[Jazper]

Quote:

Originally Posted by Dougigs With regard to ordering silver solder (mentioned by GregVDS): This isn't a big deal one way or another, but silver solder is actually slightly less conductive than good old 60/40. In order to get the 1 per cent silver (which really doesn't melt that easily), they need to drop the tin content and increase the lead content to lower the melting point. And lead is not as good a conductor as tin... So in order to get a tiny amount of silver, you're diminishing the electrical quality of the solder in general. The only place where you actually need silver solder is where you're soldering components (i.e. connectors) that have silver-plated contacts. Regular solder has a tendency to wick the silver plating off the contacts, and silver solder preserves it. Aside from that application, "regular" solder is actually slightly better electrically, and is easier to work with, and a lot cheaper (though as I said the differences aren't huge).

I've found that silver solder is slightly stronger than normal solder, but both work equally well in my experience.

 

[Dougigs]

Let me follow the above know-it-all answer with a question that betrays my larger ignorance....

I'm interested in building an M3 -- I have a Sheldon Stokes amp that I built years ago, and it's been the only line-powered headphone amp I've needed... But I'm excited by your design and especially by the great discussions on this thread.

But here's my question: Why wouldn't it be better to leave out the rail-splitter chip and use a real bipolar supply? I have one that I built for the SDS amp that uses Jung + and - regulators, and to my old-school mind this would be a more solid supply. It would have a genuine ground without a chip in the way... Now, I realize that your rail-splitter chip has a vanishingly low impedance (according to its datasheet), and while it can only handle 20ma, you're not using a ground for the power output (since it's buffered), so that probably isn't a problem. But still, I can see the benefits of rail-splitters for battery powered circuits but it isn't exactly clear to me why they'd be better in a line-powered amp than a conventional centre-tapped, dual-regulated supply.

But this is probably because my knowledge of such things is woefully out-of-date. The M3 appears to have been designed by some of Silicon Valley's best analog minds, using the latest power-supply design practices. So could one of you just spend a moment laying out the thinking behind the rail splitter?

Thanks.

 

[skyskraper]

im guessing that when designing it they would have looked to the builders who would be building it. familiarity with building single voltage supplies, plenty of commercially and diy ones available easily. its cheaper to use a single $2 part then to double the number of parts needed for the psu

oh and most wouldnt be using a jung supply so the tle keeps the ground tracked perfectly halfway between the two rails.

just my guesses, im sure amb will be along in moments.........

 

[Dougigs]

"will be along in moments..." -- easy for an Aussie to say. It's currently 3AM in northern California, home of the M3. And 11AM here in Blighty, and dinnertime in your neck of the woods. But there's no hurry.

And you've captured the crux of the question: Is the rail splitter being used to make it easier for builders (in which case I'll leave it out and go bipolar), or is it being used because it actually makes a better tracking regulator than a conventional +/- supply? (especially since this circuit doesn't actually *use* its ground much). I'm willing to entertain arguments either way.

 

[Jazper]

Quote:

Originally Posted by Dougigs

Doug, my M3 is positive jung driven (using P-A's JSR03 design) http://home.swipnet.se/~w-50719/hifi/jsr03/index.html

Works fine. (pic)
http://photobucket.com/albums/y219/K...t=IMG_0591.jpg

to quote amb's page:

Quote:

Precision virtual ground reference Rather than using a conventional split power supply with positive, ground and negative outputs, the M³ amplifier uses a single supply and internally synthesizes a virtual signal ground by using the TLE2426 precision rail splitter chip. This simplifies the power supply requirement, yet provides the advantage of a dual-tracking split supply without associated complexity. The result is improved common-mode rejection ratio (CMRR).

 

[skyskraper]

hah, he's been caught posting later then now

Quote:

Only a single supply is needed. It is not necessary to use a dual (or "split") power supply with positive and negative rails. The single supply is internally split on the M³ circuit board into separate rails, providing a virtual ground centered at one-half of the supply voltage. This simplifies the power supply requirement, provides better common-mode rejection ratio (CMRR), and gives the benefits of a dual-tracking power supply without the complexity.

seems to make brief points for both.

i guess most simple/cheaper dual supplies are simply 78XX and 79XX working together, but wouldn't track nicely like a tracking jung would.

im a fan of the simplicity of single voltage supplies with a rail splitter myself.

for you though, assuming your jung tracks tightly (im sure it would

) i could see why just using that would be the go.

 

[Jazper]

I don't see why you couldn't use a dual power supply.. pull the TLE and have the input ground in the centre pin hole, where the + and - inputs are put the +- power inputs from the supply.

I just don't see the point in going to the extent of having a crazy dual power supply when you can half of one and run a TLE chip in this instance and it'd potentially be better.

 

[Dougigs]

Wow, thanks for the pointer to that Swedish web site -- he's got a lot of nifty stuff. If I can just get my head around the kroner-euro-pound conversion rates, I'll order one of those tiny SMD Jung regulator PCBs -- They'd be an easy fit into a small enclosure. His inrush protector is nice, too (not for this -- for my voltage-isolator box, which uses a 1500VA toroid that currently is protected with a big ugly thermistor).

I'm assuming that the output topology of the M3 does not have a terribly spectacular PSRR, so it's worth our effort to be talking about fancy regulators for it... Again, I'm happy to be proven wrong.

 

[Jazper]

Quote:

Originally Posted by Dougigs Wow, thanks for the pointer to that Swedish web site -- he's got a lot of nifty stuff. If I can just get my head around the kroner-euro-pound conversion rates, I'll order one of those tiny SMD Jung regulator PCBs -- They'd be an easy fit into a small enclosure. His inrush protector is nice, too (not for this -- for my voltage-isolator box, which uses a 1500VA toroid that currently is protected with a big ugly thermistor). I'm assuming that the output topology of the M3 does not have a terribly spectacular PSRR, so it's worth our effort to be talking about fancy regulators for it... Again, I'm happy to be proven wrong.

Off topic though, P-A (peranders in this forum - the owner of the website) is really helpful, just read his instructions carefully.

ww.xe.com is an easy converter which you can use

The only main consideration with the regulator is the amount of current the M3 needs, which is about 300mA or so (more if you increase the FET bias, obviously). That being the case, be careful about the heatsinks you use on P-A's regulators, I'm running relatively small ones and they are VERY toasty..

Should also mention that increasing the capacitance on the output of the regulator, before the M3 has markedly improved response (to my ears), which I find a bit odd.

 

[Dougigs]

Ah yes, good point -- I'd be biasing it very heavily into class A since I'm using 32 ohm Grados, so I think an SMD regulator is out... better to get some big fins up. After all, portability is not the point with this unit.

I would like to see PSRR figures on the M3, since they could render this whole discussion somewhat moot.

 

[Jazper]

Quote:

Originally Posted by Dougigs Ah yes, good point -- I'd be biasing it very heavily into class A since I'm using 32 ohm Grados, so I think an SMD regulator is out... better to get some big fins up. After all, portability is not the point with this unit. I would like to see PSRR figures on the M3, since they could render this whole discussion somewhat moot.

I'm running 120mA bias and it's plenty for anything I've thrown at it.

(that makes it about 400mA overall draw)

read my thread about my M3 adventures here:
http://www6.head-fi.org/forums/showthread.php?t=123686

 

[amb]

Although I've never measured the PSRR of the entire amp, I believe it should be very good. Typically the PSRR of the amp is most adversely impacted at the input side, and here we use opamps with already high PSRR, and then added the capacitance multiplier to further isolate the opamp supply rails.

At the output side, the MOSFETs are biased by sourcing constant current out of the opamp with a pair of cascoded JFETs. Not only does this make the opamp operate in class A, it vastly improves the PSRR of the output stage over the original SDS Labs amp where the bias current flows from the positive supply rail to the negative supply rail through resistors.

Using the Tangent STEPS PSU (and locating the PSU power transformer far away from the amp board so there is no magnetic hum induction), the noise floor of the M³is very low and free of any spurious power-supply related noise. Thus, I see no advantage of swapping out the TLE2426 and going with a more complex dual power supply.

 

[GregVDS]

Ok, I read all the AMB website, and have quite right in mind the different ways to hook up grounds from input, output, Mains and so on.

My question is more related to cable shielding.

Is a simple twisted pair coming from RCA connectors on the backside to the front switch - crossfeed is sufficient (+- 1 foot, 30cm), or should I better opt for the shield-braided cable (like a microphone cable). Is here silver wire interesting, or normal copper is ok?

Again, form the output R, L and G, should I use two wires from ground to output R and L in order to have twisted pairs from pcb to neutrik jack?

And also, the output ground is ok for shielding? the input ground is ok for shielding too (since they are not the same)?

And last, Is there a minimal thickness for internal separation between M³ and PSU pcbs, in terms of aluminium foil I mean. I plan to build a very tight enclosure, to be able to have it with me at work, and come back home with. Hence, the psu and M³ will be very close. I was planning to separate them by an aluminium plyed foil (0.8mm thick), shielding the psu. The outer box will be in aluminium Alodined internally.

Looking at the dielectric constant of material, I saw Cotton is even better than teflon. Someone use cotton braid as isolant on copper or silver? Is it manageable? I read oxydized copper was rather poor conductor, but silver remains very good, even oxydized. Any comments?

All the best.

GregVDS

 

[Dougigs]

That's pretty persuasive. In fact, it seems tempting to try a completely unregulated (but filtered) supply, since between the capacitance multipliers, the rail splitter, the various current sources and the intrinsically low PSRR of the op amp front end, we've got vast amounts of regulation already... An unregulated supply (with, say, a good-sized transformer with 4% regulation, followed by a rectifier and a string of 1000 uf low-esr caps) would have a much lower impedance and maybe give the output stage more bang.

Has anyone tried this?

 

[GregVDS]

I should check if the one I bought is so, I was thinking it was lead-free.

WBT silver solder contains 4% of AG. Ag-solder is sonically useless?

All the best

Greg