[Audio-Me]

Comments and suggestions appreciated.

This is what I have in mind:

AD8610
Elna Cerafine 1000uf × 3 per rail
Wimas
Tantalum/Vishays
Stacked EL2002 buffers
Tangent's new board
Cascode current source to bias into Class A (FETs!? Cool!)
Lambda H-series linear psu 24V 1.2A - use this as a real ground?
Hammond steel enclosure w/ aluminum face plates

I'm not convinced the TI rail splitter and jung multiloop does anything for the sound, tangent, or anyone else experienced with it, please explain this to me? What improvements do they make if any? Do you recommend using them? Why or why not?

For the other parts, yes I will be getting some nice parts which are probably overkill, but I want no bottlenecks.

 

[eric343]

The TI railsplitter is required unless you have another source for a ground.

 

[andrzejpw]

just wondering . . . why will tangent's new board make the amp sound better?

 

[geom_tol]

If you're going to use an AC powersupply find a dual one, that is one with a + and - voltage. That ps will have a ground. So you won't need the railsplitter or buffer used in the meta42 ps.

The meta42 board is cool, but it was made to be very portable.

A non-portable board would probably use single opamps and separate the left and right channels completely.
The connections could be laid out so connectors can be soldered directly to the board. Eliminating wires and solder joints is always a good thing.
A real ground plane maybe?
Individual decoupling for each IC.

The Jung multiloop has some benefits, I haven't heard of any negatives from using it. And it makes the amps sound real good.

 

[aos]

I designed my PCB to have everything board-mount. As lots of stuff I make never get actually built in a box, having everything on the board and ready to move is a great plus. Also, there are no wires and no expense in that direction.

However, unless you have pre-drilled cases, board mount would be much harder to get installed in a case in an aesthetically pleasing way as you have basically no margin for error in shaping the holes. And if holes aren't perfect and small circles (and even if they are), you're very likely going to make a very ugly faceplate. Apheared used to yell at me for using square power switches...

But be careful with ground planes. In digital it's a must but with analog if you screw up - and unless you know what you're doing you certainly will - you can get instability rather than "better sound". For example you'll get parasitic capacitance and create positive feedback (oscillation). You have to remove ground around the opamp in a careful way. Some manufacturers will give you a PCB with proper layout if you order samples from them and there you can see how it looks like. In some prototypes I used ground loop that way but then I stopped bothering and stopped using ground plane for analog altogether. You might have easier time if you use low bandwidth opamps but with the latest craze that only ultrawide bandwidth opamps and buffers will do, these small parasitics will screw up stuff. Better to leave VHF circuit design to trained professionals.

In any case Audio&Me, you misunderstood some terms. Split rails and multiloop are essential parts of META42 design, not tweaks to get that last 1% (that would be stuff like current sources). If you don't want or need that, you may as well build a CMOY.

 

[Audio-Me]

Would you guys like to design me a home board?

aos - I have no problems with unprofessional looking stuff, as long as it is simple and non offensive looking. I like rocker switches btw.

ok, split rails and multiloop it is then.

geom, you can't just use a voltage splitter? Do you have any dual output linear ps recommendations?

 

[eric343]

A&M: My favorite so far is Newark part #89F7496. Medical grade linear goodness, with +/-12V, and +5V thrown in free

 

[geom_tol]

I wouldn't use a railsplitter with ac powersupply. A good dual supply shouldn't me that much more than a single, and you are going for the best possible right.

There was a thread about dual regulated ps just a few days ago. Aos also started a thread about making a very adaptable ps pcb where you can go crazy with super regulators etc etc.

Personally I wonder if it's really necessary to make these ultra silent ps for opamp designs, they have really good ps rejection you know.
I have no experience with this though, just my opinion.

 

[geom_tol]

Quote:

Originally posted by eric343 A&M: My favorite so far is Newark part #89F7496. Medical grade linear goodness, with +/-12V, and +5V thrown in free

Check that number. Don't get any matches on newark. I'm looking for something like that.

 

[erix]

Quote:

I'm not convinced the TI rail splitter and jung multiloop does anything for the sound

The rail splitter - maybe not with today's hi-perf opamps that are very flexible in their voltage requirements. You have to have positive voltage and negative voltage, however, so you might as well get them exactly the same with a part that is cheaper than two resistors.

Walter Jung's multi-loop - that *is* the sound and the whole point of the board.

It's all about stability. Why do you think so many people have had success with this design with wildly different resistor values? This design is very forgiving because of the loops.

It's about gain - in a single loop configuration you set your gain with two resistors. Typically in headphone amps it's between 5-11. Some opamps need higher gains than that to sound their best - some won't even operate below 20. In a multiloop you can set the gain of the opamp seperate from the gain of the whole amp. So you can have your opamp running at gain=200 but the whole amp is only running at gain=6.

If I was to make the uber-amp I would go with either an internal regulated plus/minus 15V power supply or one in another enclosure. You can use a big-ass toroid, move your fancy caps into the power supply section, use real regulators and still come out ahead $-wise of a medical-grade one.

ok,
erix

 

[Possum]

Maybe Audio&Me prefers the sound of a simple cmoy+buffer circuit? I don't think he's heard a multiloop configuration yet. His first ETA42 that he raved about did not have multiloop. It was a simple op amp with a buffer in the feedback loop, yet he raved about the sound in other threads. Does he want to stick to the sound of the ETA42?

 

[thomas]

eric, as i said in the other thread, i don't think these supplies you're recommending are linear. If it was linear, a 5V at several amp output could not possibly be "thrown in for free"; it would add significantly to the cost.

Quote:

Personally I wonder if it's really necessary to make these ultra silent ps for opamp designs, they have really good ps rejection you know.

exactly, what's more important is the nature of the noise as powersupply filter caps, regulators, and opamps PSSR are much less effective at very high frequencies. a cheep linear supply that puts out a small amount of 60/120hz ripple is much better than a switcher that puts out noise in the megahertz range

Quote:

If I was to make the uber-amp I would go with either an internal regulated plus/minus 15V power supply or one in another enclosure. You can use a big-ass toroid, move your fancy caps into the power supply section, use real regulators and still come out ahead $-wise of a medical-grade one.

I agree...

Quote:

The rail splitter - maybe not with today's hi-perf opamps that are very flexible in their voltage requirements. You have to have positive voltage and negative voltage, however, so you might as well get them exactly the same with a part that is cheaper than two resistors

I've always wondered about the rail splitter in the meta, i don't see the reason for using another piece of active circuitry when a pair of resistors does the exact same thing. a 1% or smaller (depending on the resistor) differnece in the ground potential should not make any difference in the sound/performance

 

[Audio-Me]

Possum, I'd like to stick with the sound of the ETA42, unless the multiloop will improve the sound for the better. Could someone tell me the sound differences?

I thought that the optimal voltage for AD8610 was +/- 13V, so shouldn't +/- 12V be better since it's closer to the optimal? Or is more juice always better than not having enough? Hehe

If it's not required, I'd like to leave out the rail splitter. So a new board, no ti chip, yes/no multiloop would be an uber eta42?

I wanted an internal linear regulated ps in the first place. Any recommendations? erix?

This is the case I had in mind, http://www.hammondmfg.com/dwg15.htm Or is that too small? All I know is, I want a rugged metal case.

geom, help me out.

aos, do you think you could design me a board, or use yours?

 

[Possum]

Here's a thread talking a little about voltages with the AD-8610:

http://www.head-fi.org/forums/showth...ad8610+voltage

 

[Audio-Me]

I like the idea of using (two?) big transformers with real regulator chips and moving the power section to a separate pcb. However, I'd like it to still be in one box.

(so this would be a diy power source? cool! just stick in a power cord to the iec jack right?)

Going above the spec voltage seems like a bad idea to me, I'd like to stay with +/- 12V.

I wish someone would tell/describe me in detail what effect the multiloop has on the sound. Unless it improves in a way that I'd like, I think I'll just do without it. Uber ETA42 it is then.

Still want the cascode current source Class A operation though.