I'm just trying to learn as much as possible. Suppose I would have to settle for an op amp that has 4x times the typical (2x times the maximum) input bias current of NJM4556. Do you think the extra DC offset would be a concern? How would you compensate without changing the op amp? Am i silly to think a resistor of equal value between feedback and ground would help in that specific case? Downsides?
How much do you want to spend? Since this is a very clean amp, you could get away with something very expensive.. Since the AMB Gamma2 is using the same case, I think they would look cute together, and it is a very good unit. It's using a 270X(or 290X I forget) for USB conversion though, so you are limited to 16/44.1. If you want to go higher, the EMU 0204 is a very good unit, does 24/192 via USB (on MAC and Linux anyway, I think Windows drivers may be available also), and is only about $129..
@hwfanatic There's no room for the four "bias compensation" resistors on the O2's PCB. I don't think even the worst case DC offset is an issue, and from what I've seen, you'll never get even close to the worst case value. One of the links in my last message here talks about the tradeoffs involved with the 40K resistor value. If you make it smaller the offset goes down but the -3 dB low frequency cut off moves up.
Single stage "Cmoy-like" designs are far more prone to DC offset issues--especially when operated at higher gains. The input bias offset is multiplied by the DC gain of the amplifier. So amps like the Grado RA1, Tangent designs, AMB Mini3, etc. will typically have larger offsets than the O2.
I'm not sure why some are obsessive about DC offset. Anything under 50 mV is harmless to any headphones I know of and very unlikely to change the sound quality. Certainly anything under 10 mV shouldn't affect the sound. The only downside I know of is when you plug/unplug sensitive headphones and there's DC offset you get some "jack noise" from the AC signal created by the contacts sliding together. But perhaps I'm missing something else?
Everything is a trade off. Certainly with DC servos, more stages, etc. you can further reduce DC offset. But that's best left for the "overkill" designs like The Wire. The whole idea with the O2 was to get everything that mattered well past the point of diminishing returns and stop there rather than make it needlessly bigger, more complex, more expensive, less reliable, etc.
Single stage "Cmoy-like" designs are far more prone to DC offset issues--especially when operated at higher gains. The input bias offset is multiplied by the DC gain of the amplifier. So amps like the Grado RA1, Tangent designs, AMB Mini3, etc. will typically have larger offsets than the O2.
I'm not sure why some are obsessive about DC offset. Anything under 50 mV is harmless to any headphones I know of and very unlikely to change the sound quality. Certainly anything under 10 mV shouldn't affect the sound. The only downside I know of is when you plug/unplug sensitive headphones and there's DC offset you get some "jack noise" from the AC signal created by the contacts sliding together. But perhaps I'm missing something else?
Everything is a trade off. Certainly with DC servos, more stages, etc. you can further reduce DC offset. But that's best left for the "overkill" designs like The Wire. The whole idea with the O2 was to get everything that mattered well past the point of diminishing returns and stop there rather than make it needlessly bigger, more complex, more expensive, less reliable, etc.
Interesting science and measurements to back it up, nice work!
I would like to build one to try it out. Are there any PCBs left?
But, you sure that it is good to parallel those NJM4556 even with load sharing resistors? Why not use a dedicated buffer, for example AD811 or BUF634...? Oh, they are a lot more expensive
I would like to build one to try it out. Are there any PCBs left?
But, you sure that it is good to parallel those NJM4556 even with load sharing resistors? Why not use a dedicated buffer, for example AD811 or BUF634...? Oh, they are a lot more expensive
Last edited:
But, you sure that it is good to parallel those NJM4556 even with load sharing resistors? Why not use a dedicated buffer, for example AD811 or BUF634...? Oh, they are a lot more expensive
I caught the smiley face, but just to be clear, op amps are paralleled all the time for various reasons. And, in this case, I've conducted lots of testing to make sure the parallel scheme works well. It's really the key to making a $0.60 op amp perform like a $10 battery unfriendly buffer for 99% of headphones.
Have you tested BUF634 (or newer high speed buffers) in op amp feedback loop? How would this compare to parallel NJM4556 in measurements?
It's a very common scheme and performs really well, there are many designs that use it. I've got a few headphone designs using those and I can't complain. But I don't have the equipment for proper measurements so I need to rely on the designers' posts.
Anyway, the quiescent current of those would eat your batteries in no time.
Anyway, the quiescent current of those would eat your batteries in no time.
No offense, but have you read the thread? I'm sure its been addressed before, either here or on his Blog. RocketScientist has done something pretty extraordinary, given the cost of then parts he has used. Like he said in his previous msg.. why use a $10 part when 2x 60 cent parts are more than capable of providing the current requirements of the majority of headphones
OK There is huge amount of text to read, so no
But... Could this mean that there is no room for improvement? It could be the end of innovation.
OK There is huge amount of text to read, so no
But... Could this mean that there is no room for improvement? It could be the end of innovation.
Not at all.. just a different design philosophy, IE, TheWire.. its using the LM49600 buffer.. but don't expect it to be battery powered.. at least not for long .
Also, start here: http://nwavguy.blogspot.com/2011/08/o2-summary.html he has thoroughly discussed the reasons behind the design (and the buffer idea has been mentioned..)
Part of his philosphy is also "Bang for the Buck" , doing the same job for less than 10% of the cost? I like that..
The Benchmark DAC-1 uses Buffers in it's headphone amp, and listeners have not been able to tell the difference between it and the O2...
Last edited:
I would like to see Ti Kan's objective response to this. He has contributed a lot to DIY audio community and seems to have also nice engineering skills and knowledge. He also has a lot of pride in his own designs. His newest creations are actually using real ground, for example the alpha20 line amplifier.
Have a look at the O2 measurements where I compare it to the Benchmark DAC1. The DAC1 uses the BUF634. While it can deliver more current into low impedance loads at lower distortion at high frequencies, I'm not aware of any headphones that need the extra current (or at least all but a very few fringe cases). And the 4556's high frequency distortion is arguably well past inaudible into any load.
If I were doing a more cost-no-object amp, that didn't have to run on battery power, I'd go with the National LME49600 (aka The Wire, National Reference Design, etc.). But for all but a very few headphones, I maintain it would only be better on paper. Otherwise I wouldn't have put up money for a blind listening challenge.
But what he doesn't have are the huevos to respond on a forum where RocketScientist can respond. He's only spoken out about the O2 over at HeadFi where he knows RocketScientist is banned and cannot respond directly.
se
Last edited:
He still might have a lot of PCBs in stock. He must get those sold first before he can give constructive ideas.
Hell, even criticism is fine. It would just be nice if he had the decency to criticize it where RocketScientist could respond directly.
se