Head-Fi.org › Forums › Equipment Forums › Sound Science › Tyll Hersten/Innerfidelity's 21 Amp Measurements...
New Posts  All Forums:Forum Nav:

Tyll Hersten/Innerfidelity's 21 Amp Measurements...

post #1 of 25
Thread Starter 

Here is a link to 21 amps that Tyll measured http://www.innerfidelity.com/images/AllAmpsJuly2013.pdf. (I couldn't find any conversation going on about them.) I don't understand measurements at all but I'd love to hear what you all think. Did any of these measure really well? Were there any surprises to you? 

 

If you were to decide on these measurements alone, what are some amps one would pick for 300 ohm HD600/HD650? I have a E09K/Qogir and I'm curious how it compares to the other amps in the list but I'm such a newb. The measurements are abstract to me, and as a consumer and sadly not a "sound science" guy (at least, not yet), I'd love to hear you all draw some conclusions from this data. 

 

What are some of the top two or three recommendations you'd make based on the measurements alone? I don't think this is just a normal "What amp would you recommend?" because the answers those threads get really aren't based on measurements... 


Edited by imackler - 7/3/13 at 2:20pm
post #2 of 25
He explains it in past updates, I remember he got some advice from nwavguy, you may want to start in that blog and get as much info as you can.

The main point I look for is low noise and flat freq response to atleast 20 kHz.
Edited by KamijoIsMyHero - 7/3/13 at 2:24pm
post #3 of 25

With regards to output impedance, the two amps with the highest output impedance measured are the Beyerdynamic A1 and Alo International which have an output impedance of 99 and 21 ohms respectively. Does anyone know why those amps are designed as such? Wouldn't a high output impedance make it harder to achieve the 1/8th rule of impedance matching?

post #4 of 25

I guess Beyerdynamic couldn't decide between 0 and 120 ohms. tongue.gif

 

http://www.head-fi.org/t/592730/beyer-amps-with-100-ohm-output-impedance-lolwhut

post #5 of 25

High output impedance is also easier and cheaper to implement, in addition to making the power output vary less with the load impedance (for a range of 16 to 600 Ω, 98 Ω output impedance is optimal in theory for that purpose). It is a simple protection from capacitive and very low impedance (e.g. short circuit) loads, and also reduces the distortion of the amplifier with a low impedance pure resistive load because it will "see" an overall higher impedance (of course, the distortion of a real dynamic headphone can increase because of the reduced electrical damping).


Edited by stv014 - 7/3/13 at 8:58pm
post #6 of 25

The better way imo would have been to reduce the output impedance and add a gain switch for sensitive/low-impedance headphones. But yeah, just adding a bunch of resistors is easier and also more "coloring".

post #7 of 25
Not sure quite what they all mean, but am curious about the Pico Slim. It is highly regarded, but the measurements don't look that great in terms of THD especially (7% into 16 ohms in the test). Anyone know what's going on there?
post #8 of 25

It is only that high at 0 dBu simply because the amp is clipping (it is not much of an improvement in terms of power output compared to using some portable devices without it). At a slightly lower output level, the distortion is less than 0.1%.

post #9 of 25
Thanks for the explanation stv014.
post #10 of 25

Generally, I think the amp measurements are mostly good to ferret out poor designs. Too much noise, 60Hz crud, high cross-talk, high harmonic distortion, high output impedance, or high IMD distortion, and I'd shy away from the amp. After that, it's hard to say whether the good measuring amps are worse sounding than the great measuring amps, but you can be pretty sure that the ones that measure really good do sound pretty darn good as well. 

post #11 of 25
Thread Starter 

I've been reading this article here http://www.innerfidelity.com/content/nwavguys-heaphone-amp-measurement-recommendations to understand these measurements. But, seriously. It's Greek to me. Actually, I have an easier time with Greek. 

 

I've got a Qogir (E09K) to a HD600. I think it sounds good. But I have to decide if there is another amp (in the maybe up to $300) budget that I should try. Someday, I'll get to a meet in Seattle, but I haven't been able to yet. I have already read thousands of subjective impressions of amp pairings. I value them. I know what amp  I would purchase if someone gave me $500 or $1000 to spend on an amp.  But deep down, I'm kind of a skeptic. It's weird to me that the pattern seems to be that almost every time you spend more it sounds better. There are exceptions, such as when you build it yourself or if the engineer is banned. Then the cost is less important. But like I said, I'm skeptic. I think if I found a $300 amp at a garage sale for $25 it would be the best sounding amp ever. I don't like that about myself...

 

Anyways, would anyone be willing to help me understand what weaknesses the E09K/Qogi has with a 300 ohm load from Tyll's measurements? If you would be willing to point out to me an amp that does better in the weak areas, I would be so much further along in balancing subjective impressions w/ the objective data from these graphs. I would really appreciate it. 

 

Right now, I can see the differences as I flip back and forth between the two charts... I just can't assign a value or interpret them. 

 

Basically, I'm mucking up the sound science forum with my ignorance. Thanks for your patience :) 

post #12 of 25

The following is simplified to hopefully make it easier to grasp:

 

To understand most of these plots you have to understand the difference between the time and frequency domain.

 

If we plot a signal in the time domain we have the time on the x-axis and amplitude on the y-axis.

A simple sine wave starts at t=0. The amplitude rises with increasing time until it reaches its peak, then falls again, crosses the zero line, goes below zero until it reaches the negative peak, decreases in amplitude again until it finally reaches zero again. This is a full cycle (360°).

If it takes a sine wave 0.1s for a full cycle the frequency is 1/0.1 = 10 Hz. In other words, we get 10 such cycles per second.

 

Now let's look at a more complex waveform, a square wave:

 

 

 


A perfect square wave can be represented as an infinite sum of sine tones. The plot above only shows 3 harmonics. If the square wave has a frequency of 10 Hz the 1st harmonic (or fundamental) is a 10 Hz sine wave, the 3rd harmonic is a 30 Hz sine, the 5th a 50 Hz sine etc.

Also note that each higher harmonic has a lower amplitude. If you sum all (1st, 3rd, 5th, 7th, ... infinity) harmonics up you get the square wave that is shown above.

 

 

Now let's plot this signal in the frequency domain:

 

 

The x-axis now shows frequencies instead of time. The y-axis still shows amplitude, but in decibels.

 

So you can see there are sinusoids at 10 Hz, with lower amplitude at 30 Hz, with even lower amplitude at 50 Hz ... just like in the time domain plot above, though normally you just see the summed square wave in the time domain.

 

In other words, going into the frequency domain is like dissecting the timedomain signal and analyzing how much of the signal lies at a given frequency.

 

 

 

Now what do you see if you look e.g. at the Frequency Response measurement graph?

Assume you send a signal into the amp that contains all frequencies, so that you should see a flat line from 0 Hz to infinity in the frequency domain.

The FR measurement shows the frequency domain representation of the signal that is output by the amp.

 

Since DC (0 Hz, essentially a flat line with an amplitude other than zero in the time domain) is bad for headphones, since it just generates heat, many amps have a DC blocking filter. Depending on the design of this filter the amp may show some low frequency roll-off. That means that as you decrease frequency the amp will output a weaker and weaker signal until it doesn't output anything anymore at DC. If this roll-off starts in the audible range (above 20 Hz) and is severe you could potentially hear weaker sub-bass than it should be with your headphones.

There's also high frequency roll-off, for reasons I won't mention here. The audible range ends at about 20 kHz (20000 Hz), so again if the roll-off is not severe our above that it's not gonna cause audible problems.

 

A much more audible problem can be channel imbalance. In the Frequency Response graph, if you see one channel (red or blue) a dB higher than the other it means that all sounds will be played slightly louder in that channel.


Edited by xnor - 7/4/13 at 5:22pm
post #13 of 25
Quote:
Originally Posted by imackler View Post

I've been reading this article here http://www.innerfidelity.com/content/nwavguys-heaphone-amp-measurement-recommendations to understand these measurements. But, seriously. It's Greek to me. Actually, I have an easier time with Greek. 

 

I've got a Qogir (E09K) to a HD600. I think it sounds good. But I have to decide if there is another amp (in the maybe up to $300) budget that I should try. Someday, I'll get to a meet in Seattle, but I haven't been able to yet. I have already read thousands of subjective impressions of amp pairings. I value them. I know what amp  I would purchase if someone gave me $500 or $1000 to spend on an amp.  But deep down, I'm kind of a skeptic. It's weird to me that the pattern seems to be that almost every time you spend more it sounds better. There are exceptions, such as when you build it yourself or if the engineer is banned. Then the cost is less important. But like I said, I'm skeptic. I think if I found a $300 amp at a garage sale for $25 it would be the best sounding amp ever. I don't like that about myself...

 

Anyways, would anyone be willing to help me understand what weaknesses the E09K/Qogi has with a 300 ohm load from Tyll's measurements? If you would be willing to point out to me an amp that does better in the weak areas, I would be so much further along in balancing subjective impressions w/ the objective data from these graphs. I would really appreciate it. 

 

Right now, I can see the differences as I flip back and forth between the two charts... I just can't assign a value or interpret them. 

 

Basically, I'm mucking up the sound science forum with my ignorance. Thanks for your patience :) 

The E09 has a rather high output impedance of 10 ohms, making it unsuitable for use with more sensitive headphones. Generally, you'd want to have the impedance of the headphones to be at least 8x that of the amp's output impedance, hence the term '1/8th rule'. Anything less would result in an impedance mismatch.

post #14 of 25

500 Hz Harmonic Spectrum (nonlinear distortion)

 

Instead of sending a signal with all frequencies into the amp we just send a single 500 Hz sine wave into it and analyze the output.

Ideally, all you should see is a narrow peak at 500 Hz, but amps are not perfect so you will see also some harmonics. Odd (3rd, 5th ..) harmonics are generally more audibly offensive than even (2nd, 4th ...) harmonics.

 

As long as these harmonics are low enough in amplitude everything's fine.

 

 

But there's more besides the fundamental and harmonics, there's also noise. This is shown better in the Noise Spectrum.

The noise spectrum shows what the amp is outputting when it really shouldn't output anything.

 

If the whole noise floor is high or there are some high peaks (especially around 3 kHz where our hearing is the most sensitive) then you will hear a hissing noise during quiet passages in your music. Can be especially annoying with highly sensitive IEMs.

 

 

IMD (intermodulation distortion)

Here we send special signals into the amp:

SMPTE: 60 Hz and 7 kHz sine tones

CCIF: 19 kHz and 20 kHz sine tones

 

To put things as simple as possible again: look for high peaks other than the input signal, they shouldn't be there.

 

 

 

Crosstalk

In the FR measurement we sent a signal that contains all frequencies into both channels and also measured both outputs. Here we send this signal into one channel and measure the other channel, and vice-versa.

Ideally, no signal will leak from one channel into the other. Luckily, our hearing doesn't need all that much separation between channels to decide that the sound is coming from one side only.


Edited by xnor - 7/4/13 at 5:47pm
post #15 of 25
Quote:
Originally Posted by Greenleaf7 View Post

The E09 has a rather high output impedance of 10 ohms, making it unsuitable for use with more sensitive headphones. Generally, you'd want to have the impedance of the headphones to be at least 8x that of the amp's output impedance, hence the term '1/8th rule'. Anything less would result in an impedance mismatch.

He's using a HD600 that has 300 ohms which is way beyond 1/8th and it measures ok with 150 ohms so it would measure even better with a 300 ohm load.

 

But still, something like an O2 would bring measurable improvements in like every area. Audibly / if it's worth it, imackler has to decide for himself.

New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Sound Science
Head-Fi.org › Forums › Equipment Forums › Sound Science › Tyll Hersten/Innerfidelity's 21 Amp Measurements...