For whoever's interested...

http://jusstuff.home.comcast.net/AmpComparison.htm

With soundcard baseline:

http://jusstuff.home.comcast.net/AmpCardComparison.htm

Full thread at Head-Fi (more explanations, etc.)

http://www4.head-fi.org/forums/showthread....&threadid=48404

Very interesting. If I compare these results to soundcards' I get the impression that the design of these amps is focussed on some aspects of 'quality' but others weren't taken that seriously (-> frequency response).

AFAIK the 'soundstage' we hear depends a lot on high frequencies and their reflection by the ear. Because of this it's possible that the 'soundstage issues' you hear are caused by the amps' frequency response and not by stereo crosstalk.

It would be interesting to set up your audio player's equalizer (or convolution plugin) in a way that frequency response of the combination soundcard+amp becomes flat - and find out if soundstage issues are reduced.

It seems the frequency response variations are only by a few dB though. Keep in mind too that headphone amps are made to amplify to a much higher level than just an analog loopback of the card itself with a relatively small amount of amplification on line out.

Also, there's no load with the soundcard loopback, where this test with the amps was done under load of a pair of headphones (300 ohms)... lots of variables there that could account for at least some of the differences.

I find the differences between the two amps the most valid in terms of comparing, because the situations are very similar there -- whereas a straight analog line out --> line in on a soundcard is very different than the amp tests.

IMD with the Creek amp is pretty bad (almost 0.3%)... the Meta42 sounds "cleaner" to my ears and that could be one reason. Not sure about your 'soundstage' observations, but it could be correct -- I find that my ears are quite sensitive to variations in stereo crosstalk with headphones though. Even just a small amount of crossfeed (the Meta42 has one built in, switched off for the tests) sounds "mono-ish" to me.

Edit -- here are the results I got with Cool Edit Pro's "statistics" function (44.1KHz/32bits recording about 10 seconds), with the amps under no load and measuring only the noise floor of each through the analog line in of the soundcard:

Creek OBH-11:

Peak Amplitude: -80.61 dB -81.01 dB
DC Offset: 0 0
Minimum RMS Power: -95.74 dB -95.5 dB
Maximum RMS Power: -91.93 dB -92.91 dB
Average RMS Power: -94.84 dB -94.62 dB
Total RMS Power: -94.82 dB -94.61 dB

Meta42 ("DoobAmp"):

Peak Amplitude: -81.56 dB -83.3 dB
DC Offset: 0 0
Minimum RMS Power: -96.03 dB -97.69 dB
Maximum RMS Power: -95.2 dB -96.47 dB
Average RMS Power: -95.62 dB -97 dB
Total RMS Power: -95.61 dB -97 dB

There was little or no variation in noise floor (e.g. hiss) with the volume pots of either amp down all the way vs. turned up to full volume.

A while back I've done some tests with different crossfeed software/plugins. IIRC typical values were -2 ... -20 dB (frequency dependant) for amplification of the signal added to the other channel. That's much more than the "Stereo Crosstalk" values your tests showed. This is why I think soundstage differences could be caused by 'equalizing' = non-linear frequency response.

Could be ... will take your word for it -- I never tried any tests like that.

The op-amp I'm using in the Meta42 amp is supposed to be a really good/accurate one for audio (Analog Devices 8620)... some DIY headphone amp builders are using older op-amps like the Burr-Brown OPA602BP and OPA637BP (ridiculously expensive too! Try $22.97 apiece/per channel for the OPA637 from Digikey) with these amps, apparently for the euphonic colorations or whatever they hear from these...

Ehm... according to these measurements, its quite possible that the headphone output of cards such as the Revo or the Santa Cruz measures better than those dedicated headphone amps. I mean, even when driving similar headphones.

Edit: maybe I'll post some measurements when I have time.

Edit -- the headphone output of those cards? Would definitely be interested to see some measurements, and also if they can drive both high and low impedance headphones equally well. I think measurements should be taken while driving a "real world" load, which I did here (used a splitter cable to drive headphones while testing).

P.S. circuit info for the Meta42 amp I tested is publicly available on the Net, here (there are some amp-specific things, of course):

http://www.tangentsoft.net/audio/meta42/

Yes, I meant driving headphones. The Revo will do better in case of high impedance headphones, since it doesn't have a very high output current capability. I believe the Santa Cruz is better in this respect. However, some people find the Revo output to be a little too low volume when driving HD580/HD600. I don't, I guess it depends on what you are used to.

If you (or someone) can demonstrate that I'm getting poor performance compared to what can be had, I'll sell these amps and find something else... I definitely want something measuring/performing as it should, particularly given that neither of these were all that cheap...

Too bad there isn't really a mass market for dedicated headphone amps... it's really a hit and miss kind of thing when all that's available is either musician/studio gear or "audiophile" stuff.

Yes, I know about the META42 design. However, I think the use of expensive opamps from Analog Devices, or A-biasing these, is useless, more given the measured performance of this amp. A cheap 5532 op-amp in standard configuration should be more than enough (this is the opamp M-Audio cards and many pro audio devices use).

BTW, the frequency response problems in this amp are strange. Problems at the Creek amp seem to be caused by a too high output impedance, since the frequency response curve more or less follows the HD600 impedance curve.

Another problem at the measurements is the very noticeable hum interference, more at the Creek amp. This could be just a problem of the testing setup causing a ground loop, this is something not unusual in these kind of tests of "external" hardware. In that case it won't be representative of actual performance. However, it can't be said for sure to be the case.

Edit: corrected explanation over frequency response differences, I mixed amps responses.

Very good performance is possible to achieve for any kind of headphone, just using a simple voltage divider with the adequate resistor values, at the speaker outputs of a minimally decent receiver or speaker amp. This is the system I usually use, and recently I saw some people at head-fi recommending it (Joe Bloggs for example).

There was some speculation on the Head-Fi thread that the frequency response variations of the Meta42 could be caused by the crossfeed circuit -- who knows.

Anyway, except for that bass hump on the Creek (which I think could and did have an audible effect on music) it seems mostly academic to speculate what's causing this with either amp... bass is down ~1dB at 30 Hz and treble is down ~1.25dB at 20 KHz... probably not anything that's going to seriously affect listening, I wouldn't think.

Sure, it's possible.. then again if there's not a speaker amp or receiver around conveniently (for example if one doesn't use speakers and/or doesn't listen to radio), imo it's still worth it getting a dedicated headphone amp. Also, dedicated headphone amps tend to be much smaller and more compact (and in my case with the soundcard source, perfect for putting on top of my PC's mid tower case and adjusting volume easily/conveniently).

Anyway... I think the Creek amp may be going on the market... that %IMD figure is really ugly, and I'd noticed it was making many of my recordings sound unreasonably poor -- subjectively speaking, almost like there was a "goop" in the background of the recordings, around the level of the noise floor.

Ah, so that's why I saw this thread come up in my "Joe Bloggs" search... Yes I'm recommending it, but then I'm a poor guy I've definitely heard better amps than what I'm using right now (Onkyo minisystem headphone out->impeder->headphones): MG Head, Sugden Headmaster, Twin Head... Come November, I'll take my setup out to the headphone meet in HK to compare with bigshot amps like these again. I don't think I'll be pleased



You do those measurements and then you can talk

fewtch: I presume that you used an elpac PSU for the creek. I would really like to see results with OBH-1 in comparison, just to see how big of a difference it really makes in technical measurments.

I'm really glad to see that you've taken the time and done the measurments and hope to see you test out other amps in the future aswell.

Those results don't really tell me anything new. Knew that Creek wasn't the highest quality amp... and always thought that it has a little bass hump. Now I know.

I did an earlier test with OBH-1 vs. Elpac (not using RMAA though):

http://www4.head-fi.org/forums/showthread....&threadid=43771

Earlier this morning, I made up a 'quick & dirty' battery pack (two 9v NiMH batteries wired in series) for the Creek amp. In my opinion the sound is further improved with battery power, but I haven't done any measurements (so at this point it's only a subjective claim). The amp seems to be outrageously sensitive to power issues and apparently has little or no power filtering in the circuit.

BTW, take a look at the RMAA graphs again -- notice the peaks at 100 Hz (and subsequent harmonics)? That's my monitor refresh rate. Apparently both these amps are subject to some interference from close proximity to my PC and monitor.

I sold the particular Meta42 I used to do the tests here with (it was overkill for my needs) but have another one on the way within a week or so.

This "budget high end" stuff is turning out to be a lot of fun, even if it's overkill in the opinion of many on HA. Next step is to pull out a soldering iron, buy some parts and try building a few simple amp designs. If the cheap/simple designs measure and sound as well as the expensive ones, it will clear up some questions in my mind.

I would have hoped a comparison to see, how it affects the poor IMD present. Does it make it worse or not. In otherwords, is the IMD issue power related or not. I have the OBH-2 wall wart (in europe it is delivered by default with OBH-11, luckily ).

I've read the OBH-1 vs elpac thread in head-fi aswell, and that thread should be read by everyone who owns the amp. But it ain't a problem to me obviously.

The results with the batteries would be great aswell, though.

OK, I'll try a quick RMAA test on battery power...stand by for a link to results (will post it in this message). I'll also turn off my monitor this time when doing the test to see if it gives a better overall noise measurement.

Edit -- here are the new OBH-11 results. Unfortunately I forgot to compare them with the old OBH-11 results, or even save the results of the new test as a .sav file I'm not gonna hook all the cables again, so you'll have to load both webpages side by side and compare that way:

http://jusstuff.home.comcast.net/OBH-11.htm

(old test: http://jusstuff.home.comcast.net/AmpComparison.htm)


Note again that my monitor was turned off this time. It seems most of the measurements improved, except intermod distortion (still not very good) and frequency response. Particularly noise floor/dynamic range improved a lot, and I think that's because of the battery power.

P.S. sorry for my poor testing procedures. I should repeat both of these with the monitor either on or off for both, and post a comparison of battery power vs. regulated wall power (but only if you're really interested -- hooking/unhooking the cables is something of a PITA).

That pretty much shows what I was looking for... the IMD isn't a power related problem. But nice to see how much the SNR and DR improved by that tweak of yours.

The 30-300hz hump is slightly flatter with the batteries... that is probably caused by slight undervoltage, as might the slightly higher THD aswell. (Just guessing though)

But flatter FR, lower SNR and DR; what else could you hope for by just changing the power supply? B)

Tests of headphone output vs normal output in Terratec Aureon 7.1:
http://astralstorm.servebeer.com/aureonrmaa/

(my 'server' can be down)

I just discovered some bad news with the battery pack -- the amp is clipping at high volume levels (peaks with orchestral music, etc). I guess maybe I should try three 9v batteries in series instead, as NiMH 9v don't generally deliver a full 9v... or maybe it's not worth it for just a bit of change in noise floor and dynamic range.

Edit -- I just measured my battery pack with a multimeter and it's putting out 14VDC (unloaded)... apparently not enough for this amp at higher volumes. I guess it's back to the regulated wall wart for now... I'll try alkalines instead though, for the fun of it.

Edit2 -- Alkalines fixed the problem... cheap too (on sale $2.99 for a 2-pack).

Here are some tests of the extremely expensive Millenium HP1:
http://medlem.spray.se/urt/Millenium%20HP1/WaveOut.htm

More info about this "audiophile extraordinaire" equipment can be found here:
http://www.netzmarkt.de/thomann/thoiw5_mil...b98385943341bb6

Edit: The tests were performed like fewtch's (driving a pair of 580s while measuring).

How did you perform the tests? Plug the headphones in, and plugged the wire going to line-in to where? Just thinking about conducting my own tests.

The amp has two headphone jacks. I plugged the 580s into one jack and a cable connected to the line in on my Revo to the other jack.

Edit: here are tests using the same jack with a split cable driving the 580s while also feeding the line in:

http://medlem.spray.se/urt/Millenium%20HP1...lit%20Cable.htm

fewtch: How did you measure the results?

So that we'd know which of those results would be atleast almost comparable.

I'm not sure what you're asking... if you mean test procedure, I used a splitter from the headphone out jack to go to a pair of headphones (HD600) and also to the soundcard's input jack.

That is just what I wanted to know.

Revo measurements, including soundcard baseline, at http://www.kikeg.arrakis.es/measurements/R...lutionHD580.htm

As you can see, frequency response is much better, but, as it was possible to happen, IMD value is not very good, and worse than in DoobAmp, due to low-freq. distortion, which is due to lack of power reserve.

Frequency response problems at both headamps would be clearly audible, but low freq. distortion at Revo output is not very likely to be, in my opinion.

With the SantaCruz soundcard, which seems to have a dedicated transistor output stage (Revo has just 2 paralelled opamps), it's possible that distortion figure is better.

And, using the resistor voltage divider at the output of a decent speaker amp would yield nearly perfect measurements.

Edit: however, if poor frequency response at the DoobAmp is due to crossfeed circuitry, quite likely it would be fixed just disabling or removing this crossfeed.

I mixed frequency response curves at the original post, I just edited it. The problems at the Meta42 are very likely to be caused by the crossfeed circuitry, as you say. The problems at Creek amp are the ones due to a too high output impedance.

KikeG, what amp do you have? Have you tried doing tests with a split cable?

Wouldn't this rise the output impedance level and negate the benefits of having a good headphone amplifier (besides 'I can play LOUD man!' one)? AFAIK one would want output impedance as low as possible.

I got rid of the "DoobAmp" (sold) and got another Meta42 instead... it seems like a good design in a headphone amp (albeit on the expensive side). Hasn't arrived yet, but will test it eventually when it does.

Today I did something interesting (and risky) -- removed/jumpered the headphone jack output resistors (470 ohms per channel) from my ca. late 80's Sony TA-AX205 speaker amp. This evened out the frequency response of my HD-580s greatly, but now I have to be careful plugging/unplugging the jack, and make sure to do it with power off (so it won't be shorted out even briefly). Also, the range of the volume control is more restricted. No big deal on either point, it was well worth it... sounds better, bass hump is gone, neutrality is greatly improved.



Picture of headphone jack from inside with resistors removed/jumpered. The remaining resistor grounds the headphone jack to the case.

Make sure that you wait at least 20-30 seconds after turning the amplifier off to allow the big power supply filtering capacitors to discharge (if the amp design doesn't completely cut off power supply to output stages AFTER the capacitors).

The amp contains a relay which I believe shuts off almost immediately when power is turned off. It takes about 4 seconds to engage after power on (which I assume happens after the PSU filter caps charge up -- they're really large ones).

I don't understand what you mean. Do you mean speaker amp?


My measurements of the Revo output were done with a splitter connector at the Revo output. That's the only way of evaluating real output performance when driving headphones.

Not if you pick adequately the resistor values. An output impedance equivalent to a few ohms is easy to achieve. The 20 ohm/2 ohm divider, for example, works well in this sense. The only thing you need to be careful is that the 20 ohm resistor is rated several watts.

I thought you performed a headphone amp test.



I disagree. Normal listening does not include a splitter cable that leaches current away from the headphones (Kirchhoff). I agree it can be useful to see if the amp can take the load.

HD580 impedance is ~300Ohms. Input impedance is 10Kohms+ so after a Kirchhoffian calcualtion you'll realize that the input added load is negligible.

I see.

lucpes is right. As to my test, I just measured performance of my Revo built-in headphone output.

Those were some damn nice results. Did you use your Audiophile to do the recording? I can't get anywhere near those results when doing a loopback with my Revo:

http://medlem.spray.se/urt/Revolution/2444.../Comparison.htm

No, I used the Revo input. If I use the Audiophile input, I get a nice ground loop that causes a lot of low-freq. noise.

As to your measurements, they show low frequency hum noise (50 Hz and its harmonics). It could be caused by a long cable or defective connector picking this 50 Hz hum noise. Try using a different, shorter cable, and move it around till the hum dissapears or gets lower. If it doesn't, it could be kind of an internal ground loop caused by the same card, maybe a semi-defective unit? Still, the noise is quite low level, and may happen just in loopback tests, not when playing music.

I upgraded to RMAA 5.1 and cleared the area behind the computer from electrical wiring as much as possible, and now my results are more in line with yours.

http://medlem.spray.se/urt/Revolution/RMAA.../Comparison.htm

I tried rmaa-ing my Creek 11se w/obh-2. I don't have a splitter so I'm wondering if my results therefore are not comparable with yours...?

M-audio revo with 3.5mm -> 2xrca connected to the amp, AKG K-501 connected to output 1 and output 2 connected back to the revos line-in.

The results