[samsquanch]

Ok, before I start, and because two other threads that discuss the O2 have been locked, I'd like to say let's keep this fun, it's audio, let's enjoy it!  Let's keep arguing about whether or not to do something has merit or not, and if someone is as dumb as tree moss for doing such a thing, to a dull roar.  (eg. "Why did you replace partA with partB, do you even know how electricity works, doofus?")
 
Alright, well I've built two separate O2's, one that follows the letter of the BOM, and one that has some very slight modifications (ceramic capacitors in feedback loops replaced with polystyrene caps) which will also become my guinea pig for as long as my interest holds.  I also built myself a Beta22 almost a year ago, three channel, two cases.
 
Now I've spent a lot of time with my Beta22, and completely love it, loved it from the first time I played DSOTM through it.  The O2 on the other hand, I was not blown away by, I found it to be cluttered around 80Hz (open E speed picking), and a few other issues, but I don't hate the thing, it is my portable headphone amp of choice at the moment, and I've been spending a decent amount of time listening to it.  
 
What I said in the previous paragraph has caused a fair amount of bickering in other threads, something to the effect of me being biased, imagining things, or preferring the sound of a colored amp (I'm fine with that).  Not the worst insults ever thrown at someone, but I did take issue with the implication that I don't know how or what instruments should sound like.  With the Beta22 snare drums sound like standing next to someone hitting a snare drum, not a recording of a snare drum, speed picking on a guitar's open e string has dynamics and is not just turned into a buzzing guitar.  To me that's the difference between the O2 and Beta22.
 
Ok ok, enough about me...
 
Usually my minor gripes were responded to with some iteration of, "Do some measurements!"  Well now that I've built my second O2 I've decided, Hey WHY NOT!?!?!?
 
First test I've done is Total Harmonic Distortion + Noise (THD+N)
 
Test was conducted with an NTI Minilyzer and Minirator.
 
Minirator was producing a a -10dBu sine wave, all amps volume knobs turned up to their maximum.
 
Numbers in a grid like fashion (color separations are octaves):

 
Graph:

 
Nothing too crazy yet, the O2 fares better than the Beta22 up till 160Hz, then they're pretty much even all the way till 20kHz. And the polystyrene caps seem to have no effect on THD+N.  
 
NEAT!
 
Now I don't have a pile of test gear laying around my house, so I have to figure out what I have at work (I work for an AV company) and how to use it.  The more equipment I find, the more tests I'll do.  And as I am admitting right here and now, I may misuse a piece of test gear, if you feel that I am, and that it's corrupting my results, please let me know, I'll re-test!
 
Got suggestions for tests?  I'll do my best, bare in mind I don't have deep pockets for purchasing things, but I am pretty handy with a soldering iron, so if I can build something to use to test, let me know, I always like making new things!
 

 

[blockhead]

Nice post thanks for the testing. I wonder how you could test for the "clutter" or congestion as some people call it? It's obviously not caused by the kind of distortion you tested for. Can you tell the difference between them blind? I wouldn't usually ask but Sound Science and all Thanks again for the effort. 

 

[Head Injury]

Test for IMD, crosstalk, and of course frequency response.
 
I'd also test closer to a normal listening voltage. One of the amps could produce more distortion than the other there. NwAvGuy does his THD sweeps at 400 mV.
 
Those measurements are also remarkably close everywhere but the bass. Are you sure the equipment you're using isn't bottlenecking the measurements, so you're ultimately getting a read of the test equipment's THD instead of the amp's? And what is the load impedance?

 

[Willakan]

Yeah, my concern would be the resolution of the test equipment. THD is meant, with the O2 at least, to drop below 0.001% at 1khz into some loads, which needs some relatively snazzy equipment (doesn't help that I can't see the graphs from where I'm browsing - friggin' blocker). What load are you using, by the way? (Apologies if this appears on the graph I can't see!) Testing at 32ohms could be very revealing of differences, whilst testing them unloaded doesn't show much.
 
 
EDIT: Looking at the specs of your test equipment, I suspect that the performance of the O2 is better than the analyser! The Beta22 may be as well in at least a part of the spectrum.
 
EDIT2: Looking further, the distortion performace of the analyser (as tested in the "Downloads" section of the company site) seems somewhat inadequate - 0.025% THD is considerably higher than the THD of the O2 under most circumstances.

 

[mikeaj]

Indeed, what's the test load and output voltage from the amp?
 
Unfortunately, it looks like you're not getting interesting results yet because the NTI Minilyzer seems to not have good enough specs to really distinguish between the three.  The website says it has a residual THD+N of less than -74 dB (not too impressive) with unbalanced input at 0 dBu to +14 dBu.  Probably at lower levels there's more noise so the figure is worse than that.  The signal generator seems to be less of a limiting factor, at -90 dB or -96 dB THD+N for probably the best case, depending on the model.  Obviously it's much harder to make audio analyzers and signal generators with as high performance as headphone amplifiers, so it's not surprising that the gear may not be up to task.
 
http://www.nti-audio.com/Home/Products/Minstruments/MinilyzerML1/Specifications/tabid/151/Default.aspx
http://www.nti-audio.com/Home/Products/HandheldExelLine/MiniratorMR2/Specifications/tabid/159/Default.aspx
 
However, -74 dB is 0.02%, and you're not even getting results that low right now.  Certainly both amps should have THD+N below that amount into most loads.  Instrumentation is definitely not my forte, so I'm confused as to how or why you're seemingly picking up worse performance from the Beta22 at the low frequencies.  It shouldn't be that poor in reality, right?
 
 
I recommend loading the amps with around 16 ohms and a level of about 2V rms (~8 dBu; O2 will clip higher than about 2.2V into 15 ohms) to challenge the amps enough so hopefully one misbehaves enough for a difference to be picked up.  An IMD test should be a little more challenging than just a single-tone THD test.  This output level is in the sweet spot range for both the analyzer and signal generator anyway.
 
 
edit: oh wait I realized you said you were using -10 dBu.  Crank it up.  What's the gain of both amps?  Check the output voltages, in any case.  Good luck.

 

[stv014]

Quote:

Originally Posted by mikeaj /img/forum/go_quote.gif
 
However, -74 dB is 0.02%, and you're not even getting results that low right now.  Certainly both amps should have THD+N below that amount into most loads.  Instrumentation is definitely not my forte, so I'm confused as to how or why you're seemingly picking up worse performance from the Beta22 at the low frequencies.  It shouldn't be that poor in reality, right?

 
0.02% is pretty bad, even a good sound card with software signal generator and analyzer would be better. The poor distortion at low frequencies could be caused by many reasons. Is it possible that it is determined with some kind of FFT based analysis, that does not have enough frequency resolution at low frequency (I have run into this type of problem) ?
 
As for what is worth testing (most of these are already measured and reported on the NwAvGuy web site anyway):
  - frequency and phase response with actual headphone loads
  - maximum usable output voltage (with high impedance load) with low distortion (possibly with swept frequency)
  - maximum usable output current - basically the same test again with very low impedance load
  - output serial resistance and capacitance (if present)
  - noise (preferably also as a function of the volume control)
  - stereo crosstalk (it is usually low enough), and cross-distortion
  - THD vs. voltage and frequency
  - IMD (both 60 Hz+7 kHz and 19 kHz+20 kHz, preferably at different voltages, but definitely at the highest possible voltage where it is still low enough)
  - any other test signal you can come up with

, e.g. square wave response, intermodulation with low frequency square wave and high frequency sine wave, IMD with various frequency combinations/swept frequencies, etc. I have even tested a sound card by playing and recording actual music, and then creating a difference signal against the original signal convolved with the previously measured impulse response.
  - for DACs, try a square wave (possibly not just with 50% pulse width) that is rendered without anti-aliasing but has a cycle time of integer number of samples, with nearly the maximum amplitude (e.g. 90%), and the rest of the available headroom - 10% with the previous example - used by a high frequency (not an integer multiple of the square wave, e.g. 100 Hz square + 7050 Hz sine) sine wave. Bad DACs and sound cards that are prone to clipping peaks fail this test and produce high amount of intermodulation between the test signals.
Again, it is best to do the testing with actual headphone loads. Amplifiers tend to do much better without a real (especially low impedance) load.
 

 

[Willakan]

I can now see the charts: the THD for the O2 is ludicrously high compared to what NwAvGuy reported (0.25% with some frequencies!), but generally consistent with the residual THD of the analyser. The incredible similarity in results further implies that you are simply charting the limitations of the measuring equipment on a graph. As others have said, tough loads (I'm assuming that test was done with a load?) are the only way you could hope to make any even vague distinctions here.

 

[samsquanch]

Ok, first hurdle, Minilyzer.  It has an input impedance of 40k Ohms, so what if I were to order a 32.2 32.4 Ohm wirewound resistor and throw it in parallel to the input of the minilyzer, which should provide a load of ~32.17 32.37 Ohms?  Would this work, or am I thinking about this wrong?
 
I'll crank up the output of the minirator later today and see if that helps anything.  The co-worker that I borrowed the testers from showed me how to use them, and I don't think he fully knew what I was trying to do, so he set the output to -10dBu and said it should be good, I didn't know any better so that's where it stayed, oh well, test again!
 
I've seen people dismiss RMAA test results in the past, buuuut, are there any tests that the software can perform that are reliable?  I have a Lexicon Alpha Pro usb ADC/DAC that's fairly decent, I've used them for program recording in sports stadiums, not the most expensive piece of gear ever, but it's never let me down for what I use it for.  If I were to use this, how should I set it up to get reliable results?
 
I'm going to dig around the office some more and see if there's any other test equipment laying around that might be of use.
 
Thanks for the help, I want to keep this going, I'm having too much fun already.

 

[nick_charles]

 
Quote:

I have a Lexicon Alpha Pro usb ADC/DAC that's fairly decent, I've used them for program recording in sports stadiums, not the most expensive piece of gear ever, but it's never let me down for what I use it for.  If I were to use this, how should I set it up to get reliable results?  

The USB device you mention may still have more background THD+Noise than desirable - though it is rather better than the one I used for my cable tests

Line Inputs	(2) 1/4" TRS balanced or unbalanced
Input Impedance	20 kOhm balanced, 10 kOhm unbalanced
Maximum Input Level	+12 dBu
Frequency Response	+0, -0.2 dB 20 Hz - 20 kHz, ref. 1kHz
THD+N	<.009% A/D, 20 Hz - 20 kHz

 

 

[Willakan]

 
Quote:

Ok, first hurdle, Minilyzer.  It has an input impedance of 40k Ohms, so what if I were to order a 32.2 32.4 Ohm wirewound resistor and throw it in parallel to the input of the minilyzer, which should provide a load of ~32.17 32.37 Ohms?  Would this work, or am I thinking about this wrong?
 
I'll crank up the output of the minirator later today and see if that helps anything.  The co-worker that I borrowed the testers from showed me how to use them, and I don't think he fully knew what I was trying to do, so he set the output to -10dBu and said it should be good, I didn't know any better so that's where it stayed, oh well, test again!
 
I've seen people dismiss RMAA test results in the past, buuuut, are there any tests that the software can perform that are reliable?  I have a Lexicon Alpha Pro usb ADC/DAC that's fairly decent, I've used them for program recording in sports stadiums, not the most expensive piece of gear ever, but it's never let me down for what I use it for.  If I were to use this, how should I set it up to get reliable results?
 
I'm going to dig around the office some more and see if there's any other test equipment laying around that might be of use.
 
Thanks for the help, I want to keep this going, I'm having too much fun already.

The problem with RMAA is that it is inconsistent and various aspects of its functionality are opaque: the mysterious "calibration" process' effect on the measurements is not fully revealed whilst other tests can produce erroneous results upon occasion due to various intermittent bugs (hasn't been updated in a while!). Even with the best-case RMAA setup, the results are only really comparable with other RMAA tests done on that setup and even then the conclusions that can be drawn are pretty limited.
That said, assuming you're testing with a load you're probably safe with FR - but that will likely be spectacularly uninformative, as I would be astonished if either of the amps get within any real distance of audible deviation from flat.

 

[stv014]

Quote:

Ok, first hurdle, Minilyzer.  It has an input impedance of 40k Ohms, so what if I were to order a 32.2 32.4 Ohm wirewound resistor and throw it in parallel to the input of the minilyzer, which should provide a load of ~32.17 32.37 Ohms?  Would this work, or am I thinking about this wrong?

 
Ideally, you want to test with a headphone load, using a splitter cable to connect both the ADC and the headphone to the amplifier output. A simple resistor is better than nothing, but it does not reproduce the reactance, and even non-linearity that dynamic drivers have. Note that the headphone can also act as a microphone, even if a not particularly efficient one.
 
Quote:

Originally Posted by samsquanch /img/forum/go_quote.gif
 
I've seen people dismiss RMAA test results in the past, buuuut, are there any tests that the software can perform that are reliable?

 
I am not sure I can help too much choosing the software, but at least you should look for a good spectrum analyzer/oscilloscope/tone generator application. There are some reasonable freeware ones (e.g. Visual Analyser, but it does have some bugs), however the really advanced packages tend to be shareware/commercial. RMAA should not be dismissed entirely, but you should take its - and any other potentially unreliable program's for that matter - results with a grain of salt, and preferably try to verify them with other software as well. It is not bad for quick basic testing if you are aware of its limitations and make sure that the tests are performed correctly (e.g. no Windows system sounds played on the audio output device while testing, no buffer over- or underruns, input and output levels are set correctly, etc.). Of course, any good wave editor can be useful for examining the recorded signal, both in the time and frequency domain, and possibly also for tone generation. But for generating special test tones, I use programmable software synthesis, although I am unsure if that is useful for you.
 

 

[mikeaj]

If you test with headphones, be sure to note the max input they're intended to handle.  Most 16 ohms IEMs are not going to take kindly to 2V rms sine wave testing, for example.

 

[samsquanch]

Didn't think I had given up yet, did ya?
 
Ok, so I've been a little busy with work, but I did some more talking and more digging, and found another piece of gear in our shop, it's another piece of gear from NTI, it's the A1  (http://www.nti-audio.com/Portals/0/Products/Discontinued/A1_Manual.pdf)
 
From the manual, on THD+N:
 
 
THD+N Max:  < 0.005% (-86dB) or 25µV
______ Typ:  < 0.0025% (-92dB)
 
The input is 100kOhms, but I'm going to build a box that has a headphone in, headphone out, and an rca for each channel to connect to a tester, which should, with headphones (32ohm) plugged in, drop the impedance to just under 32ohm.
 
It can also test for crosstalk, with an accuracy of +/-.3dB
 
Well, there's that to perk up your ears again.  As always, let me know of any questions, concerns, suggestions you may have, I'm hoping to do this test tomorrow evening (if I remember to drag my Beta22 in to work...)

 

[Willakan]

That seems quite a bit better: still not sufficient to fully showcase the O2 (

), but hopefully we should start seeing some difference in loaded tests between amps!