HydrogenAudio

CD-R and Audio Hardware => Audio Hardware => Topic started by: markanini on 2011-12-30 10:23:04

Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2011-12-30 10:23:04
Motherboard is a Gigabyte Z68XP-UD4.


Line out to line in:
http://nerdmoments.byethost7.com/24_44_1.htm (http://nerdmoments.byethost7.com/24_44_1.htm)
Output set at "50", input at "100", assumed to be nominal levels(?).

Headphone out at max output, unloaded, to line in. (Input reduced to "25"):
http://nerdmoments.byethost7.com/24_44_1_HP.htm (http://nerdmoments.byethost7.com/24_44_1_HP.htm)

I can do some loaded test of the headphone pre-amp if anyone's interested. I have Ety HF5, Beyer DT250-80, Senn PX-100,  Koss KTXPro1, Senn HD201 and Creative EP-630 at hand. I assume I need a 3,5 mm Y splitter with one male and one female connector at the split ends?

Note that while output is 24-bit capable input is 16-bit(oddly, both are 24-bit according to datasheet).
Title: ALC889 RMAA results at 24/44.1
Post by: xnor on 2011-12-30 11:02:26
I assume I need a 3,5 mm Y splitter with one male and one female connector at the split ends?


Or a normal Y splitter (2 female ends) and a loopback cable. I'd be interested how measurements change with ~30 ohm headphones as load.
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2011-12-30 11:15:06
I assume I need a 3,5 mm Y splitter with one male and one female connector at the split ends?


Or a normal Y splitter (2 female ends) and a loopback cable. I'd be interested how measurements change with ~30 ohm headphones as load.


That would be the PX100(32Ohms). I'm picking up a splitter today.
Title: ALC889 RMAA results at 24/44.1
Post by: xnor on 2011-12-30 14:31:01
Alright. The audio chip is a Realtek ALC889 right? (typo in the thread title)
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2011-12-30 16:10:30
Yes.
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2011-12-31 06:44:53
Results for PX100:
http://nerdmoments.byethost7.com/24_44_1_PX_100_66.htm (http://nerdmoments.byethost7.com/24_44_1_PX_100_66.htm)
http://nerdmoments.byethost7.com/24_44_1_PX_100_61.htm (http://nerdmoments.byethost7.com/24_44_1_PX_100_61.htm)
http://nerdmoments.byethost7.com/24_44_1_PX_100_43.htm (http://nerdmoments.byethost7.com/24_44_1_PX_100_43.htm)
http://nerdmoments.byethost7.com/24_44_1_PX_100_32.htm (http://nerdmoments.byethost7.com/24_44_1_PX_100_32.htm)

Would love to hear your comments, xnor.
Title: ALC889 RMAA results at 24/44.1
Post by: saratoga on 2011-12-31 06:53:19
Those look really bad.  I'm assuming thats the headphone out and not line out?
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2011-12-31 08:32:40
Some sort of best case result with Etymotic HF5: http://nerdmoments.byethost7.com/24_44_1_HF5.htm (http://nerdmoments.byethost7.com/24_44_1_HF5.htm)

I guess the conclusion one could draw is that the line-out is clean, no less clean than pro-audio interfaces I've used(!), but better to use a separate headphone amp if signal quality is a concern.
Title: ALC889 RMAA results at 24/44.1
Post by: ErnestS on 2011-12-31 10:01:14
Make sure you change in realtek software "speaker configuration" from speakers --> headphone.

Here is my RMAA measurements of ALC888. I did a loopback with a y-splitter cable and AKG K530 attached.

The order is:
ALC888 + AKG K530 + headphone option in realtek software "speaker configuration"
ALC888 + AKG K530
ALC888

PDF (http://www.mediafire.com/?o3abqhnbt6hrk6f)

Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2011-12-31 11:33:42
For the sake of comparison split loopback driving Etymotic HF5 through a dedicated heaphone amp built by a guy named Knut: http://nerdmoments.byethost7.com/ALC889+headphone%20amp.htm (http://nerdmoments.byethost7.com/ALC889+headphone%20amp.htm)
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2012-01-01 05:03:45
Again for comparison, Sansa Fuze. Totally beats previous results for frequency responce THD and IMD:
http://nerdmoments.byethost7.com/Fuze (http://nerdmoments.byethost7.com/Fuze) -6dB.htm
(Driving Etymotic HF5)

I think with something like a Fiio E5+ALC889 should be able to produce very high quality sound to headphones. One certainly doesn't need to spend big $$$.
Title: ALC889 RMAA results at 24/44.1
Post by: xnor on 2012-01-01 11:01:28
Indeed, the problems seem to be output impedance (-> deviations from flat FR depending on impedance curve of the headphones) and IMD + noise when driving a load. As a source it should work pretty well.
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2012-01-01 15:02:48
Yeah, high output impedance.

For fun here's the most ambitious looking on-board I've seen:(http://assets.vr-zone.net/12863/g1_sniper_2_3.jpg)
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2012-01-01 15:47:31
I googled and couldnt find a straight answer to this one: Is there an easy and accurate way to measure output impedance on and headphone output with a multimeter? Depending on where I place the prods at the TRS I get 8, 16, and 32 Ohm on my dedicated amp.
Title: ALC889 RMAA results at 24/44.1
Post by: xnor on 2012-01-01 18:34:12
Hmm, from 1.5 kHz to 70 Hz there's a difference of about 4 dB with the PX100 (impedance ranges from about 34 to 60 ohm) so the output impedance seems to be a lot higher. Could be as high as 100 to 200 ohms?

I think you could play some 50-60 Hz sine wave (e.g. use Audacity to generate them) and measure ac voltage without and with load (e.g. 30 ohm resistor). Then calculate Zout = (Rload * (Vnoload - Vload)) / Vload. Not the most accurate way however.
Title: ALC889 RMAA results at 24/44.1
Post by: saratoga on 2012-01-02 17:03:41
I googled and couldnt find a straight answer to this one: Is there an easy and accurate way to measure output impedance on and headphone output with a multimeter? Depending on where I place the prods at the TRS I get 8, 16, and 32 Ohm on my dedicated amp.


You have to measure the voltage with 2 known resistances and then solve for the output impedance. 

See:  http://en.wikipedia.org/wiki/File:Source_a...d_circuit_Z.png (http://en.wikipedia.org/wiki/File:Source_and_load_circuit_Z.png)

You would control ZL, measure VL, and then solve for ZS.
Title: ALC889 RMAA results at 24/44.1
Post by: ErnestS on 2013-03-15 08:52:39
I forgot about this topic.

Yeah, high output impedance.

According to the spec sheet (http://www.realtek.com.tw/downloads/downloadsView.aspx?Langid=1&PFid=28&Level=5&Conn=4&ProdID=173&DownTypeID=1&GetDown=false&Downloads=true) the amplified output impedance for ALC889 is 2Ohms. Line out is 200Ohms.
Title: ALC889 RMAA results at 24/44.1
Post by: skamp on 2013-03-15 09:06:44
The output impedance isn't just tied to the DAC chip, it will depend on the implementation on your PC / laptop.

An easier way to measure output impedance is to play and record a sine wave at 1 kHz, without a load, and with a load of known impedance (like headphones with a stereo splitter). Determine the RMS value of both recordings (e.g. with "sox file.wav -n stat -rms"), then use the following formula:

Zout = (Zhp * (RMShigh - RMSlow)) / RMSlow

You can adjust that value by taking into account your ADC's input impedance, if you know it:

Zout = 1 / ((1 / Zout) - (1 / Zin))

I've managed to make reliable measurements that way (verified by comparing my results to other results available online). Don't ask me about the math, I got it from stv014, but I know it works.
Title: ALC889 RMAA results at 24/44.1
Post by: stv014 on 2013-03-25 17:22:21
According to the spec sheet (http://www.realtek.com.tw/downloads/downloadsView.aspx?Langid=1&PFid=28&Level=5&Conn=4&ProdID=173&DownTypeID=1&GetDown=false&Downloads=true) the amplified output impedance for ALC889 is 2Ohms. Line out is 200Ohms.


The datasheet also recommends (in the "Application Circuits" section) 75 ? serial resistors (and also 100 uF capacitors) on the headphone output, so an implementation that follows the reference design would have an output impedance of 77 ?.
Title: ALC889 RMAA results at 24/44.1
Post by: xnor on 2013-03-25 18:45:47
Z77 Extreme 4 motherboard - I measured about 80 ohms for the front panel headphone output and 200 ohms for the line out in the back.

Some semi-pro/home recording audio interfaces have similar output impedance for the headphone out, or lower (down to about 20 ohms).
Title: ALC889 RMAA results at 24/44.1
Post by: Arnold B. Krueger on 2013-03-25 19:02:14
Motherboard is a Gigabyte Z68XP-UD4.


Line out to line in:
http://nerdmoments.byethost7.com/24_44_1.htm (http://nerdmoments.byethost7.com/24_44_1.htm)
Output set at "50", input at "100", assumed to be nominal levels(?).


In other measurements I've found that a lot of chips used for on board audio interfaces have asymmetrical performance. IOW their ADCs aren't up to the quality levels of their DACs.

I recently did some testing of an ALC 887 chip on a Gigabyte system board and when looping it back through itself, found results similar to those posted above.

I then redid the tests looping the ALC 887 output through an M-Audio AP 24192 input - knowing that this interface has native performance with up to 110 dB dynamic range.

I obtained the following results:

Frequency response (from 40 Hz to 15 kHz), dB:   +0.01, -0.08   Excellent
Noise level, dB (A):   -92.4   Very good
Dynamic range, dB (A):   92.4   Very good
THD, %:   0.0028   Excellent
IMD + Noise, %:   0.030   Good
Stereo crosstalk, dB:   -89.8   Excellent
IMD at 10 kHz, %:   0.0093   Very good

While I'm at it, I discovered a way to circumvent a problem that sometimes arises while testing different audio interfaces mixed and matched.

In this test I was never able to get a recording level of greater than -12 dB, which the Rightmark program had problems with.  I simply had the Rightmark program save the wave file that it recorded, normalized it with Audacity, and then analyzed it.  Since the noise level of the AP 24192 card is about 110 db below its normal FS input, the test was run under a condition where this was effectively 98 dB below the tests, and so errors due to the circumvention were still well below the dynnamic range that was observed.
Title: ALC889 RMAA results at 24/44.1
Post by: markanini on 2013-03-25 21:19:09
Try setting the main output to headphone mode. I remember gettting more volume and better rightmark results in loopbaack mode that way.
Title: ALC889 RMAA results at 24/44.1
Post by: Arnold B. Krueger on 2013-03-27 13:52:48
Tried setting output to headphones instead of speakers and obtained:

Frequency response (from 40 Hz to 15 kHz), dB:   +0.01, -0.07   Excellent
Noise level, dB (A):   -92.5   Very good
Dynamic range, dB (A):   92.6   Very good
THD, %:   0.0028   Excellent
IMD + Noise, %:   0.029   Good
Stereo crosstalk, dB:   -90.1   Excellent
IMD at 10 kHz, %:   0.0092   Very good
Title: ALC889 RMAA results at 24/44.1
Post by: db1989 on 2013-03-27 13:58:04
How not to post (http://i.imgur.com/zL5ere3.png)
Title: ALC889 RMAA results at 24/44.1
Post by: stv014 on 2013-03-27 15:52:33
ALC887 line output (rear jack) on an ASUS motherboard. The front panel is not connected, because it is used by a sound card, so I cannot test the headphone output, but I measured it earlier to have an output impedance of slightly more than 75 Ω.

Output impedance (http://i45.tinypic.com/14qkg1.png)
Output impedance phase (http://i49.tinypic.com/2nroqc6.png)

The output impedance of ~200 Ω is consistent with the datasheet. There are apparently also 10 uF capacitors on the output (calculated from the output impedance and the frequency at which the phase is -45 degrees).

THD vs. level at 3 kHz unloaded and with 220 Ω load (http://i47.tinypic.com/1zc3hi0.png)

With the low impedance load, the line output clips at high level. This probably would not happen on the headphone output.