[SinnerG]

So I'm contemplating building a CMOY of some sorts, but would most likely use the AD8620AR opamp I ordered about a week ago. I see it can draw up to 45mA per channel, +-30mA typical according to the DS.

I was also reading up on the virtual ground layouts using TLE2426, but they can sink around 20mA.

Then I saw the PenguinAmp ... this uses 2x TLE2426 for the virtual ground circuit. Why? How? One for each opamp in a dual package?

Yes, I'm a newb. Homework first. Soldering iron into eye later. I like a bit of safety margin so it didn't seem like the TLE2426 could pass enough current for the AD8620. As with all newbies though, I'm looking for something with the least amount of bits and pieces, yet a little up from the standard 2 resistor voltage splitter.

 

[tangent]

If you just replace the divider resistors in the CMoy with the TLE, a single TLE will do just fine. The return current from the headphones mostly gets soaked up by the rail caps, which are charged and discharged from the low-impedance power supply. The TLE is only there to keep the virtual ground pinned at 1/2 V+, which doesn't take much current.

You can instead rearrange the power caps so they are only from rail to rail, leaving the TLE "naked" on the virtual ground, in which case during heavy loads you can indeed overload it. I've tried that, and it can indeed make the amp sound bad when the load gets high.

 

[SinnerG]

Thanks Tangent.

As you can well see, you are dealing with an aging amateur here.

I need to continue re-schooling on basic analog electronics. I think what I might have been thinking was that a max'd out AD8620R would be pulling 45mA per channel right out the back-end of the TLE. I have a strange feeling I've got it all backwards most of the time.

I better keep reading up on info before I put a soldering iron into the wall socket. Otherwise there's going to be a newsflash of someone off'ing themselves with a mere 1pA.

I wonder why they did that on the PenguinAmp though?

 

[error401]

You could also substitute the TLE2426 with a buffered resistor divider if you really did need more output current. It adds a couple parts, but you can easily use a high current opamp to drive the virtual ground. See tangent's page on virtual grounds.

 

[SinnerG]

I saw that. Been through Tangent's pages over and over. Just highlights how little I know, being that I'm a novice.

It's a very good resource as well as the links attached to it. I just need to keep reading/refreshing on the basics before I properly understand it all. Otherwise we're going to be in trouble.

I'll go with the TLE for now as Tangent pointed out. Can't believe it is the only "ground reference" IC on places like digikey.

 

[SinnerG]

Hmmm...

Rail Splitter, from Abraham Lincoln to Virtual Ground - Maxim

Down to 0.035% accuracy between resistors. Wasn't that the whole issue with trying to match resistors for a good split?

 

[linuxworks]

I've seen the TLE splitter with a single cap across the +- rails as well as dual caps from the rail to ground.

I'm still not clear which choice you'd use and why. is there an easy answer?

I do worry about the low current ability of a TLE. is the TLE chip supposed to be 'for a single chip at a time' kind of thing? ie, each place you'd need a 'split' you'd have a local 3 pin TLE there? is THAT why they are 'only' 20mils?

 

[SinnerG]

It just dawned on me that the suggestion from Tangent is how the MINT amp is configured. To top it off, the MINT shows an AD8620 as well. Hmmm... perhaps me thinks this is one to look at for me.

 

[tangent]

Quote:

Originally Posted by linuxworks /img/forum/go_quote.gif I'm still not clear which choice you'd use and why.

If you stick caps across the TLE's outputs, as in the MINT, it becomes essentially a DC-only component. At AC, its output current limit prevents it from charging the caps faster than the load demands, so it's completely ineffectual. This means that the caps are not a perfect solution to too high a load -- if it gets high enough, the inability of the TLE to keep up will still cause virtual ground shifts, which in a simple topology like the MINT means distortion, since everything is grounded to this same point.

This is why we buffer the TLE in the PIMETA and PPA. The TLE then only has to deal with tiny dynamic currents, such as that from the op-amp feedback loops. It keeps the ground reference steady for the output ground buffer, which shunts the headphone return current to the power rails, where it's soaked up by the power caps. Much lower distortion results.

Quote:

is the TLE chip supposed to be 'for a single chip at a time' kind of thing?

You have to be very careful about paralleling any voltage source. They will never be equal in voltage, so their low output impedance means the higher voltage one constantly tries to "correct" the lower-voltage one. This is why the famous A47 amp has those 47-ohm resistors on its outputs: raise the impedance and this fighting becomes unimportant.

Quote:

is THAT why they are 'only' 20mils?

They're 20 mA because it's a complicated little circuit in a TO-92 package. Note that op-amps in bigger packages only go to about 40 mA typically; a TLE2426 is basically a resistor divider with an op-amp buffering its output.

They're successful because in most circuits, it's adequate.

Also keep in mind that there's apparently not so much demand for these ICs that there's competition. If there were, maybe we'd see some with different parameters.

 

[SinnerG]

Thanks again Tangent for your insightful comments on all of this. I'm sure you could have told us all to go read tons of other threads on diyaudio, etc.

Just the awesome stuff there gets way to complex for the beginners.

Quote:

Originally Posted by tangent /img/forum/go_quote.gif This means that the caps are not a perfect solution to too high a load -- if it gets high enough, the inability of the TLE to keep up will still cause virtual ground shifts, which in a simple topology like the MINT means distortion, since everything is grounded to this same point.

Would the most likely case of that happening be in the scenario of low impedance headphones? I currently just use A900s, which are efficient and low impedance. They really don't need much to drive them well.

Quote:

Originally Posted by tangent /img/forum/go_quote.gif This is why we buffer the TLE in the PIMETA and PPA. The TLE then only has to deal with tiny dynamic currents, such as that from the op-amp feedback loops. It keeps the ground reference steady for the output ground buffer, which shunts the headphone return current to the power rails, where it's soaked up by the power caps. Much lower distortion results.

And this leads to more complexity further down the line, right? You reckon one your schematics of TLE and opamp splitter on the virtual ground ideas page on your website would be fine to work with the rest of the MINT amplification section? The PIMETA is way more technical than I can handle right now.

 

[schubert]

If you're looking for nice ground circuits, I'd recommend the one used by Ben Feist for his A47 (link, also described elsewhere) or Sijosae's discrete splitter (usually on Headwize, but it's down; also here - link). I built the A47 as described on the site (sounds really nice with 470uF 16V Black Gates as power caps), and used the Sijosae with a CMOY based on the AD843. As a fellow newbie I can say that they're not difficult to build, for I made them and they both worked really well. Good luck!