[morsel]

Sovkiller, I already warned you in private email about derailing this thread. I am now taking moderator advice and putting you on my permanent IGNORE list, something I have never done before to anyone. Do not post here again. No posts referencing Sovkiller will be responded to.

 

[ppl]

Morsel wht is your issue with Sovekiller and might i ask how he is derailing this thred.

 

[aos]

Well, even if you could legally build it based on his circuit, I don't think it'd be right to do so. Anyhow, I can't tell others what to do, I just know what I'll do.
I can only recommend allowing for a crossfeed daughterboard mount. People
using crossfeed are minority anyway so it's just an inconvenience. You can
always build it on a protoboard, it's simple enough, and just add it to the amp in that form.

Also, even if it's "legal", you might run afoul of Head-Fi policies. Head-Fi
policies are much stricter than the law and as Jan is one of main sponsors
if he complained I'm sure this thread wo go poof (if it contained crossfeed). I'm actually quite surprised this thread was not brought on Headwize, as Morsel herself stated during ppa development days that she will do so because a lot of people protested due to percieved "commercial" status of PPA. Also, there'd be less "noise" on the thread. Though right now I'm probably making some myself.

By the way, MOSFET should be biased to the most linear region, and it should be more than the load will ever need to use. cameron made a good point
that ground channel might need to sink 2x.

10 Ohm might indeed be a bit much. I say experiment.

 

[aos]

Oh yeah, MOSFETs, as I recall it was tough to find actual complementary pair. I recall looking at IRF at that time and even those few they claimed were complementary, were in fact quite apart. Of course N and P devices differ and can never have exactly same parameters (due to difference in electron and hole mobility mostly, I think) but I recall Toshibas were much better matched. Just a thought, if you're going to go with IRF, it may be worthwhile to comb their offering and see if they have matched complementary pairs - I checked the datasheet and haven't seen them actually mentioning that these two you
specified are complementary. As their intention is to sell these for switching, they might not bother offering complementary pairs though.

 

[amb]

I know there are fans of single-ended fans out there, but I am not too keen on such a topology, just a personal preference. To me there is an elegance with push-pull. Going single-ended means either a rather large, hot running resistor (yuck), or a constant current source of some sort. The quoted article used a LM317 voltage regulator for that, and if I were to be forced to go single-ended I would instead find a MOSFET solution (a la the Borbely headphone amp), but that would require more parts. Ultimately it comes down to preference, but so far I am not convinced that there is any advantage with single-ended, measurably or sonically.

The source resistors are there as a form of local feedback (the MOSFETs are themselves operating as simple source followers) to equalize the matching between the N and P channel devices. Without them you'd need to hand-select and match the MOSFETs from a big handful on a curve tracer, and even then it won't be too close. Better matching means lower distortion. They also happen to be very convenient for setting the quiescent current, because you just measure the voltage drop and turn the Vbe multiplier pot. 10 ohms is a good value because you could read the voltage and translate that into mA in your head without a calculator. The resistors are within the global feedback loop, so they do not contribute significantly to the output impedance. In fact, I measured an actual output impedance of less than 0.1 ohm.

-Ti

 

[aos]

That's true, how else would you measure the bias? And matching transistors as well. You need resistors, and 10Ohm is a nice round value but there's nothing stopping the builder from going a bit lower, say 4.7Ohm.

I'm sure you'd find a lot of buyers for matched MOSFETs if you have the ability to match them on a curve tracer

.

 

[Sovkiller]

Quote:

Originally Posted by ppl Morsel wht is your issue with Sovekiller and might i ask how he is derailing this thred.

"You are with me, or against me" kind of an attitude right? Simple as that.....not a very polite attitude, for a very good professional and talented person, as her....but every single person is a separate world, modesty is a virtue not everybody have.....

 

[Sovkiller]

Quote:

Originally Posted by aos Well, even if you could legally build it based on his circuit, I don't think it'd be right to do so. Anyhow, I can't tell others what to do, I just know what I'll do. I can only recommend allowing for a crossfeed daughterboard mount. People using crossfeed are minority anyway so it's just an inconvenience. You can always build it on a protoboard, it's simple enough, and just add it to the amp in that form. Also, even if it's "legal", you might run afoul of Head-Fi policies. Head-Fi policies are much stricter than the law and as Jan is one of main sponsors if he complained I'm sure this thread wo go poof (if it contained crossfeed).

Please AOS do not misunderstood me, I'm not suggesting the use of such device against Jans willingness.....even when maybe it is legal, this won't be ethical, in my book neither, and I won't do it neither, but it will be a nice addition indeed, even when now I know that this is not a posibility to consider given some attitudes

End of the "Crossfeed" chapter, for me at least......

 

[morsel]

Aos: amb put a lot of effort into studying the IRF catalog. I did the same and called IRF tech support for additional advice and confirmation before we chose the IRFZ24N and IRF9Z34N. They are not officially designated as complimentary pairs, but we feel confident about our selection.

The source resistors are in the negative feedback loop, which makes the apparent output impedance really low, but it does not change the fact that the impedance before feedback is higher than it would be with smaller source resistors. Low output impedance before feedback correlates to good bass with low impedance headphones. I believe it is important to have the lowest possible distortion and output impedance before feedback. Unfortunately, as amb points out, lowering the source resistors may increase distortion before feedback.

We will test the output stage in open loop configuration (without feedback) using source resistors from 10 to 0 Ohms and let you know how it goes.

 

[aos]

Quote:

Aos: amb put a lot of effort into studying the IRF catalog. I did the same and called IRF tech support for additional advice and confirmation before we chose the IRFZ24N and IRF9Z34N. They are not officially designated as complimentary pairs, but we feel confident about our selection.

If you did the research already, then all is well.

 

[DCameronMauch]

AMB: I am totally following you about the resistors as local feedback. I had forgotten about that. Makes a lot of sense. Decreasing the value also decreasing the "matching" effect. And you are right, they should have minimal on the final output impedance. The mosfets are also inside the feedback loop, so the opamp should "linearize" them. But opamps, like everything else, are imperfect. It is always best to make the output stage as linear as possible, to make the opamp's job that much easier.

Now the single ended topology does not need any device matching. There is only one transistor driving. So there is that advantage. Free to test mosfets without matching p-type. The bias current should be similar, since this design looks to be keeping the push-pull topology class A mode all the time. Come to think of it, the single ended topology is essentially what you get when you use a current source to the negative rail with opamps. I have had the most linear results from IRF with the IRLIZ24N. 2nd and 3rd harmonics are about 4dB lower than the IRFZ24N before going into a feedback loop. But I don't think there is any matching p-type mosfet.

Attached is the same schematic with the current source expanded. Just to show a possible implementation. And that parts count can still be quite low. One power resistor, two small resistors, and one TL431. Not bad.

 

[morsel]

Cameron: Thanks for the nice drawings. It's tempting, but don't we lose part of our supply voltage range across the current source MOSFET this way?

 

[morsel]

One output stage, both inline and folded. Oops, I just realized the MOSFETs have to be on the same side of the heat sink to make the inline folks happy.

Not that it matters, as the linear space wasted with the inline configuration seems prohibitive. The smallest PAR-Metals case is 12×8×2" which is not tall enough for 2" heat sinks. It also comes in 12×8×3". 8" is not enough depth for 3×3" long channels side by side. Rotating the board 90° means the input stage is at the back of the case, not along the side, which means the volume control leads have to run back and forth or the pot must be rear mounted with a shaft extension, whereas if the input stage is along the side, the volume leads can be shorter.

 

[amb]

DCameronMauch, thanks for the schematic. That is roughly how I would do it if I were to go single ended as well. However, there is another consideration that makes push-pull attractive to me, is the fact that push-pull is twice as efficient as a current-sourced single-ended topology in a class A stage. This is an important practical concern that cannot be overlooked for an amp that is intended to be built by many people.

Nelson Pass describes this well in his class A 40W power amp design paper:

http://www.passdiy.com/pdf/a40.pdf

Although this is less of a problem with headphone amps than it is with speaker amps, still we are constrained by the amount of heat sinking, ventilation, and power supply dissipation too (assuming the use of voltage regulators on the output devices, which I'd like to do). All this also directly correlate with size and cost as well.

-Ti

 

[aos]

morsel, you're correct, you lose part of the available voltage swing on the current source (or the resistor which is a simpler version). That's actually where the loss of efficiency occurs - a lot of power is dissipated on the resistor (or current source).