching power supplies around here, they can be put to some use)
That said, I was wondering if using a linear regulator to regulate down to the half supply voltage would be considered a better idea than say... a buffered virtual ground wrapped within a opamp's feedback loop to decrease the output impedance?
I tend to think that linear regulators are fairly similar to an opamp-buffer topology, however they tend to burn off a lot more energy regulating to voltage. I was wondering if this could be a better idea if power consumption is not an issue.
(Too many unused laptop switching power supplies around here, they can be put to some use)
@nereis, I'm not sure how well a regulator would work. Most op amps used for audio have excellent common mode and power supply rejection so they can put up with quite a bit of variations/noise on the power supply rails. But ground noise is different. And a fixed voltage regulator would create ground and asymmetrical power supply noise.
A conventional regulator is a fixed voltage when what you really want is half the power supply voltage. In a battery powered amp it will create severe asymmetrical clipping as the battery gets low.
You would also end up with asymmetrical noise on the "rails" with respect to the virtual ground. The regulator will reference the virtual ground to the real ground (which from the amp's perspective is the negative rail). But the positive rail will have some ripple on it from the audio signal with respect to the virtual ground the regulator cannot correct for.
When you consider just 0.0001 volts is -80 dB from 1 volt it doesn't take much to cause potentially audible problems in an amp. That's why the Mini3 falls on its face in several distortion tests.
A conventional regulator is a fixed voltage when what you really want is half the power supply voltage. In a battery powered amp it will create severe asymmetrical clipping as the battery gets low.
You would also end up with asymmetrical noise on the "rails" with respect to the virtual ground. The regulator will reference the virtual ground to the real ground (which from the amp's perspective is the negative rail). But the positive rail will have some ripple on it from the audio signal with respect to the virtual ground the regulator cannot correct for.
When you consider just 0.0001 volts is -80 dB from 1 volt it doesn't take much to cause potentially audible problems in an amp. That's why the Mini3 falls on its face in several distortion tests.
I've tried both regulating with LM7812 then splitting & buffer to +/- 6V from a 19.3V isolated laptop switcher as well as simply splitting & buffering the laptop supply, both of were rather satisfactory noise-wise powering opamps CMOY-style with 7x gain.
Your comment on asymmetrical noise in different rails was rather insightful. I've noticed that isolated switching power supply while maintaining rather good voltage regulation between the positive & negative outputs, tends to move up and down when referenced to "earth".
However noise on the two rails tends to be rather symmetric on a good switcher and is effectively unseen by load for the most part. I suspect TLE2426 rail splitter might actually be quieter than a linear regulator if there is a lot of noise/asymmetry on the rails, since it tends to move its virtual ground potential according to noise. However I may just rebuild one with a LM317 to see if the potentially stiffer v-gnd improve crosstalk. (and I suspect it will)
I am sure there is still plenty ultrasonic noise left and may affect overall performance, but nothing audible with sensitive IEMs plugged-in.
Your comment on asymmetrical noise in different rails was rather insightful. I've noticed that isolated switching power supply while maintaining rather good voltage regulation between the positive & negative outputs, tends to move up and down when referenced to "earth".
However noise on the two rails tends to be rather symmetric on a good switcher and is effectively unseen by load for the most part. I suspect TLE2426 rail splitter might actually be quieter than a linear regulator if there is a lot of noise/asymmetry on the rails, since it tends to move its virtual ground potential according to noise. However I may just rebuild one with a LM317 to see if the potentially stiffer v-gnd improve crosstalk. (and I suspect it will)
I am sure there is still plenty ultrasonic noise left and may affect overall performance, but nothing audible with sensitive IEMs plugged-in.
To really measure the sort of noise I'm talking about you need to either have an audio analyzer or a really good RMAA set up where you can properly set levels into a proper load and run THD+N sweeps. I'm not talking about "hiss", but rather non-linearities created by the audio signal. The Mini3 has fairly low noise in terms of SNR. But all the non-linearities in the ground show up as higher THD+N--especially at high frequencies into low impedance loads. Even at only 400 mV it was around 0.5% THD+N at 20 Khz.
If the main supply rail is regulated that obviously solves the asymmetrical clipping issue. But you still have ground noise and the asymmetrical "rail" noise from the amp's perspective. And much of that noise will be proportional to how much current the amp is delivering to the load.
There is mentioned in PPA v2 docs that you can use dual supply on it. Tangent writes:
"Single Voltage or Dual Voltage Power Supply?
The PPA will only work with single-voltage power supplies unless you modifiy the circuit. If you have a dual-voltage power supply, it’s probably better to run the amp from the outer terminals and ignore the ground lead than to try and hook up the ground lead.
If you wanted to try it anyway, you could run the ground lead from the power supply to the ground plane, and jumper across the TLE positions. This should work, but it is doubtful that it will give improved performance relative to the standard power supply configuration. Since a dual supply costs more than a single, the only reason to try this is out of curiosity."
There is another argument that maybe has not been thought of, dual power supply has double the parts count, but the cost of 3 channel equalizes this. He thinks it doesn't give better performance to use dual supply.
I was wondering if I could mod my M3 from AMB to work with dual supply (LM317/LM337). I also have PPAv2 but I am more interested in M3 now as I just spent some money on parts.
"Single Voltage or Dual Voltage Power Supply?
The PPA will only work with single-voltage power supplies unless you modifiy the circuit. If you have a dual-voltage power supply, it’s probably better to run the amp from the outer terminals and ignore the ground lead than to try and hook up the ground lead.
If you wanted to try it anyway, you could run the ground lead from the power supply to the ground plane, and jumper across the TLE positions. This should work, but it is doubtful that it will give improved performance relative to the standard power supply configuration. Since a dual supply costs more than a single, the only reason to try this is out of curiosity."
There is another argument that maybe has not been thought of, dual power supply has double the parts count, but the cost of 3 channel equalizes this. He thinks it doesn't give better performance to use dual supply.
I was wondering if I could mod my M3 from AMB to work with dual supply (LM317/LM337). I also have PPAv2 but I am more interested in M3 now as I just spent some money on parts.
active supply splitting will cost efficiency,
with gnd plane/hybrid "star" using TRS gnd as input, feedback and output star "point" there is little inherent performance difference btween dual PS and active splitter
but that is not the case with "3-channel" topologies that separate the input and output gnd with the gnd channel amp
you would need to do a little more violence to the PPA to convert it to run on dual supply - like removing the 3rd channel as well, wiring dual supply gnd to output and input gnds (in "the right" order)
with gnd plane/hybrid "star" using TRS gnd as input, feedback and output star "point" there is little inherent performance difference btween dual PS and active splitter
but that is not the case with "3-channel" topologies that separate the input and output gnd with the gnd channel amp
you would need to do a little more violence to the PPA to convert it to run on dual supply - like removing the 3rd channel as well, wiring dual supply gnd to output and input gnds (in "the right" order)
I've offered several examples of real world performance differences and the results are fairly clear. The problem with statements like the one above is I doubt it's based on proper measurements. What DIY designers think, and reality, are often very different. I've provided several examples with AMB.
The O2's power supply adds less than a dollar to the parts cost to have a true bipolar power supply. So I think the "double the parts count" is a fairly weak argument when the 3rd channel op amp in the Mini3 (which isn't up to the job) is over $4 and another M3 channel is higher still.
You should be able to run the M3 from a proper dual power supply. The trick will be getting all the ground routing correct. Ground routing can have a huge impact on performance and the M3's PC board isn't designed for a proper single star ground.
I am trying to get most performance out of those parts. Maybe I should sell the PCBs
What if the ground channel is just left out, when using dual power supply? Schematic: http://www.amb.org/audio/mmm/mmm_sch.png
Layout: http://www.amb.org/audio/mmm/mmm_pcb.png
You can still get in on the O2 PCB buy Why not build an amp that has fully documented performance into proper loads and is based on solid engineering instead of myth? I strongly suspect the O2 will outperform the M3 in several areas.
Why not build an amp that has fully documented performance into proper loads and is based on solid engineering instead of myth? I strongly suspect the O2 will outperform the M3 in several areas.I am in that group buy alreadyYou can still get in on the O2 PCB buy
You cannot suspect that it will outperform M3 if you haven't done a comparison with your equipment
It is sad that they used quite a bit time in the design process of [SIZE=-1]M³[/SIZE] and it was not only Ti's work.
Last edited:
You're correct I can't be certain, I merely said I suspect the O2 has some measurable advantages. I say that, in part, because of everything in this thread, my 3 channel article, and my experience with the Mini3 and AMB's response to valid concerns raised by me and several others. As JCX and others have confirmed, splitting the grounds as they are in the M3, and using a virtual ground, can only hurt performance. It raises distortion, degrades crosstalk, and roughly doubles the output impedance. It can also create rather significant inter-channel distortion as seen in the Mini3.
It's entirely possible the M3 still performs well enough for its flaws to be inaudible. In which case it would very likely be indistinguishable from the less expensive O2 in a blind listening test.
It's entirely possible the M3 still performs well enough for its flaws to be inaudible. In which case it would very likely be indistinguishable from the less expensive O2 in a blind listening test.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.