Hifiman HE-6

When you increase load impedance, then voltage swing dominates, and typically, power draw drops so at least it's reasonable to assume that the amp won't have any problem driving the higher load, yeah. But still, expect current draw to still be significant enough that most headphone amps (especially tubes) won't be able to deliver. If you look up specs for op amp parts that are present in most lower-end to some mid-fi headphone amps, you'll find they distort as soon as you reach 30mA, which is barely enough to hold 1.5V steady with the HE-6. Buffers get closer to 200mA and can hold steady up to 10V, but I'd say... it's seriously not a good idea to push the boundaries of an amp. So a speaker amp with some headroom to spare is still a better idea overall.

And ya, HE-6 is probably the hardest non e-stat headphone to drive in existence IMO. I made the mistake of thinking they were not that hard to drive myself, until I realized that people were modding it with ear pads and fuzzors that likely drop a lot of sensitivity in certain parts. The end result is a headphone with which they'd then pump moar and moar powah through in order to get the mids to their listening levels. It's truly kinda crazy what people are doing with this thing.

On that note, though, it's still true that the stock HE-6, in all its "sound kinda like HE-560 but more better" form, isn't really that much harder to drive (nor does it sound that much different IMO). But almost every single HE-6 owner will tell you to mod it. I concede that it does sound better after modding, but I can't help but wonder if a part of the enhancement is also due to the amp itself, since amping requirements jumped up so high.

 

Maybe my I'm misunderstanding the way amplifiers work but I have been under the assumption that an amp doesn't work harder at higher volume. In other words I've been told that an amp runs "at 100%" all the time, meaning all the voltage reaches the headphone jack (or speaker jacks) before being choked by the potentiometer (volume knob). I suppose it's possible for the potentiometer to sit a multiple locations in the amp and accomplish a similar result but that's neither here nor there. Is this a wrong assumption? Do topologies (er... classes?) change the way this works?

I think it's immediately intuitive to assume an amp works much harder to play at higher volumes (to allow more voltage through the jacks), but the above assumption says otherwise. If the above assumption is correct, I'm not sure how one could push the boundaries of an amp. Once you take the pot out of the equation (zero resistance, max volume), does the amp care?

And, yeah, holy cow this headphone is modded to death. It's quite sad, really (but at least I got to make a thread out of it, hah).

I really appreciate the long replies, Bill. Good discussions going on here

 

Ive tried fuzzor with audeze pads vs without fuzzor.
Fuzzor is overhyped, I guess for those who have bad amplifier fuzzor will make it better as mentioned.

 

Actually, amps do work harder at higher volumes. This is how I understand it:

Essentially, an amp works by taking an input signal (say, from the DAC?) then apply a "gain" (basically a "multiplier") on that signal, and then push the "gained" (or "multiplied") signal to the headphone. The secondary task is for the amp to hold that gained signal stable/steady, which may actually be the reason why some amps sound different, but I think that's for a different discussion.

So however much voltage/current the amp is delivering to the headphone depends on the input signal. Say, if the amp is connected to a DAC that outputs a sine wave at 2Vrms, and assuming the "gain" of the amp is set at 2x, then at 50% volume knob, the signal the headphone will see is 2Vrms, and at 100% volume knob, the headphone will see 4Vrms. That is pretty straightforward.

However, it is entirely possible for the amp to "clip" when the gained signal exceeds power supply capabilities. For instance, if your power supply feeds the amp 20Vdc (that's +10V and -10V relative to Ground), and assuming transistors in the circuit are perfect (they aren't... so expect 75% efficiency or even lower), then you're looking at roughly, say... 14Vrms realistically. Assuming input signal is 2Vrms, that means the amp can only handle a maximum gain of 7x. Anything higher than that is just asking for troubles. And in fact, based on the efficiency issue mentioned, that means max gain that can be safely assumed for that amp is 5x. Fun note: NwAvGuy's O2 has a gain of 6.5x, but obviously the power supply isn't up to par, so the amp itself has gain issues, and this is noted in his blog as well, so he kinda admitted his amp had limitations but people don't seem to read the entire thing?

So is it totally cool then, if we design a power supply that does way above our gain requirements? Say... 40Vrms? Yeah, that's possible. But then you'll be combatting heat, stability, linearity, noise, distortion, and other issues that may ensue because the power supply may work cool at constant 40Vrms drawn, but it may throw fits if your amp is at lowest gain and drawing like... 0.1Vrms for IEM. That's about 0.25% of the maximum output, so the power supply need to be able to stay cool in terms of noise and distortion at roughly less than 0.1% of that figure for things to not make a significant difference (audibility is still a different matter altogether), so... you're really looking at 0.00025% overall noise and distortion requirement just for the power supply, and you're also kinda hoping the amp section doesn't contribute anymore to that. Urg... suddenly, things don't look so hot anymore, eh? Cue the quoted number games by manufacturers here. Fun note again: some IEMs need much less than 0.1Vrms to reach 90dB. For instance, Apple's in-ear headphone needs like 0.035Vrms to reach 90dB, so...

All of this is just to say... it's not easy (if not downright impossible?) to design an amp that can take on IEMs and scale all the way to the levels of the HE-6 and still maintain composure against all kinds of other parameters like input signal levels, and so on. Because of limitations of measurement tools (seriously, the numbers will only get super small), I think it's more reasonable to purposefully build amps that only fit some requirements, rather than trying to please everything and everyone. Now, this is why it's totally okay for something like the EC Studio to have some hiss and noise with IEMs. And it's also reasonable to assume that most amps that will play nice with your IEMs will not necessarily play nice with your HE-6.

And you're welcome, dude. I used to ponder all this stuff about amps until I tried to design my own and realized what insanity I got myself into. Maybe I've misunderstood a few things here and there but at least the rule of thumb for me now is to always consider the power supply and all of its distortion, noise, etc... stuffs first, before jumping into the rest of the design. In this way, though I kinda don't want to admit it (trust me, I butt heads with HE-6 people all day long regarding their "big" amps), it's true that a bigger amp with a bigger power supply that has better regulations and better specs can technically deal with the HE-6 better. Hence... speaker amps.

 

Use my felt ring instead.
http://1drv.ms/1GlXKGr

 

@Bill-P helps you appreciate the Ragnarok more when you explain all of that.

 

Bill: I got super learned from your post, thanks a ton for the good info. I can't imagine the struggles of designing your own amp. I think I would need a different degree for that.

I saw a pic of this over at HF, I think. Did you make this because you removed the dust cover from the vegan pad?

 

I'm still using the dust cover but I can try with a pair of pads with no dust cover. This is my replacement for the fuzzor which doesn't reduce clarity but minimizes reflections.

 

Oh, okay. I think I will better understand once my vegan pads show up.

 

out of the box thinking..

Only true when you have a linear volpot.
Most amps have a log type volpot so expect around 10% of the input voltage X gain.
So a 2x gain amp with the volpot set to 50% will have around 0.4V on its output when a 2V input signal is applied.

Actually the max output voltage depends on the topology of the output stage.
It is possible to design an output stage that almost reaches the power supply rails.
In that case an output voltage of +/- 9.5V is quite possible.
This translates to 6.7V rms, not 14Vrms

Not issues, just practical input voltage limits that have to be observed.
The high gain setting is NOT intended for use with 2Vrms sources.

That is assuming the circuit has 0dB power supply rejection.
Most designs are wayyyy better than that.

That is something I can agree with.

 

An amp for sensitive IEMs and an HE-6? Gonna need this for a volume knob.

 

Actually, this beast does both extremely well.

 

I'm not an IEM guy but based on reviews... Project Ember?

 

I wouldn't say "most amps" since this is something that is not really specified in any specs sheet. Sometimes a linear pot would be better because they can be done in a way that allows better tracking at every level (stepped attenuator still best), even though a log pot would allow for a "better sweet spot" since technically, you won't reach high levels too soon.

But I was using that as a simple example, yeah? Nothing to get too technical about.

Uhm... while it is possible to design an output stage that reaches the power supply rails, efficiency is always an issue, so it's always best to assume 80% of maximum as being the case. This is actually referenced in the datasheet of most op amps that are used even in amps up to $500. Some op amps even have different peak swing values for the negative and positive rails.

But on that note, I stand corrected on the Vrms measurement. I somehow confused Vpp with Vp. However, it only causes the worst case scenarios I mentioned to be more significant.

Well, but people who purchase the O2 from some other place, like, say... JDS Labs, have absolutely no way of knowing that.
https://www.jdslabs.com/products/35/objective2-headphone-amplifier/
https://www.jdslabs.com/pdf/Instructions_O2.pdf

Because it is not referenced anywhere except buried on the designer's website... written as one line in an article that has hundreds of lines and a bunch of graphs. Those who buy this amp without knowing its history will most likely not know this fact at all. And I actually doubt even the ones who implemented the design know about it, or else they would have noted in the instruction manual regarding the reasons why.

Yeah, by use of negative feedback, you mean? This is an entirely different discussion (negative feedback versus none), so I won't go too much into it.

But... here's just a passing thought: the headphone/transducer itself is still directly connected to the power supply via the ground line/path. Depending on the topology, this ground may actually not be the same as the ground that's on the source device (DAC), and as such will cause weird things to happen even when the amp's circuit has good power supply rejection. Proof? Look up NwAvGuy's measurements of the Mini3 amp. So it is not true that just because the amp's circuit has good power supply rejection that the power supply can be ignored altogether.

Balanced topology supposedly "hides" ground from the headphone/transducer, however, ground is still used as reference, and if the "reference" is not "absolute", can we trust the signal?

This is not just me speaking theoretically, but it's also based on my experience building amps.

 

Most amps FOR CERTAIN, very few amps (mostly cheap or some portable amps) use linear pots.
You can use a linear pot in a feedback loop and get kind of a log type behaviour.
Also you can use a log pot and create a semi-log pot by mounting a fixed resistor from tap to ground of around 1/10th of the pot value.
BUT the input resistance will vary substantially with the volpot position

Efficiency and output voltage swing are 2 entirely different things.

Nope, by using a clever design like balanced circuitery and proper current sources, even without feedback it is quite possible to design an amplifier with a high PSSR.

Again you are mixing things up.
NwAvGuy talked about the Mini 3 because it has a VIRTUAL ground.
Even in a virtual ground topology the input and output ground are one and the same.
Problems arise when a ground is designed improperly. A wrongly executed starground, 2 separated ground planes, connections at the wrong points of ground planes etc.

As mentioned above ...

 

Showerthought: someone should extract the HE6 drivers and solder em into a new HFM cup/headband shell (HE400S/i/560 should all be compatible?). In theory that would mitigate a lot of the comfort issues while still staying HE6.

 

Ember driving UERM directly has a ton of noise... unless I use a UE noise buffer adaptor. The end result is not necessarily reference, but quite interesting.

 

So how do you know "for certain"? Knowing that almost any kind of pot can be used in any situation? I think you're really just roping things in for an argument at this point.

And input impedance will not vary substantially if the vol pot is implemented not at the input but at a different place like with the Objective-2 design. Though that will create other issues.

Eh? I'm talking about "efficiency" of the amp circuit being able to make use of the voltages it's provided. Obviously it won't be 100%, or in other words, in real life, the amp circuit won't be able to reach close to 100% of the voltage rails. More like 80% in most cases.

Please reference actual real data:
http://www.analog.com/media/en/technical-documentation/data-sheets/AD843.pdf

Figure 2 on page 4, output voltage swing versus power supply voltage. Notice how it's consistently not 1:1?
Also Figure 3 on page 4 shows output voltage swing versus load resistance, which I alluded to: maximum output voltage swing is NOT constant with regards to all loads, and is actually more or less going to vary in a non-linear fashion even when current limits have not been reached.

And you will tell me now that most amps implement a "clever design" like "balanced circuitry" and "proper current sources" now?

I'm talking about reality here, not what is "theoretically possible". Also, amplifiers that use op amps have power supply rejection ratio that vary with frequency (oh yeah, that actually happens), and they'll actually perform worse and worse at higher frequencies, thus necessitating a proper power supply design still. Again, you will not be able to ignore a power supply in a real world situation.

Again, please reference real data:
http://www.ti.com/lit/ds/symlink/opa627.pdf

Figure 11 on page 8, please... and yes, that variation of PSRR is within the audio band. You may also find it in the previous datasheet that I linked above.

Is it possible to design a discrete circuit that does better? Yes! Is everyone doing that? No!

So is "ground" now not a part of the power supply?

Also the Objective-2 has a VIRTUAL ground as well. If you can actually realize that any amp (no exception) operating on DC adapters and/or battery actually has to adopt a virtual ground implementation in its power supply, you'd know what a silly thing you just wrote there.

And I won't even go into NwAvGuy's intentions to pick on the Mini3... but I'll simply say that... he had an agenda. We can continue that discussion at our leisure elsewhere.

I rest my case. I don't think you know half of what you are on about (though, again, I admit you did correct me on Vrms calculations). Please try to actually design, then build an amp, measure it, listen, and then go back and tell me if I'm being crazy.

But seriously, I think I'll stop responding here. This is dragging on, and obviously none of this relates to the HE-6 anymore.

 

It would be nice if Hifiman sold said parts like all other Headphones companies do.