[NelsonVandal]

Sorry for one more post, but is this hunting for current really relevant. If I use 300 Ohm HD650's and play at 1 - 2 V (I rarely play louder), isn't the current needed only 3 - 6 mA?

 

[holland]

Yes, you don't need much current to drive the HD650. Current is the primary concern for low impedence headphones.

How are you using the opamp to create the virtual ground? Are you using a resistor divider and the opamp as a buffer (this is basically the TLE2426, btw)? If so, you can probably use your other jisbos.

 

[NelsonVandal]

God damn it! Am I just kidding myself using opamps as virtual ground. I never really understood how ground, sinking and sourcing works. Is half the sound produced by the ground circuit? If it's so, it should be twice as good if I can get a JISBOS to work properly as active ground. That would be something. I think the sound was better with 3 JISBOS channels, but I had some weird offset drift depending on the length of the headphone cable.

If I split the batteries, will the batteries then affect the sound, or will there be no coloration at all from the ground channel? Should I parallell capacitors with the batteries from rails to ground? What would the ideal ground circuit look like when using batteries as power source?

If the railsplitter colors the sound, how does a TLE2426 sound? Everybody is using them. Should we try rolling railsplitters just as we roll opamps? There's an old 33201 opamp in Xenos OHA-REP as virtual ground. Is this why I never got it to sound good until I modded it to active ground with LMH6642?

 

[holland]

Quote:

Originally Posted by NelsonVandal /img/forum/go_quote.gif Am I just kidding myself using opamps as virtual ground. I never really understood how ground, sinking and sourcing works. Is half the sound produced by the ground circuit? If it's so, it should be twice as good if I can get a JISBOS to work properly as active ground. That would be something. I think the sound was better with 3 JISBOS channels, but I had some weird offset drift depending on the length of the headphone cable. If I split the batteries, will the batteries then affect the sound, or will there be no coloration at all from the ground channel? Should I parallell capacitors with the batteries from rails to ground? What would the ideal ground circuit look like when using batteries as power source? If the railsplitter colors the sound, how does a TLE2426 sound? Everybody is using them. Should we try rolling railsplitters just as we roll opamps? There's an old 33201 opamp in Xenos OHA-REP as virtual ground. Is this why I never got it to sound good until I modded it to active ground with LMH6642?

It's still not clear to me how you are using the opamp as the virtual ground. How are you splitting the voltage? You've got to have a resistor divider and the center point as the non-inverting input, right? How did you mod your xenos to "active" ground? Did you replace the opamp or add it after as a buffer.

The TLE2426 is a resistor divider + opamp in one package. Yes, there's a color. No, I have no idea what it is. I'm not sure of it's affect on sound. If you really want to build your own, you'd use a resistor divider with an opamp + buffer after it, or resistor divider + buffer.

Sinking and sourcing refers to the direction of the current. Sinking is the current flowing into the TLE2426 (for example). Sourcing is the current flowing out of. Just to confuse you. Electrons are negatively charged and flow in the opposite direction of the current.

I'm not a fan of splitting batteries, but as long as the batteries are fresh and from the same lot, it should mostly be OK in the middle of the charge span for a NiMH battery.

 

[NelsonVandal]

Quote:

Originally Posted by holland /img/forum/go_quote.gif It's still not clear to me how you are using the opamp as the virtual ground. How are you splitting the voltage? You've got to have a resistor divider and the center point as the non-inverting input, right? How did you mod your xenos to "active" ground? Did you replace the opamp or add it after as a buffer.

I'm using the TLE as a voltage splitter since it was already there. Connected to the (unity gain) opamp +input. All grounds are connected to the opamp output.

I'm angry at myself for not knowing the basics of electronics. I'm just after that golden sound.

About Xenos: I removed the LMH6642 used as buffers. An AD8066 is used unbuffered for L/R. Virtual ground goes to the input of one LMH6642, and the output is connected to phones = OG. The old connection from ground to phones is removed of course. It's just like a Mini3 in a Xenos box with that old opamp as rail splitter. This is a major improvement. I've modded that Xenos like crazy and this is the only setup I'm happy with. To my ears, and with HD650, it sounds better than Mini3 with AD8397/OPA690. To mess it up further, it's class A biased with JFET's (2N5486), 6 mA, but I'm not sure it improved the sound.

 

[amb]

Quote:

Originally Posted by NelsonVandal /img/forum/go_quote.gif I'm using the TLE as a voltage splitter since it was already there. Connected to the (unity gain) opamp +input. All grounds are connected to the opamp output.

While this gives you increased ground current capacity, you do not get the performance benefit of a 3-channel active ground topology because all ground are still tied together. You should connect all grounds to the TLE2426's output, except the headphone ground return, which would be connected to the opamp's output. This way, the headphone return current does not pollute signal ground. The signal ground would not have any significant current flow, so it could serve as a stable ground reference for the amp. Study the schematic of the Mini³ and M³ amps -- note the difference between IG and OG.

Oh, and yes, if you must use a 3-channel topology with rail splitter, then three JISBOS buffers would be best. If you run into headphone cable related issues, probably there is too much capacitance. Adding some ferrite beads to the outputs of each channel may be the cure.

 

[majkel]

OK, so this is how symmetry might be realized.

You have two double op-amps with identical (or close) current efficiency. Ground path (lower schematic) has to sink current from both channels (upper schematic). Assuming currents from both channels are identical, ground channel has to accept twice the current from the channel. Otherwise ground channel becomes a bottleneck when you want to drive power hungry headphones or listen very loud without distortions and compression of dynamics.

Some remarks here:

- there is a simple resistor voltage divider. Unlike the TLE2426, voltage divider has got no sonic signature.
- It's a closed loop ground channel so you cannot decouple any supply pins to the ground using capacitors, etc. You can only decouple the supplies by adding a capacitor between the rails.
- the circuit needs single supply, one battery or set of batteries with no center tap connected anywhere

Sorry for going off topic but you know, I promised...

 

[kvant]

Nelson,

I just wanted to add a cartoon to my previous post illustrating how the signal goes through the virtual-ground buffer. The thread drifted a bit, but since I already drew it...

From this picture it seems that quality of the buffer in the virtual ground (BUFG) is as important as quality of L/R channels (JISBOS here). However, there is some experimental evidence that BUFG might not be that critical. In the PIMETA, for instance, people say that stacking, i.e. improving, buffers on L&R channels helps the sound quality, while stacking ground buffer does not make any difference. Perhaps it has something to do with the fact that output of BUFG sits steadily at V/2 while output of the L/R buffer swings the signal voltage.

You probably saw the the following material on virtual grounds, but just in case you didn't...

1. Tangent's notes
2. thread on headwize

 

[NelsonVandal]

Thank you very much for your replies, Amb, Kvant, Majkel, Holland. I really appreciate it. I'm willing to learn. I've realised that the sonic impact of the ground channel is the same or almost the same as of L/R. Now I understand why my Sijosae AB amp with discrete rail splitter sounds so different from a similar amp with TLE.

I'm going to try the JISBOS buffer as active ground again, and add ferrites if needed. If there's still offset drift, I'll stick with it as rail splitter.

I'm going to try to use resistor deviders instead of TLE2426 to see if this improves SQ. I've seen your mosts on this before, Majkel, but I never uderstood how crucial this part can be. I'm going to try this in other amps like LISAIII clone and Mini3 as well to see if there's any difference. If it colors the sound sinking currents from the headphones, doesn't it color the sound sinking the small currents as well?

 

[amb]

Quote:

Originally Posted by NelsonVandal /img/forum/go_quote.gif I'm going to try to use resistor deviders instead of TLE2426 to see if this improves SQ. I've seen your mosts on this before, Majkel, but I never uderstood how crucial this part can be. I'm going to try this in other amps like LISAIII clone and Mini3 as well to see if there's any difference. If it colors the sound sinking currents from the headphones, doesn't it color the sound sinking the small currents as well?

Although you're free to do whatever you want with your projects, for the record I do not recommend replacing the TLE2426 in the Mini³ with a resistor divider.

The TLE2426 has far lower output impedance than any resistor divider (datasheet says 0.0075Ω typical). In the Mini³ it only serves as a superior and stable virtual ground voltage reference, and neither sources/sinks current nor swings voltage, thus it does not impart any sonic signature. It also consumes much less current (195µA at 12V) than a resistor divider, so it wastes less battery power.

In short, you will experience worse performance if you replace the TLE2426 with resistors in the Mini³. I didn't just throw parts into the circuit for no reason...

 

[TzeYang]

amb,

I'm having lots of troubles finding the input jfets for this buffer. Do you think i can substitute the complementary input jfets with the "usuall" 2n5457 and 2n5460 and eat the cake too?

Yes, the ratings are farrrr worse and they are not as robust as the Toshibas. Just wondering

thanks.

 

[amb]

The 2SK170/2SJ74 have higher transconductance and lower noise than 2N5457/2N5460, and are far better devices for the input stage.

I don't know why you're having trouble finding the Toshiba JFETs. I offer them in matched quads, which is perfect for two channels.

 

[majkel]

I agree. You cannot use voltage divider wherever you want. In my GC proposition there is a fully compensated feedback loop against both DC offset and high frequency oscillations. Also, most ground channels act as current pump for electrolytic capacitors. My GC is a real third channel, there are no electrolytics caps playing a role of output current bypasses. I think it's another advantage because you get rid of high influence of electrolytic capacitors on the sound. Regarding power consumption, 2x150k divider draws 30uA of current from a 9V battery so talking about power savings is unjustified here IMHO. Telling TLE2426 has no sound signature is also false. It is enough to refer to the datasheet and read that it's an op-amp with a built-in precision feedback loop and a voltage divider. So it must have a sound signature.

 

[TzeYang]

Quote:

Originally Posted by amb /img/forum/go_quote.gif The 2SK170/2SJ74 have higher transconductance and lower noise than 2N5457/2N5460, and are far better devices for the input stage. I don't know why you're having trouble finding the Toshiba JFETs. I offer them in matched quads, which is perfect for two channels.

I live in Malaysia. Shipping is a pain to pay for using my country's puny currency. I can't see myself spending any much more after splurging monies on the film caps and manly tools for case work recently. I might save up some more monies and buy around 10 pairs from you in near future i guess since i really do miss that kind of quality service you offer at your site.

While on the other hand, the itch of building a Jisbos is no longer tolerable LOL. So i was wondering if i can just use the cheaper 2N5457/5460 with far worse ratings instead temporary just for experiments.

 

[amb]

Quote:

Originally Posted by majkel /img/forum/go_quote.gif Also, most ground channels act as current pump for electrolytic capacitors. My GC is a real third channel, there are no electrolytics caps playing a role of output current bypasses. I think it's another advantage because you get rid of high influence of electrolytic capacitors on the sound

You are mixing up virtual ground rail-splitter duty with output ground channel. In the Mini³, M³, PPA or Pimeta amps, the two are separate (IG vs. OG). The output ground in these amps do not have electrolytic caps to the rails. The 3-channel β22 amp doesn't even use a virtual ground, but it still has an output ground channel. Each of these are just as "real" a third channel by your definition.

The cmoy amp is a notable exception, with a single virtual signal ground (whether implemented with a resistor divider or TLE2426) that also serve to source/sink the headphone return current. That is a main reason why cmoy is a lesser amp than the others I mention above.

Quote:

2x150k divider draws 30uA of current from a 9V battery so talking about power savings is unjustified here IMHO.

I was not directing the current consumption comparison specifically at your circuit, although your high resistor value of 150KΩ would be a significant source of noise. If I had to use a resistor divider, I'd choose something much less to minimize noise (e.g., The cmoy uses 4.7KΩ, which flows 1mA at 9V, and is more than 5x increase over the TLE2426).