InnerFidelity October 2012 Update

Been a good month for some unusual stuff. Stuff, unfortunately, that I can't quite tell you about. But there's some cool things I can talk about.

Headphone Measurments
Let's start with this month's headphone measurements. As usual, I've updated the AllGraphs.pdf of course, and added the cans to the measurement datasheets page. New measurements include:

  • You've probably seen the recent article about the Logitech UE9000 and UE6000, which measure very well in passive modes, but I also got in the UE900 IEM for measurement. I liked it even though they seem a bit bright to me. They've been sent off to ljokerl for review consideration.
  • He (ljokerl) also requested a set of MEElectronics A161P in-ear headphones and had them shipped to me for measurement before listening to them for review. I liked these as well, though again, I thought them a bit bright.
  • Also got in two other IEMs: PNY Midtown and RBH EP1. They just showed up on my doorstep. Probably should have just left them there.
  • Head-Fi member JupiterKnight was pretty excited about the JVC HA-S500-Z inexpensive on-ear cans. They're not available in the U.S., so I told him I really wasn't too interested. But he pressed on and said I should really have a, they're pretty good for a cheap phone. I've called JVC and asked for a sample of the HA-S400, which is available in the U.S. If they're as good as the 500 they'll probably get a review from me.

Beyerdynamic Custom One Pro Measurements
121001_Update_photo_CustomProBoxGraphI'm not sure I like them well enough for a full review, but they sure do have an interesting feature: an adjustable bass port. This four position switch on the rear of the earcups allows you to select four different levels of bass.

The image to the right shows a graphic on the packaging for the headphones, which has a graph showing what EQ change you can expect for each switch position selected. I figured I'd measure the headphones for each position. You can find the full sheets in the AllGraph.pdf and the download page, but I also made a graph overlaying all the curves, which you'll see below. It doesn't really look like what's on the box, but it is doing stuff to the bass.


Headphone Amplifier Measurements
Ugh. What a process. It's like untangling spaghetti. I've started to take various measurements with a variety of amps to get a sense of what the scaling of the graphs will be. It seems the THD+noise vs. Output Voltage is a bit of a sticky wicket. I been trying to make the test as common as possible for all amps, but it looks like I'm not going to be able to use a common set of loads or maximum voltages for all amps. The OTL tube amps and, I suspect, other high output impedance amps, just aren't going to like the 32 Ohm and 15 Ohm loads. The amps aren't designed to drive low impedances, so it's really not fair to measure the amps with them. And portable amps aren't going to deliver voltages nearly as high as home amps. Unfortunately, that means separate tests, and spreadsheets that aren't directly comparable.

Anyway below you'll find two images from the AP tester showing THD+noise vs. Output Voltage. For the tests below, I set the amps' volume controls so that with 1Vrms in they gave 1Vrms out (unity gain) and then ran the sweep from 0Vrms to 4Vrms input voltage. The following curves show THD+noise vs. Output Voltage with 32 Ohm and 470 Ohm loads.


The above graph show the six amps tested into a 32 Ohm load.


The above graph show the six amps tested into a 470 Ohm load.

As you can see, the tube amps (Woo, Cary), are not liking life with the 32 Ohm load. The O2 measure great until you hit the limits of its lower voltage power supply rails. Not sure why the Burson performed poorly with the 32 Ohm load. And it's pretty obvious the solid-state Apex Butte and Lake People amps need to be driven to higher output voltages to see where they start to distort. Bottom line: My idea for a one-size-fits-all amp test is likely going out the window.

Upcoming Stuff
Parrot Zik - I've got one in-house, and I think it's pretty cool. I've just received a Bluetooth transmitter that I can hook up to my AP tester so that I can measure wireless performance. The Zik will be the first victim. I'll go back and measure the Logitech UE9000 as well.

Sennheiser HD 650 and HD 600 - My persistent bugging of Sennheiser's PR people finally paid off, and I've just received an HD 600 to go along with the HD 650 I have here. I've been wanting to do this review for quite some time. These are a couple of great, classic cans.

Beyerdynamic DT250-250 - There's been a bit of chatter lately on the boards about these headphones, so I caller Beyer for a review pair. I've loved these cans for a long time. I'm so glad they're still making them.

V-Moda M-100 - Seems that Val has pushed the button on these, and has told me a pair will be showing up shortly. Been waiting quite a while for him to be happy with this new baby. My fingers are crossed.

Rocky Mountian Audio Fest - I'll be taking off next week Thursday for RMAF, and the whole middle of this month will be filled with reports from the show. I'm very excited! It'll be a great chance to catch up with some of the industry folks and get some gear coming in for winter. Hope to see you there!

October is my daughter's favorite month; she loves to dress up for Halloween. I grow three days worth of face fuzz and wear the same costume every year.

Scary, isn't it.

mward's picture

Nice to see measurements of the UE900. 

Any plans to measure Apple's EarPods? I'd be glad to loan you a set if that's what it takes! I'm interested to see how they stack up vs the old Apple EarBuds. 

Tyll Hertsens's picture

Thanks for the offer. But, I'm gonna try to get in touch Apple. I'd love to do a review, and this would be a good way to make a connection with them.

ultrabike's picture

Love to hear more about the HD600/650 and the M100s.  As far as your amp results, IMHO they look pretty promissing. Usually headphones are divided into low and high impedance. Sounds reasonable to compare amps when driving at least two kinds of loads (and gain modes.) Would be intersting to know what voltage is required to drive certain headphones to 90 and 100 SPL. Might help with pairing the headphone to the appropriate amp... Who knows...

Also, cool outfit mang yes

johnjen's picture

There are 2 green traces but I figure the O2 is the lower of the 2 in both graphs.  Perhaps it should be a dashed line?

Just say'n is all…wink

I too look forward the papa V-Moda cans.  The 80's are very nice.


Tyll Hertsens's picture

Unfortunately the AP can't do that. 

DJ's picture

Great to see progress in amp measurements.

For the tests below, I set the amps volume controls so that with 1Vrms in they gave 1Vrms out (unity gain) and then ran the sweep from 0Vrms to 4Vrms input voltage.

I understand why you used this method and it's OK for majority (maybe all) of amps out there but it's not optimal for O2 amp. What you basically did with the other amps is attenuate input signal to amps gain stage with volume pot by Av (Av is gain of amps gain stage) to obtain unity gain of amp as a whole. As O2 amp is different than others, with the volume pot after gain stage, by turning volume pot you did attenuate output signal and obtain unity gain but you didn't attenuate input signal to its gain stage, and therefore with AC power and low gain setting ( Av=2.5 ), output signal from input gain stage will reach the voltage of power supply rails when input is at app. 2.8V ( 7/2.5=2.8 ), which means clipping and rapid rize of THD+N thereafter. You can see that clearly at your graphs. With high gain and/or battery power the threshold would be at even lower voltage. 

The O2 measure great until you hit the limits of its lower voltage power supply rails.

You didn't reach the limits of its lower voltage power supply rails, you just overdrive input gain stage by amplifying input signal larger than 2Vrms (not likely to happen in the real use) 2.5 times . It would be fairer if you could turn volume pot all the way to the right (max) and use your own pot before O2 amp to attenuate input signal by 2.5x to obtain unity gain. The THD+N should be higher but not significantly (by amount of THD+N of input gain stage attenuated with volume pot) and should not rise until app. 7V which is enough for app. 119dB with insensitive HiFiMAN HE-5LE or 300ohm Sennheiser HD 800.

Other than that GREAT WORK. Looking forward to see further improvements and measurements.


Tyll Hertsens's picture

In this case, I just sort of arbitrarily picked a volume control setting that might work for a broad range of amps. Setting the amps to unity gain just seemed like a good start for exploratory purposes. 

For actual testing I'm not sure how I'm going to set the pot. it'll be obvious in some case (max gain measurement for example), but for others not so obvious. I'd like to just turn it all the way up and leave it, but then it means the amp will be operating at pretty high gain, and some amps may be noisier than they otherwise would be in normal use.

The other half of the story is what the input signal levels should be. The easy answer is 0dBu (0.775Vrms) , but that's likely to be too hot for some portable amps. Drop it to -10dBu and you might run into problems showing higher signal to noise measurements than users would really experience. 

"You didn't reach the limits of its lower voltage power supply rails, you just overdrive input gain stage by amplifying input signal larger than 2Vrms (not likely to happen in the real use) 2.5 times ."

My understanding is that with solid state amp it usually amounts to the same thing as they can usually amplify signals almost to the rails, which is why they clip so hard when they do start clipping.

Which brings up another point: It would be nice to clarify by measurement the highest input voltage before clipping the input stage, and then separately characterize the THD+noise vs. voltage of the output. 


Ugh, what a mire. I think I'll try to talk with John Atkinson at RMAF about it. And I'll see if I can chat with NwAvGuy as well.  They're both pretty top notch on the subject.

DJ's picture

As usually it's all about best compromise.

1. Measure Vout with arbitrary Vin and calculate total gain of amp (Av_tot=Av_pot*Av_amp=Vout/Vin)

* You can not change Av_amp (it's constant or it could be changed with gain switch if amp has this option) but you can change Av_pot.

* Also note that higher Av_amp leads to higher THD+N so you should do measurements for every gain settings separately.

* Volume pot at 12 o'clock is problematic because different pots have different taper meaning Av_pot would not be constant (UNFAIR)

* If you turn volume pot all the way up, Av_pot=1 for all amps ( FAIR), THD+N would be worse  (UNFAIR)  but all amps would be at the same ballpark (FAIR)

* If you would rather set volume pot at lower position note that this way O2 amp would start clipping at output voltage that is lower than its power supply voltage. (UNFAIR)

* Once set volume pot position should not be changed through the measurement procedure


2. With sine wave generator and oscilloscope find at what output voltage amp starts to clip (Vout_clip)

* Vout_clip is a threshold that defines working scope of amp and therefore scope of measurement

* As you can see Vout_clip is different for every amp so unfortunately you can't use predefined Vout=4V or any other fixed threshold


3. Set Vout_clip as threshold value for sweep and start automated measurement procedure

* Or you could set threshold for input signal Vin_clip=Vout_clip/Av_tot

* Optionally threshold could be Vout_clip+Vconst (eg. Vconst=0.5V) so that we can see rise of THD+N in clipping area


PS1: For every measurement (graph) except load resistance you should also add Vout_clip and Av_amp values along amps name.


PS2: IMHO, even now we can see big difference between amps and although character of >1% distortion is more pleasant (for someone even sweet sounding) with tube amps I don't want to listen sweet sound of distortion (be it tube or solid state) but rather music as it was before reaching amp. Basically "wire with a gain" and O2 is quite close to that goal.



Willakan's picture

I'm not sure how easy it would be to do, but if you could extend each line on the graph to the point of clipping, to a maximum of 8V, that would cover the overwhelming majority of headphone amplifiers. I'd imagine this might be difficult for combining multiple lines on one graph automatically, but for individual datasheets that seems like it would work out.

The exact output of the amplifier before clipping into different impedances could perhaps be noted in a separate table, to accomodate any ridiculously high-power amplifiers.

Anyway, the measurements are looking great - with extended plots they could work pretty well. I don't think it's unfair to ask these amplifiers to drive a wide range of impedances: the Woo Audio 6 page proudly proclaims that it works with any headphone from 8-600 ohms, so the results into 32 ohms pertain directly to their claims. Very few amplifier manufacturers say that their amps can be used with a small subset of headphones: when they make a claim and their amps can't deliver, this failure can and should be documented.

As for the different power requirements, perhaps the amplifiers could be subdivided into "High" and "Low" power, with the latter graphs only extended to 3V on the THD+N scale? Even if the same graphs are used for both, the information is still there, just not as tidily presented.

Tyll Hertsens's picture

At first my plan was to use loads of 600 Ohm, 150 Ohm, 32 Ohm, and 16 Ohm (NwAvGuy's suggestions). You may be right about just running the tests that way. Don't want to blow up any amps though.

The problem is I have to use the input signal to run up the output of the amp, and I don't know for sure if it's characterizing the input or output stage. It would be nice if I could leave the input fixed and slowly turn the pot up and have the AP gather the data...but it doesn't really work that way.   

yuriv's picture

If you set the volume control on the amp for unity gain, like you have now, it's possible on some amps for the input to overload before you reach the amp's maximum output. The test is still useful because many digital sources have line outs that are hotter than the 2V RMS Redbook standard, sometimes much higher. A FiiO E9, for example, can handle only 2.1V, so its inputs will be overloaded by the line outs of some CD players. It would be nice if your test suite also determines how much voltage it takes to overload an amplifier's inputs.

If I recall correctly, the O2 sets its gain at the input, then the volume control attenuates the signal just before the output stage, which are just unity-gain followers with 4556s. With +/- 12V rails minus a diode drop, and 2.5x gain, the op amps at the inputs will start hitting the rails before the input reaches 3.2V RMS  (3.2V x sqrt(2) x 2.5 = 11.4V). And that's assuming the input op amps can do rail-to-rail swings. The 2068s can't, so the O2's inputs should overload below 3.2V. I think that's what we're seeing in the chart that you have above. If you do the same sweep with a 600-ohm load, you'll hit the same 3V wall.

Dialing the volume control on the O2 for something higher than unity gain--say, 2x (6 dB)--should get you higher output: well above 3V for a 32-ohm load, if I remember correctly. In this case the, the input will be well under 3V RMS, the maximum that the O2's inputs can take at low gain before clipping.

Instead of sweeping from 0V to 4V RMS to determinine the maximum undistorted output, maybe sweep from 0V to 2V RMS, because most amps ought to be able to handle Redbook standard. Maybe set the volume so that the sweep reaches 1% THD+N at the top of the range.

Edit: Just saw DJ's note above after a browser refresh. It's voicing a similar concern.

Tyll Hertsens's picture

As I've been reading you comments and the two above yours (which are all good, btw) I went back and read some email exchanges with NwAvGuy. His suggestion was to set the pot at either unity gain or 12 o'clock (since that's where they end up getting used usually).  

Your point (and NwAvGuys as well) about stopping the trace at 1% distortion is a valid one...for solid state amps. But tube amps will have to go higher, I think given the charts above. The trick there is I'm trying to automate the tests. Brian and I have talked about writing some code that detects some threshold THD+noise value and automatically stops the test at the right level. Just another molehill/mountain to clime.


AstralStorm's picture

What about detecting the doubling of distortion over best case 1 kHz baseline, like in general bandwidth calculations?

(That means 6 dB more harmonic distortion.)

Brentagon's picture

So you didn't like the RBH EP1?  That's disappointing, as I was looking forward to them.  The measurements don't look too bad.  What didn't you like about them, Tyll?

Tyll Hertsens's picture

...they had a huge imbalance between the channels in the low frequencies.  And then they just sounded way too bass heavy.

boniceman's picture

If you can please make some comparisons like:

Logitech U9000 VS Parrot zik

Logitech UE900 VS Shure SE535

Logitech UE6000 vs Beats by dre Studios vs Yamaha pro 500



Tyll Hertsens's picture

I'll certainly compare the UE9000 with the Zik in the upcoming Zik review.

ljokerl will have to decide if he'll review the UE900 and what he compars it to.

And the UE6000 slays the Studio in passive mode.  Noise probably sorta close---both are kinda mediocre, but I'd bet I prefer the 6000.

donunus's picture

I am waiting to see your thoughts on the dt250-250 and how it stacks up with current headphones. I'm also interested in the upcoming review of the hd600s and 650s and compare your thoughts with mine on those too. These three just happen to be my favorites :) 

Lunatique's picture

Will you be reviewing the Westone 4? It's one of those headphones that a lot of people consider to be one of the best universal IEM's out there, and some like it more than the Shure SE535.

Tyll Hertsens's picture

I'll have a chat with Westone at RMAF and see if I can get some for ljockerl to have a listen to and consider for review.

iCupboard's picture

Dear Tyll. 

I have no idea where to ask this question so I am just doing it here. :)


I am on the verge of buying a headphone, but I am torn between two headphones. Those being the AKG K550 and the Philips Fidelio L1 headphones.


I am a "on-the-go listener" that listens to these genres:


  • Indie-rock
  • Indie folk
  • Pop
  • Rock
  • Some dubstep (very little)
  • Audiobooks (those are very important to me)


Which headphone would you suggest out of those two?

Tyll Hertsens's picture



It depends on whether you want a good seal or not. They're roughly the same sound quality.  But I do think the UE6000 is better than both on balance, and with your penchant for rock, I'd go with the gentler-on-the-ears UE6000 over both mentioned.

e_resolu's picture

That would be nice

Tyll Hertsens's picture

Just sent a note off to philips to see if I could get a review pair. They're pretty responsive, so we'll see.  If I get some in they'll have measurements in one of the Updates at a minimum.

itsastickup's picture

...have real difficulty seeing comparison graphs where dotted lines aren't used.


I'm supposing it won't be a problem with individual amp tests but where there are multiple lines for different devices, it would be great if dotted lines could also be used.


A possible alternative might be to use colours that avoid the red/green issue (which is the 5% of chaps who are affected by this form of colour blindness). For example a light green and a yellow or a light orange are all going to get jumbled up. If the colours are differentiated by making the shades more different (ie. a deeper orange, but not too close to red). Unfortunately, while my own colour blindness is mild, I don't think this idea will work for many others.

Tyll Hertsens's picture

...there are numerous limitations with the AP tester I use.  As you say, when the final testing is done, amps will appear on separate sheets. Can't please all the people all the time, unfortunately. But I'll keep it in mind, it's a good point.

itsastickup's picture

....didn't realise it was the tester itself. 

elmura's picture

A fair comparison of THD+N vs Output would apply the principle of applying input levels that are most commonly supplied by typical sources. Looking at the products on test, one would typically wish to feed it with a quality DAC source. So 2Vrms would be a great input level. Another test would use a typical portable device output level of 0.5V and 1V. So, ideally, we compare at 0.5V for the ipod/iphone brigade, 2V for the DAC source, and 1V for everything in between.

Next we should compare at a fixed output level so the volume control would have to be adjusted on each amp to produce a specific output level for a particular load. This level should be set to reflect real-world application of say 85dBfs for an average headphone sensitivity level.

To compare maximum outputs before distorting

I strongly believe that varying the input level above 2V is wrong as it doesn't reflect real world application. To increase the output level for a given fixed input, you must gradually turn the volume up and plot what happens. Or adjust the volume to produce say 2.5v, 3V, 3.5v etc and plot. More time consuming, but a much higher quality test with real-world comparison application.

Then, if there is a Gain switch, you would run the same test at each gain switch position.

Elmura Audio Visual

Tyll Hertsens's picture

Adjusting the volume control for the THD test might be untenable though. 

Lunatique's picture

I'll keep begging for a Westone 4 review until I finally see it. :D It's widely considered one of the best universal IEM's on the market, and worth reviewing.