[Syzygies]

I am experimenting with PIMETA output resistors, not to reduce hiss, but to convert my Etymotic 4p earphones to 4s. Many of us who bought the 4p did so to be able to use them directly with, say, an iPod. We waded through reviews and manufacturer literature describing how the 4s, with 100 ohm resistance, was vastly superior to the 4p, but needed amplification.

Something besides hiss is at play here. I met the Ety engineer, he's a smart guy.

Meanwhile, I can't find any links on our forum that discuss this. Can one plausibly fulfill the 4p-to-4s 75 ohm conversion inside the feedback loop? What's the difference? Can anyone summarize this old debate, with a modern eye to the 4s issue?

I caved on my request for socketed output resistors outside the feedback loop, in recent board design review threads. I now regret this. It's much easier to experiment with socketed resistors than with changing output jacks. Rewiring my output jacks was my one nightmare of an otherwise incident-free Pimeta build, and it continues to be an issue.

Here's what tangent said for the META42, which offered both options:

Quote:

Originally Posted by tangent If you are getting audible hiss at normal listening volumes with the source disconnected, put a 47 to 100*? resistor at R8 or R9. (This mainly happens with low-impedance headphones.) There are frequent arguments in the audio DIY forums about whether it's better to put the output resistor inside (R8) or outside (R9) the feedback loop, and I don't intend to try to settle the argument here. I suggest you initially jumper across both positions and only put in an output resistor if you have hiss problems.

 

[tangent]

Quote:

Can one plausibly fulfill the 4p-to-4s 75 ohm conversion inside the feedback loop?

No, it doesn't have the same effect as putting the 77 ohms outside the loop.

Quote:

What's the difference?

When the resistor is inside the loop, it is viewed as an error source, and feedback tries to correct it. The output impedance of the amp therefore remains low. All I can tell you beyond that is that this in an incomplete view: feedback cannot possibly be making it disappear entirely, or it would have no effect. I have guesses at the reason, but nothing I'd be willing to put my name on in public.

When it is outside the loop, the op-amp has no control over the resistor; it simply adds to the op-amp's output impedance. The mechanisms at work here are much clearer. What is happening is that the extra resistance changes the headphones' transfer function: the way current changes as voltage changes. You cannot mathematically predict how this function will change without having a complete model of the headphones, but in the case of the Ety 4P, someone's already done the engineering for you. Just add 77 or so ohms, and they look like 4Ses.

Even if you cannot predict the exact impact of adding series resistance, some things are clear. It increases the load impedance, which makes the drive stage happier; less direct distortion. Also, this always requires more gain to reach the same volume level, so that effectively pushes the noise floor down.

 

[setmenu]

I added sockets to the outputs of the buffers of my amps to allow for tuning
the output impedance of the amp.[outside the feedback loop, as Tangent says]

http://www6.head-fi.org/forums/attac...achmentid=4554


It works very nicely.
You could even experiment by attaching a pots with flyleads and fine tuning
the resistance until the sound suits your tastes.


Setmenu

 

[Syzygies]

Yup, adding an output resistor socket looks like how I'll go on this, too.

Quote:

Originally Posted by setmenu You could even experiment by attaching a pots with flyleads and fine tuning the resistance until the sound suits your tastes.

A six step "attenuator", 0 to 75 ohms in 15 ohm steps, would probably sound better, and be plenty of choice in practice. I like e.g. what 75 ohms does to my Grado SR80s, amazingly one gets tighter bass without losing the bass.

My sense is I'd like best some intermediate resistance, for my Ety 4p phones.

 

[n_maher]

not to be overly picky here, but the resistance difference between the 4ps and 4ses is 73ohm according to Etymotic.

Nate

 

[Syzygies]

75 ohms is an available value at Mouser, as Vishay-Dale RN55E75R0BB14, $1.02 each. The official Etymotic cables cost $65 for a reason, and my trial cables made with RadioShack metal film resistors sounded like absolute crap. If "boutique" resistors like Welwyn might sound even better, I'm there, but I don't have a theory I believe that tells me they would.

I love how tangent remembered I was off by 2 ohms, but flipped the sign. He notes this is approximate.

That said, the Ety 4p and 4s represent two poles, I can find aspects I don't like to each pole, and I'd bet a great many people would prefer various resistances in between. My next order will include a range of values between 0 and 75 ohms.

 

[looser101]

Syzygies,

You could always buy an adapter to try:

Buy:
http://www.fixup.net/products/
More Info:
http://www.fixup.net/tips/ety/ety.htm

Or you can DIY. See here for some suggestions on how to:

http://www.headphoneamp.co.kr/bbs/zb...esc=asc&no=131
http://www.headphoneamp.co.kr/bbs/zb...esc=asc&no=133
http://www.headphoneamp.co.kr/bbs/zb...esc=asc&no=123

You could use an extra opamp socket for the resistor socket and try different values. Once you are satisfied with your choice of resistor value then make your own adapter or add it inline with your pimeta outputs. Having it external could be of benefit if you want to use other headphones with your amp.

Cheers

 

[SnoopyRocks]

Quote:

Originally Posted by tangent No, it doesn't have the same effect as putting the 77 ohms outside the loop. When the resistor is inside the loop, it is viewed as an error source, and feedback tries to correct it. The output impedance of the amp therefore remains low. All I can tell you beyond that is that this in an incomplete view: feedback cannot possibly be making it disappear entirely, or it would have no effect. I have guesses at the reason, but nothing I'd be willing to put my name on in public.

It won't disappear, just get much smaller. I don't like the error source description though. To evaluate this, model the amplifier as a voltage controlled voltage source with series output resistance, Rout. Ground the positive input. Add a resitive voltage divider between the output and negative input. This gives you a noninverting OpAmp topology. Apply a test source (Vt with current It) to the output. Solve for Vt/It. This is the effective output impedance under linear conditions (no slew). Go through the math and you come up with

Rout_effective={[Ro/(1+Beta*Ao)]||[Rf/(1-Beta)]}
=Ro*Rf/[Ro*(1-Beta)+Rf*(1+Beta*Ao)]
~Ro/Beta/Ao
where
Beta=Ri/(Rf+Ri) -> the feedback factor (noninverting configuration)
Rf -> feedback resistor (from output to OpAmp negative input)
Ri -> resistor from Opamp negative input to ground
Ao -> open loop low frequency gain
Ro -> output impedance of the amplifier's last stage

We can observe that the amplifier output impedance is reduced by one plus the loop gain. Since the OpAmp inputs are virtual grounds, the overall output impedance is the reduced output impedance in parallel with the slightly modified feedback resistance. The end result is that the actual output impedance is miniscule when Ao is large and that a 77 ohm series resistor at the output inside the loop doesn't do much, just like you said. Your intuition was right and now you have the proof.

 

[tangent]

Quote:

A six step "attenuator", 0 to 75 ohms in 15 ohm steps

If you're going to do that, you might as well go up to the other common nonzero output impedance, 120 ohms.

I'm not sure I'd bother trying to go in linear steps, either. Log laws are everywhere in audio, so sticking to standard 5% resistor values to get a loggish curve should be fine. 0, 22, 33, 47, 68 (or 73 as a special case, if you like), and 120 would work. That last jump is a bit big. It should really be 100 ohms, but we're making exceptions here, right?

Quote:

I love how tangent remembered I was off by 2 ohms, but flipped the sign.

Actually, what I did was misremembered the impedance of 4Ps. I thought they were 23 ohms.

Quote:

To evaluate this...Go through the math and you come up with

Thanks for doing the analysis. I can't do that, yet. I've got some books, but I've hidden them under the bed, where all the monsters we haven't yet faced belong.

Quote:

~Ro/Beta/Ao

I don't see how that can be true. I could believe Ro*(Beta/Ao), because the effective output resistance reduction should decrease as Beta becomes larger, shouldn't it?

I did my own reduction from your first step and came up with Ro/(Ao*Beta), which I suppose is the same thing as what you came up with, as the numbers come out the same. If there's an error, then, it must be in that first equation.

 

[SnoopyRocks]

Quote:

Originally Posted by tangent ... Thanks for doing the analysis. I can't do that, yet. I've got some books, but I've hidden them under the bed, where all the monsters we haven't yet faced belong. I don't see how that can be true. I could believe Ro*(Beta/Ao), because the effective output resistance reduction should decrease as Beta becomes larger, shouldn't it? I did my own reduction from your first step and came up with Ro/(Ao*Beta), which I suppose is the same thing as what you came up with, as the numbers come out the same. If there's an error, then, it must be in that first equation.

The math is pretty simple in this case actually: KCL at the output and hardly any algebra. The tricky part is knowing how to model the problem appropriately.

We might be thinking different of things by Beta. I defined it to be the inverse of the low frequency closed loop gain which is Acl=1+Rf/Ri=(Ri+Rf)/Ri -> Beta=1/Acl. Then

Rout_effective~Ro/(Beta*Ao)=Ro*Acl/Ao

This says that the effective output impedance gets bigger as you increase the overall amplifier gain. Another way to look at it is to say that the available open loop gain is reduced. It is this reduced open loop gain that lowers the output impedance. Does that make more sense?

 

[Syzygies]

Ok, I should be packing for a trip, I moved some more classic rock onto my iPod, and ended ripping out the Ety 4p to 4s conversion inside my Pimeta. I salvaged the old jack for an external 4p-to-4s cable. I pray this is the last time I redo my outputs, one screw is already stripped, and I'm not even going in to change batteries!

Yo Yo Ma sounds incredible through this amp, AD843 op amps, Ety 4s canalphones. His recordings are impeccable.

However, the whole Ety line teeters on the edge of brittle in its handling of high frequencies, on any amp. Any of my friends who tried my 6i's and didn't like them, this was the reason. I'm reminded that the best grades of chewing tobacco include ground glass to aid absorbtion. I don't need ground glass in my ear.

If Jimi Hendrix can't sound great on 4s earphones, I'm a 4p man.

 

[tangent]

Quote:

We might be thinking different of things by Beta. I defined it to be the inverse of the low frequency closed loop gain

Ah! I knew Beta from reading about the original op-amp equations by Black, and I thought it was the gain, not the feedback division factor. Now your formula makes sense.

Quote:

This says that the effective output impedance gets bigger as you increase the overall amplifier gain.

Precisely what I expected, and what led me to question the formula.