[P701]

building audio electronics can be its own hobby - but really the 1st question should be for what headphone - they vary by orders of magnitude in required power for the same SPL, have different impedances, the ratio of I to V,  this changes amp/buffer device and power supply choices

 

I'm using Sennheiser 598's (50 Ohm). Well generally, I guess those would go quite fine without a dedicated amp for them, but I've understood that headphones behave better when driven with low output-impedance devices, and also better current capability can further improve low-impedance headphones.

The basic CMoy I built to see if it'd improve the sound at all, seemed as if the soundstage "opened" a bit more on mid-high frequencies.
Can't say if there were any benefits in the lower frequencies, but I think it didn't sound that "muddy" while using CMoy, but still as if there was something missing, as if the lower freq was quieter or being cut-off a bit.
I used 0.68uF coupling caps with 200k resistor in input's lowpass, so it cant be due to it. (also 470uF in power bypass). Also ran some RMAA tests with the CMoy, and all seemed fine.
I was thinking if it's due to using resistor divider for vground, I changed 10k resistors to 2k, and it (possibly) improved it a bit.
So for now, I'v thought of improving some things and building an amp with buffer on vground, and also on the outputs within op-amp's feedback loop. And class-a bias mod.
Or would that be just unnecessary for the 598's?

 

[jcx]

@ 112 dB/Vrms the Senn 598 should be fine in many cases for listening from DAP which usually manage about 1 Vrms output, have internal op amps intended to drive even lower Z loads
 
 
some might aspire to having an amp with the dynamic headroom to hit rare 120 dB SPL peaks without amplifier clipping when rocking out at live levels of jazz club or even some unamplified Classical
 
http://headwize.com/?page_id=266 is good for both how low you should listen for all day and how loud the dynamic peaks of real live music can require
 
with the 598 for 120 dB SPL you need >~3.5 V, 70 mA peak amplitude - the peak current is beyond many standard DIP through hole packaged audio op amps, even 2x paralleled
 
the O2 headphone amp project blog has a section on op amp selection - he ended with NJM/JRC4556 and paralleled them
 
Class A and portable battery operation aren't a good combination: 2x 70 mA SE will drain 200-250 mA NiMH in <2  hours, "optimum push-pull" cuts the required standing bias by half - still looking at ~ 3 hrs battery life
 
biasing the output to be Class A for a smaller fraction of the full output could be an OK compromise - only 7 mA each ch would get you to 100 dB SPL in Class A
 
 
the O2 uses dual battery "passive ground" - buffer/active ground is useful for splitting a single battery V but doubles the headphone current drain on the battery
 
active/virtual gnds are seldom an improvement - rather a necessary evil if you want to avoid DC blocking capacitors at input and output - the "3-channel" fad was technically unjustified on most claims

 

[P701]

Thanks for the valuable info!

Hmm... One thing came to my mind that I started wondering now tho.
If I used a single-supply wall powersupply (24V) for the amp instead of batteries, would it actually work with the op-amps and buffers the same way as with batteries? Since they are specified to use positive and negative supplyvoltage, and now I have +24V and 0V?
Eventhough I have read that op-amps only care about the voltage difference that you use to supply them with, but this just came to my concern. And how about buffers then?

Also, if it will work, would the virtualground be now unnecessary, and couldnt I just use the power supply 0V as a ground?

Heres how the schematic would look like at the moment (power part only):
http://i.imgur.com/QvLz5zf.png

EDIT: Further discovering, op amps will work with single supply, and I would still need the virtual ground as the ground reference for them as I already had.
But how about buffers? Does anyone know if they support single-supply voltage (24V and 0V)?

 

[wakibaki]

Your diagram is FUBAR.
 
You have negative power connections to the opamps, current sink and buffer but no negative rail. You have an output but no signal input. The rail-splitter cannot source/sink sufficient current to supply 2 channels of typical audio opamp, never mind running them in class A + the buffer, even if the diagram wasn't full of errors. 
 
Stop trying to walk before you can run and settle down and complete a design that you know is within your capabilities. Learn how to bias an opamp for single supply and AC couple it. Once you have understood how and why to do this, plenty of time to move on to virtual grounds, class A, buffers and multiloop feedback.
 
w

 

[P701]

Your diagram is FUBAR.


 


You have negative power connections to the opamps, current sink and buffer but no negative rail. You have an output but no signal input. The rail-splitter cannot source/sink sufficient current to supply 2 channels of typical audio opamp, never mind running them in class A + the buffer, even if the diagram wasn't full of errors. 


 


Stop trying to walk before you can run and settle down and complete a design that you know is within your capabilities. Learn how to bias an opamp for single supply and AC couple it. Once you have understood how and why to do this, plenty of time to move on to virtual grounds, class A, buffers and multiloop feedback.


 


w

 

Maybe the schematic was somewhat unclear, or not sure if I just "cant".

You are wondering about the negative connections...
The negative rails on opamp and buffer (V- & VDC-) were just left there as I draw the component with ExpressSCH, they would connect to 0V rail. Same with the pin near JFET transistor (V-).
OP Amps will work with singlesupply, as they only care about voltage difference.

Wouldn't the input (half the supply voltage = "vground") be already created with TLE2426 rail splitter, which will output it to VO pin of it?
Isn't the buffer (along with op-amp) handling the current? With TLE2426's VO just setting the desired voltage level?

Here's where I looked for references, and I don't really understand why wouldn't the vground work in my schematic, if it would here http://tangentsoft.net/elec/vgrounds.html At the very end of the page. (as long as you can supply buffers with singlesupply voltage, like opamps).

And here for Class-a mod http://tangentsoft.net/audio/opamp-bias.html but maybe that wont work when using singlesupply voltage?

EDIT: correcting some mistakes..
How about this? http://i.imgur.com/08T9qkF.png Or should I stop...

 

[P701]

Above done also with LTsPice. 40Hz sine wave generator. R1, R2 demonstrating TLE2426.

It would work for as long as the PSU 0V is isolated from the ground of the audio input's ground?

 

[Avro_Arrow]

Any amplifier is only as good as it's power supply.
 
With that in mind, lets set the order of preferred power
supply architecture.
 
For a desktop amp.
1. Dual rail power supply using a center tap transformer or dual secondary transformer.
2. Dual rail power supply using a voltage doubler rectifier and single secondary transformer.
3. Single rail power supply with some form of rail splitter or ground channel.
 
For portable use.
1. Center tapped battery pack.
2. Battery pack with some form of rail splitter or ground channel.
 
For class A bias mod...
The current sink goes in the left and right channels, but not the ground channel.
 
Edit:
 
Doh!
Of course, you can exchange the rail splitter or ground channel for an output capacitor.
 
more Edit:
Facepalm!
Nikongod will point out you can just use a transformer....

 

[wakibaki]

OK, I see what you are doing now, but it's hardly surprising that I misunderstood the first time around, since there s no good reason on Earth to have an opamp biassed into class A interposed between the rail-splitter and the buffer. The purpose of biassing an opaamp ino class A is to obviate crossover distortion, but since you are driving the opamp with a DC level there is no "crossing over" going on. The rail-splitter can anyway provide plenty of current to drive the buffer, so the opamp is doubly redundant.

If you are determined to use a buffered virtual ground, use a buffer for each channel's ground but drive both direct from a single rail-splitter.

 

[wakibaki]

 
edit: Sitting thinking about this, a possible problem arises with dual channels when the 2 ground systems are connected together. It's possible that some voltage offset could arise due to differences in the buffers... everything might be OK, but everything might not. With very little resistance in the system a not inconsiderable current might flow buffer-buffer. Which would make it desirable to build a single discrete buffer with sufficient current handling capacity for both channels, or introduce some resistance in the ground-ground link, which is again not 100% desirable. The alternative is to have totally isolated channels in the HP and source, but this would require further work to guarantee that. It just goes to show how problems can creep in unnoticed as systems get more complex.
 
This is just one reason why virtual ground systems are, IMO, best avoided. Run the whole thing from 2 batteries or a dual-rail supply and that particular problem just goes away.
 
w

 

[P701]

Ah, didn't realize that the class-a bias mod is not needed for the ground channel. I just had seen it done on some other amps.

OK, I see what you are doing now, but it's hardly surprising that I misunderstood the first time around, since there s no good reason on Earth to have an opamp biassed into class A interposed between the rail-splitter and the buffer. The purpose of biassing an opaamp ino class A is to obviate crossover distortion, but since you are driving the opamp with a DC level there is no "crossing over" going on. The rail-splitter can anyway provide plenty of current to drive the buffer, so the opamp is doubly redundant.



If you are determined to use a buffered virtual ground, use a buffer for each channel's ground but drive both direct from a single rail-splitter.

 

Yeah, my bad. The schematic was actually somewhat confusing back then.
I had an understanding that when buffer is used inside op-amps feedback loop the op-amp would correct any offset in output that might be introduced by the buffer.

 


edit: Sitting thinking about this, a possible problem arises with dual channels when the 2 ground systems are connected together. It's possible that some voltage offset could arise due to differences in the buffers... everything might be OK, but everything might not. With very little resistance in the system a not inconsiderable current might flow buffer-buffer. Which would make it desirable to build a single discrete buffer with sufficient current handling capacity for both channels, or introduce some resistance in the ground-ground link, which is again not 100% desirable. The alternative is to have totally isolated channels in the HP and source, but this would require further work to guarantee that. It just goes to show how problems can creep in unnoticed as systems get more complex.


 


This is just one reason why virtual ground systems are, IMO, best avoided. Run the whole thing from 2 batteries or a dual-rail supply and that particular problem just goes away.


 


w

 

Do you mean discrete ground channels for Left and Right?
I actually had that in mind when I started wondering how I could "improve" from the basic CMoy, but realizing that wouldn't it be just the same thing as having 2 parallel buffers in one ground channel. Since, the ground channel is same for left and right anyways.

 

[wakibaki]

I mean a single discrete buffer for the ground channel. You can't parallel parallel the IC buffers, the LME49600 (basically the same as the BUF634) has a worst-case output offset of +/- 60mV, you could double that if you paralleled them and one buffer would just be driving current into the other. Suppose you joined the 2 ground systems with a 1 ohm resistor from each, that would give you 60mA. That's about a quarter of the available current. Of course, it could be a lot less than that, but it's all just turning into the kind of ugly kludge that most people would rather avoid designing into their amplifiers.

So now you need a discrete buffer with 1/2 amp+ current capacity

 

[00940]

So now you need a discrete buffer with 1/2 amp+ current capacity

 
A bit of an overkill... 150ma should be enough for the hd598 to get a decent virtual ground (I fully agree with JCX's figures). 250ma is very conservative.
 
One LME49600 in each channel and one LME49600 to buffer the ground would be more than enough.
 
Here's a suggestion. Obviously, the very basic buffers can be replaced by buf634/LME49600. All grounds return to vg (virtual ground). All inputs/output must be isolated, especially power in. Decoupling caps must be added near the opamps too.
 

 
 
 
To be honest... as soon as you try to "improve" the cmoy, you get into designs which are a lot more complex. Which means a lot of chances to mess the layout and get bad/suboptimal results. If you're not doing it for the fun of it, I'd just buy a O2 kit.

 

[wakibaki]

   
A bit of an overkill... 150ma should be enough for the hd598 to get a decent virtual ground (I fully agree with JCX's figures). 250ma is very conservative.
 

 
Overkill, yes. Very conservative, yes. That's where the OP has been all through this thread. 
 
w