[TzeYang]

Quote:

Originally Posted by Koyaan I. Sqatsi /img/forum/go_quote.gif Yes, it is. And a four pin XLR too! *whispering* Though perhaps you could introduce him to the plastic surgeon who did the nose job on yours. se

Quote:

M³ can use a simple TLE2426 rail splitter, because the varying return current from the headphones is sourced from or sunk to the power rails through a third amplifier stage. A precise, well-decoupled reference is still needed because the input signal is referenced to ground. In M4, the input signal return current passes through the feedback loop to the complementary input line, and so there is no ground current at all. In fact, M4 essentially has no ground reference. A network of 1M ohm resistors ensure that the floating power supply remains centered relative to the common mode voltage of the input, but that's all.

On a side note, I did not see you responding to this :|

 

[Steve Eddy]

se

 

[j4cbo]

Quote:

Originally Posted by TzeYang /img/forum/go_quote.gif It's basically m3 with a fully differential input and output, so sound wise they should be pretty similar. I'm more curious on the PCB layout since there are so many stages on the PCB. How did you do your grounding? Star ground or ground plane? What about the power traces? IIRC, Jung circuits are a pain in the ass to make them perfect.

Almost all of the actual amplifier components are on the top layer, so V+ and V- are both on the bottom layer. The power traces branch after the filter cap bank; +in and - run separately to the Jung regulator and the 7824/relay circuitry. The only connection between signal and power is through the 1M resistors, which are located near the amplifier's input pins.

I haven't done any line/load step response testing on the Jung regulator... that's up next, I guess. SMT makes a tight, clean, elegant layout way easier than through-hole.

 

[j4cbo]

Quote:

Originally Posted by Koyaan I. Sqatsi /img/forum/go_quote.gif The load current still flows through ground. As I've said before, it doesn't just hit the power rails and then magically disappear.

What is "ground"?

Most amplifiers have a pair of power supplies, whether they be tracking or otherwise; one bunch of parts has the job of keeping V+ however many volts above ground, and another bunch of parts has to keep V- some ways below ground. Both the M³ and the M4 are exceptions to this; there's only a single voltage regulator driving V+ with respect to V-.

Then, generally, you have a bank of capacitors from V+ to ground and from ground to V-. They can source current into the rails as well, and yes, that current flows through the ground node. The M³ is a bit different; there are no capacitors from V+ to ground or V- to ground, only from V+ to V-. There are, however, caps between Vop+ and ground, and Vop- and ground. The M4 doesn't even have those. There really isn't a path for current to go through the load, into one rail, through ground, and back to the other end of the load. If there is, please point it out to me.

The only node in M4 which could accurately be called "ground", I think, is the one connected to R39 through R44. I'd have a hard time believing that any significant current flows through it.

This all gave me a hell of a time trying to hook up a 'scope to the damn thing, for what it's worth - my 'scope's inputs all have a DC path to ground, so I couldn't connect the scope's ground pin anywhere useful without causing current to flow through the earth pins in the wall.

Quote:

What the M4's input is referencing is a voltage (half the supply voltage). But that voltage is referenced to ground. And the M4's input has a DC path to ground. So it would still be susceptible to picking up IR drops due to load currents if proper grounding practices aren't used.

The input probably has a DC path to ground, but not necessarily - if the source has separate drivers for the noninverted and inverted output pins on the XLRs then yes, but if it's transformer-coupled, perhaps not. In some sense, yes, everything is "referenced to ground" - but nothing in the M4 cares.

 

[FallenAngel]

So... where do I get my hands on a board?

 

[smeggy]

you get in line

 

[amb]

Quote:

Originally Posted by j4cbo /img/forum/go_quote.gif ... This all gave me a hell of a time trying to hook up a 'scope to the damn thing, for what it's worth - my 'scope's inputs all have a DC path to ground, so I couldn't connect the scope's ground pin anywhere useful without causing current to flow through the earth pins in the wall.

This is a generic "problem" with using a scope to test balanced/differential gear. The need to "float" the AC earth ground from the scope is fraught with safety and measurement issues. Tektronix has a couple of good article about this:
Tektronix: Technical Brief > Floating Oscilloscope Measurements ... And Operator Protection
The Three Facets of "Floating" Measurement Solutions > Application Notes and Technical Documents : Tektronix

 

[mattcalf]

Oh god, just as I attempt to settle on the M3. This design comes along.

 

[00940]

Very nice idea and an apparently perfect implementation. More economical than building bridged amplifiers. The opa1632/ths4131 opamps are neat little things, with a decent quiescent current. How hot do they run in your build ?

Power dissipation could be reduced by the use of BJT instead of mosfet for a given power output but to each one his tastes

My only quibble is the use of smt relays... you dont' save place (at the contrary) and they are annoying to solder.

 

[johnwmclean]

j4cbo, congrats that really is a brilliant accomplishment, you must be very proud! I’m staying tuned for listening impressions.

 

[j4cbo]

Quote:

Originally Posted by 00940 /img/forum/go_quote.gif Very nice idea and an apparently perfect implementation. More economical than building bridged amplifiers. The opa1632/ths4131 opamps are neat little things, with a decent quiescent current. How hot do they run in your build ?

Toasty. It's a bit worse than it would otherwise be since they're in Class A (the output stage is driven with 3 mA per phase) but that's still not much compared to the quiescent current.

I calculated the dissipation at about 450 mW per chip. The datasheet claims the SOIC-8 can dissipate about a watt 25 °C, so I'm not worried.

Quote:

Power dissipation could be reduced by the use of BJT instead of mosfet for a given power output but to each one his tastes My only quibble is the use of smt relays... you dont' save place (at the contrary) and they are annoying to solder.

The space used is about the same, but I hate dealing with flipping over the board without having all the components fall out. Plus, this would allow someone to use the hotplate reflow technique to solder almost everything at once. There are a few parts on the back which would have to be done by hand, but not many.

 

[Steve Eddy]

se

 

[j4cbo]

Quote:

Originally Posted by Koyaan I. Sqatsi /img/forum/go_quote.gif Yes, there's only a single voltage regulator because there is only a single supply, the bottom of which is "ground" and variously labeled "-UB" and "V-" in your schematic.

That is not correct. -UB and V- are not the same node. Look at the schematic.

For instance, the varying current through the opamp and Vbe multiplier in the output stage near the top of the schematic passes generally through the following components and nodes:
- From the negative end of C5/C6/the capacitance multiplier through V- into the constant current source made of Q1-Q2
- Through Q22 (mostly, but also the resistor divider)
- Through R9 into the opamp
- Out the positive supply of the opamp into V+
- Back to the other end of the capacitance multiplier and filter capacitors.

Quote:

All of your P channel MOSFETs have their drains tied to ground, so you tell me.

No, they don't, they have their drains tied to -UB. Why don't you try building one of these and connecting a wire from -UB to earth ground, and let me know how well that works out for you?

 

[qusp]

wow, this could be just what I need, bookmarked!!

nice work!!

 

[grawk]

I just deleted a bunch of posts trying to get this thread back on course. Take the ground debate to the other thread.