Working on Amp Measurements

Going to be doing a big article this summer comparing the new HiFiMAN HE-1000 and Mr. Speakers Ether with a variety of other high-end headphones and using a handful of high-end amps. In preparation I did some more work today on the amplifier measurement routine.

I only got three full sets of measurements today, click for the .pdfs: Simaudio Moon Neo 430HA; AURALiC Taurus Mk2; and JDS Labs O2 AC Powered.

I got rid of the spectral plot of 500Hz as it didn't seem to be that revealing, and added in a DIM/TIM (dynamic intermodulation/transient intermodulation) distortion plot. It's not particularly easy to interpret as the result is just a couple spikes in the spectra, but I do also record the DIM percentage as a number in the single point measurements at the bottom right.

I've also decided it was a dumb mistake to do the measurements with 0dBu as the reference as no one uses it anymore. The question is what should the standard output voltage be for measurements of amps? 0dBu is about 0.25Vrms, which pretty loud on most headphones. I'd be uncomfortable testing all amps at 1Vrms as many portable amps might really be straining at that point. Anyway, open to suggestions there.

THD+noise is quite a problematic measurement, for reasons I won't bother to explain, but you'll note it has stair steps in it sometimes. I have called Audio Precision about it and they said is was a normal measurement artifact as the machine changes ranges internally...ergh. Anyway, I might play around with the scale of this plot. Some amps might be able to go up to 15 volts before hitting 1% distortion. The problem is, in Excel and log scale I have to jump up to 100V if I want to go higher than 10V, which kind of screws up the usefulness of the graph. I could go to a linear scale, but a lot of things happen below 1V sometimes, and it would be hard to see with a linear scale. Open to suggestions.

I don't really want to say too much about the plots themselves as I've got so little experience interpreting them and seeing a wide selection of amps, but... Boy, the AURALiC has a lot of cross-talk. The Simaudio is a beast—half Ohm output impedance and can dump 6 Watts into 16 Ohm before hitting 1% distortion. And the O2 is very low noise even when run off it's wall wart.

Enjoy the plots! I'll try to do some more clean-up on scales and axis and do some more measurements soon.

logscool's picture

Your Measurement sheets all show up as 2 pages. The second page just shows the far right side of the first page again.

Bob Katz's picture

Hey, Tyll... how about my vertical scale calibrated in dB SPL based on the sensitivity of a popular headphone. Multiple scales on the axis: volts and dBu as well.

Bob Katz's picture

Why are people taking THD+N measurements anyway? Better to see individual harmonics and total of harmonics versus the fundamental. THD+N going up as signal goes down is (probably) not an indication that distortion is going up--- it's an indication that noise is a larger proportion of the measurement.

zobel's picture

I think this holds up well as an overview of all relevant measurements being done on amps. I appreciate the practical, real world view of the parameters, and the explanations of acceptable values and ranges in some of the factors measured.

Tyll was wise to ask this guy to provide this synopsis, which he did for free;

zobel's picture

It does show how much downstream noise is present, you know, the noise present after the volume control.

zobel's picture

if we knew which harmonics were present, we could come up with a weighting system to attach a value to there relative audibility. There would have to be an agreed upon system of weighting even vs odd harmonics with respect to their respective levels and frequencies. You, and other educated folks have the ability to make some inferences from the graphs you are mentioning but without standardization, a total weighted value of distortion as a single number wouldn't be possible. We would be making the "guesstimates" we are now doing with the graphs of headphone frequency response.

I think Tyll's raw data for headphone SPL/frequency is too hard to make valid interpretations as to the audibility of sounds above 1.5 kHz. Everything below that seems good to go though.

One thing that would make these SPL/Hz graphs acceptably useful would be to know the best, most representative curve that subjectively represents flat to most people as measured at the eardrum. This of course means at the approximate position of the eardrum in a specific dummy head with it's unique size, shape and microphone/electronics package. Olive's curve is getting close, but it needs refinement.

Once this curve is finally established for the individual dummy head being used, it can then serve only as a generalized, one size fits all compromise, since we all have different ears and brains. But it would be helpful, as it could be used as a correction curve for any particular measuring head. Maybe with some added smoothing of the curve in the upper frequencies to down-play comb filtering and resonances up there, a flat SOUNDING pair of cans will graph out as a flat line. That is what we are trying to guess at now with the raw SPL data, making corrections in our heads out of wacky lines we try to imagine conforming to an iffy model.

Bob Katz's picture

.. but he did not discuss a harmonics only measurement as I recommend. THD+N for the traditional value and then you have to qualify it when the signal is low that it is clearly dominated by noise. A harmonics only measurement does not get higher with signal unless there is crossover distortion or other nonlinearities in the DAC or amplifier. So it is more "discriminatory.

I also found it interesting that the engineer with the initials recommended 110 dB peak level as the max that we should be interested in. It was the same value that I arrived at as well. If Tyll marks his graph axis in dB SPL for a few headphones it can tell us something. I know headphones have various sensitivities, but when you translate watts into volts, it's amazing how much more closely they line up. I'd rather see an SPL level for one headphone and a chart showing what the SPL would be comparing two other extremes of headphones.

zobel's picture
Bob Katz's picture

I wish he wouldn't use unpronounceable initials for a pseudonym. It depersonalizes him and puts him on an unreachable pedestal. Can't even call him by his first name. Really.... if for personal reasons he doesn't want to reveal his real name, then do what actors and actresses do, make up a first name and last name that people can call you by.

zobel's picture


I agree, North West Audio Video Guy is a dumb name. He wasn't dumb though. Quite the opposite. He did remain anonymous for personal reasons, and then disappeared, poof. He became famous as a straight shooter who ripped into audiophile mythology, and all the snake oil and BS in the audio industry. He was sort of a Robin Hood for truth and justice in the industry. Tyll just measured the O2 amp that he designed and offered in kit form for no profit.

So, you are really Bob Katz? You have to admit that sounds a bit like a pseudonym. If you have kids would you name them Tom, House, or Wild, or maybe Hep? You should ask Tyll if he is a Bobcats fan. He is from Bozeman, the home of the "Cats". He would probably get it.

Bob Katz's picture

Can you please show the graphs larger? Graphs are very hard to read, and when expanding the size it pixelates.

tony's picture

Click on the link below, the graphs are beautiful.

Tony in Michigan

tony's picture

Summer is looking up !

I hope you do this like a HBO TV series.

Start with the Pilot, than do Episodes, hmm, 1 thru 9, close with a Cliff hanger for Next Seasons Episodes.

Headphones is Evolving at a rather quick pace, all these devices being released, phew, you can be the one to tell this story to the World. You could even have someone translate it into German and some of the other languages ( Japanese ).

This can be great journalism for our technical readership. ( don't crash your bike )

You've got Joker, Steve G., Bob Katz, Grandberg and you, all men of integrity.

Manufactures realize all this and will cooperate, they want a fair presentation of their facts.

That "New" Simaudio does seem a powerhouse, plus you can get the DAC built in for a bit extra. Betcha it could power all the headphones in a Classroom Lecture.

I've heard that the 1000 is a finesse headphone, not all that power hungry and I've heard that it will likely be the Product of the Year for headphones. Hmm., bet they'll be as common as HD800s at all the headphone meets.

I've found a DAC that makes all my music sound better.
( there is some talk about a new super DAC that is rather hard on the lesser music recordings, I ain't mentioning any names )

In for a Penny, In for a Pound!

I'm In!

Tony in MIchigan

castleofargh's picture

maybe use a different voltage for reference of portable amps and non portable?
or maybe more relevant, amps for portable headphones, and amps for non portable hard to drive headphones. as some portable amps are actually very powerful and more likely to be used for that reason with hard to drive headphones.

the voltage value for noise is a real nice point as it may give a hint when to avoid the amp if we use very sensitive IEMs. that way I can even go and try to get how loud that noise will be into a few headphones. \o/

it would be nice to have a little tuto or index "how it's done, what are the values I used". obviously you can do that when you're set on definite values ^_^.

anyway it would be very cool if you could become a main spec provider for headphone amps like you are with headphones. looking forward to the future, thank you Tyll.

Tyll Hertsens's picture
I thought about using different reference on portables, but I really want to remain as standardized as I can. I already know I'm going to have a different routine for tube amps---have to go to higher distortion levels on the THD vs. output voltage.
zobel's picture

having an issue measuring a low enough THD + noise level with your current setup? Are the results showing too high of levels for some reason? Is there some AC or other grunge creeping into the tests somewhere?

zobel's picture

Your amp THD + Noise vs output is looking good in that NwAvGuy's own measurements that he posted pretty much agree with your graphs.

Tyll Hertsens's picture
Yeah, I'm pretty sure it's just background acoustic noise. I have a "sound proof" test chamber, but it only cuts down noise so far.
Jan Meier's picture

Dear Tyll,

nice to see you're progressing on the amp measurements.

Personally I do find the SMP-IMD measurements most interesting. People in discussions often limit themselves to harmonic distortion, probably because these signal components are more easy to understand. However with audio signals intermodulation-components are normally much stronger and much more obnoxious as they're not harmonic to the musical contents.

With the scale you're showing the SMP-IMD results it's not really possible to discern the IMD-portion of the signal. Around the large 7kHz peak you should have smaller peaks at 7000 +/- n*60 Hz. These smaller peaks originate from the interaction of the 60Hz and the 7kHz signals. You may consider a separate graph showing the frequency-range of 6500..7500 Hz only.

Just a suggestion though.



Tyll Hertsens's picture
Thanks Jan, I'll look into that.
veggieboy2001's picture

I'm sure this will be part of your work, but I'm hoping there will be a "how to interpret" the graphs, much like for the headphones in the resource section. I'm happy to say that this website is growing somewhat parallel to my own journey as I am researching portable headphone amps and I'm hungry for info...
Thanks...I'm looking forward to the fruits of your labor!

Bob Katz's picture

If my name was AVguy you could ask if AV was my first name and guy my last name. Once in a while I get questions about the legitimacy of my name, but all the alley katz know it's my real name. :-)

zobel's picture

it's probably more important that the Phat Katz know you by you name, otherwise you wouldn't have as much opportunity to hang with the Cool Katz who make the music. Do you have a son named Tom?

zobel's picture

How about Kit? Any relation?

StefanP's picture

Vinyl exhibits a crosstalk ranging between -30 to -15 dB. So what will the -60 dB "only" imply for the Auralic? I love it and appreciate its natural non-fatiguing sound with plenty of drive and low-level grunt. My current favourite being the AKG 712 and the Hifiman HE-500, saving up for the HE-6. Pro. But for electronics this seems to be a lot. Between unit variation or intentionally designed?

chik0240's picture

It's surprising to see that the Taurus Mk II have such a high cross-talk figure, any thought on that?

StefanP's picture

Yes, me too! Less analytical? More vinylish?

StefanP's picture

An update: I am comparing the Taurus with the Burson Soloist SL using the AKG 712 Pros and the Burson is more spacious-sounding but the Taurus gives a more solid presentation with deeper and tighter bass. The price ratio here in Sweden is 3 between these units. Maybe the AKG is not good enough to give a better separation. Anyway I though that higher X-talk would give a less tight sound but not here. Anyone with better understanding of this parameter?

mat's picture

Don't crosstalk and THD significantly vary with load impedance?

From NwAvGuy's blog:
"THD and crosstalk are meaningless if the load isn't specified and it almost never is."
"The O2’s crosstalk, top to bottom, at 15, 33, 150 and 600 ohms ... measured about –65 dB, –72 dB, –91 dB and –95 dB respectively."

I think it might be helpful to specify the load impedance that was used for these measurements, or did I just miss it on there somewhere?

mat's picture

Ah, sorry. I see that you show the load impedance for THD, but I'm not seeing it for crosstalk.