If it graphs bad then it is bad; yes or no? - Page 2

Okay, if

(1)  Z_s is non-trivial compared to Z_L

(2)  Z_L varies widely with frequency across the audio range,

then V_L varies widely with frequency across the audio range.

V_L = V_s * Z_L / (Z_L + Z_s)

So when you measure V_L vs. frequency (implicitly assuming an equal input at every frequency), like for the FR graphs posted, you may see some bumps.  Any added series resistance or source output impedance or impedance of the cable can be lumped into the Z_s figure, pretty much for the sake of this model, which works pretty well in reality for audio-frequency stuff and this scenario.  In practice, output impedance also varies with frequency, but not by a lot usually, except whenever they have a capacitor on the output for DC blocking.  You only see that cap on relatively cheap designs; that's the cause of bass rolloff when some of these devices are tested with headphones.

edit again: as mentioned before, all HiFiMAN IEMs and headphones are very resistive (as far as I know), so they would be minimally affected.  That is, except the HE-300, which nobody cares about anyway.  And it's just ~55-90 ohms, so it'd be a minor bump only.

@scuttle: It can make a big difference. Loudspeakers are also microphones. Basically you want to short them (Zs = 0) so that the back EMF can do its work and damp the speakers ringing.

As a general rule, yes. If it graphs bad then it is bad.

Some people like to wriggle on this point, somebody always wants to quote some amplifier that 'everybody' thought sounded wonderful that had crap measurements, and it's very difficult, even for a convinced engineer, to stand up against the enormous weight of antiscientific sentiment that seems to have found expression in the audio arena. You only have to look at the huge mass of contributed material on Head-Fi, the very large majority of which is complete and utter rubbish, to get some sense of the preponderance of popular opinion which would really prefer not to have it's fantasies contradicted by scientific evidence. Practically every mother's son thinks they know better than Einstein.

You still have to be careful. Not everything actually works out the way your intuition suggests. That's why we have the expression 'counter-intuitive'. Even given that though, there's far more blindly accepted bullsh1t out there than there is unrecognized truth.

I like your attitude, scuttle. Sometimes your lack of expert knowledge in some area or another gets you into a bit of trouble, but fundamentally you're sound as a bell. Just don't hesitate to recognize when you've made a mistake. You probably don't anyway. There's no shame in being wrong, everybody's wrong from time to time. The ability to change your position is absolutely essential in a rational person. Like they say, the man who never made a mistake never made anything.

Don't take any wooden nickels.

w

Quote:

Originally Posted by scuttle 

 

Ok. But in this case it looks (ie from xonr's diagram - thanks for that) like adding a resistance in series is has an effect that is never achieved by varying iem resistance (unless maybe you have a very low iem resistance - in this case loweer than 8 ohms, which is not going to happen.) So who cares?

Going back to xnor's diagram:

Let me put names on some of this.  Vs is the amplifier.  Zs is it's internal impedance of the amplifier.  Zl is the load, in this case the Shure SE530.  Vl is the voltage measured across the load, which is what is being graphed in your original graphs.  This circuit is essentially a voltage divider made up of Zs and Zl.  I found this link about voltage dividers, it seems to be one of the more simple explanations out there:

http://www.aikenamps.com/VoltageDividerRule.htm

Now for a few details that are very important to understand.  In any amplifier Zs is not really just a resistor.  It actually is a complex impedance made up of a resistor, inductive reactance, and capacitive reactance.  The latter two may be very small and insignificant, or may have an easily measured effect.  We don't really know what's in the Hifiman, but most solid state amps, and many tube amps too, can be considered to have mostly resistive source impedances.  This is a small but important detail.  The Hifiman amp output impedance is about 32 ohms.  That isn't actually specified, but it can be inferred.  But we run into a small problem here.  The 801 isn't current product, and there are no published specs for it on the Hifiman web site. Their current 602, however, shows this information:

Output level: 1.1v at 32 Ohm; 2.2v at 150 Ohm

We can estimate the output impedance of that unit to be around 32 ohms.  Why?  Because the combination of the output impedance of an amplifier and the load create a voltage divider.  When the two are equal, exactly half the unloaded amp voltage appears across the load.  The specs show 1.1V at 32 ohms, 2.2V at 150 ohms.  150 ohms isn't technically "unloaded", but it's high enough to get close to the right answer.  The 602 has an output Z of approximately 32 ohms, and is likely almost purely resistive. There is a high probability that the 801 used a similar output stage.  Sorry, I thought I'd seen the 801 specs somewhere, can't find them again.

Now for Zl.  In the case of the Shure SE530, it's not a just a simple resistor, but in reality it's a complex load made up of a resistance, inductive reactance, and capacitive reactance. But, more complex than even that, the result is an impedance that varies with frequency.  The graph from headphone.com is not very detailed, but at least shows what's going on:

You can see that it's somewhere around 30 ohms at the bottom, swings up at 1.2K, then dips quite low at between 5K and 6K.  Recall that the amp Z and the load Z create a voltage divider.  The load Z is varying with frequency.  So the frequency response measured across the load will vary too as a result, with dips at the minimum impedance points and peaks at the maximums.  

Now lets look the same circuit, but this time with the Clip as the source.  It has a source impedance of about 1 ohm or less.  Looking at that 1 ohm Z and the load again as a voltage divider we can see that the 1 ohm resistor has very little loss, so nearly all of the amps voltage will appear across the load.  The fact that the load varies with frequency doesn't alter the measured response across the load much because the source Z is so low.  Even at the lows point of the SE530's impedance curve, the source Z is still about 1/8 of the load Z.  

Now, if you wanted, for some odd reason, to make the Clip measure just like the HM-801, you just need to look at their biggest difference: source impedance.  Clip, 1 ohm, 801 32 ohms.  My suggestion was to add, at the output of the Clip, a series resistor of 27 ohms (again, suggested because it's the closest 5% standard value to 32 ohms, and easily purchased).  The circuit would be like this:

The total source Z is now 1+27=28.  Yes, I know, it's not 32, but you could make it 32 if you wanted to, it's just a little more trouble.  Again, I chose the value in case somebody actually wanted to try this and didn't want to buy a lot of little resistors to make up 31 ohms.

Now see how it compares to the equivalent circuit using the Hifiman:

From the standpoint of the load looking back at the amp, the two are nearly identical.  The results of a frequency response test across the load would also be nearly identical.

I home that helps to clear up some of your confusion.  If not, let me know, I'll try again.  

Thanks to xnor for the original drawing and discussion, and mikeaj for his post as well.

although I use graphs to an extent to base my judgement on a headphone's quality, some of these FR graphs are quite sketchy, can't tell if their set-up is properly compensated or not, case in point is innerfidelity, Tyll seems to be able to actually match(engineering wise "close enough") a manufacturer's own graph while changstar and headroom don't, this give me a distrust of many sources of FR graphs, innerfidelity is the only one so far that run consistent results

30 Ω is also a standard 5% value (less available than 27 or 33, but still easy to find if >1W power is not needed - obviously, power is not an issue with the Clip, so 1% metal film resistors rated for 0.4-0.6 W are fine):

20%: 1.0,        1.5,        2.2,        3.3,        4.7,        6.8
10%: +     1.2,        1.8,        2.7,        3.9,        5.6,        8.2
 5%: +  1.1,  1.3,  1.6,  2.0,  2.4,  3.0,  3.6,  4.3,  5.1,  6.2,  7.5,  9.1

From what I remember of the measurements that went around at release, its not just the output impedance with the HM801 - correct me if I'm wrong, but I think I remember that even lineout from the DAC was -4/5db at 20khz, a fairly steep roll off too starting around 5k and -3db (i.e. audible) at around 15khz, where it could actually affect playback with a lot of music. 

Quote:

Originally Posted by stv014 

 

30 Ω is also a standard 5% value (less available than 27 or 33, but still easy to find if >1W power is not needed - obviously, power is not an issue with the Clip, so 1% metal film resistors rated for 0.4-0.6 W are fine):

 

20%: 1.0,        1.5,        2.2,        3.3,        4.7,        6.810%: +     1.2,        1.8,        2.7,        3.9,        5.6,        8.2 5%: +  1.1,  1.3,  1.6,  2.0,  2.4,  3.0,  3.6,  4.3,  5.1,  6.2,  7.5,  9.1

Yes, but the point was readily available, and that means Radio Shack, and that means 27 ohms.  .25 watt would be just fine.  Come to think...a 100 ohm pot might be fun.  You could "tune" how much impedance-induced response wobble you want.

Quote:

Originally Posted by EddieE 

From what I remember of the measurements that went around at release, its not just the output impedance with the HM801 - correct me if I'm wrong, but I think I remember that even lineout from the DAC was -4/5db at 20khz, a fairly steep roll off too starting around 5k and -3db (i.e. audible) at around 15khz, where it could actually affect playback with a lot of music. 

Wow, that would be "color"! Guess they wanted their amp to sound "different".

Quote:

Originally Posted by EddieE 

From what I remember of the measurements that went around at release, its not just the output impedance with the HM801 - correct me if I'm wrong, but I think I remember that even lineout from the DAC was -4/5db at 20khz, a fairly steep roll off too starting around 5k and -3db (i.e. audible) at around 15khz, where it could actually affect playback with a lot of music. 

That is because it uses a non-oversampling DAC, which likely also has higher than usual amount of high frequency IMD and ultrasonic imaging, and not very good low level linearity. However, some may prefer it simply because of the "warmer" sound resulting from the treble roll-off.

Quote:

Originally Posted by stv014 

 

That is because it uses a non-oversampling DAC, which likely also has higher than usual amount of high frequency IMD and ultrasonic imaging, and not very good low level linearity. However, some may prefer it simply because of the "warmer" sound resulting from the treble roll-off.

Do we know this for a fact?  Non-oversampling with a gentle filter is pretty old-school.  I'm mean like 25 years old-school or better. Hard to imagine that would ever be better than oversampling with an out of band filter.  Where would you even buy such a DAC these days?

Triple fi 10's FR is being twisted like this:

and intermodulation with the same load (might not be comparable with other measurements, RMAA is not the best tool, but it's still worse than most other players measured by the same guy):

Almost any ipod, iphone, sansa clip or fuze, ... does (a lot) better in pretty much any (measurable) way.

Quote:

Originally Posted by skamp 

Some roll-off on a frequency response graph means that technically, the device isn't "high fidelity", but the audibility of it is not necessarily obvious. It depends on the amplitude of the variation, and one's particular sensitivities.

Same goes for frequency variations due to high output impedance. My iPod Classic, for instance, causes a 1.5 dB boost in the very high frequencies with my Shure SE425s, but I find that it doesn't bother me in the least. But generally speaking, a device with a non flat frequency response and a high output impedance is technically bad, and best to be avoided (IMO).

It's all a question of degree.  Your Classic's 1.5dB boost may be inaudible based on amount and total spectrum affected.  Affect more of the spectrum and that boost might become audible, etc...

Agreed that non-flat and high output Z should be avoided.  That 's what I do.  Every day. Ok, that's an exaggeration.   Just those days that end in Y. 

Quote:

Originally Posted by xnor 

Triple fi 10's FR is being twisted like this:

 

and intermodulation with the same load (might not be comparable with other measurements, RMAA is not the best tool, but it's still worse than most other players measured by the same guy):

 

Almost any ipod, iphone, sansa clip or fuze, ... does (a lot) better in pretty much any measurable way.

Well, that's pretty nasty, but that's a plot of headphone response, right?  They aren't supposed to be flat. Do you have a measurement done with an iPod, etc, for comparison?