Blind Listening Tests & Amplifiers

Status
Not open for further replies.
nw_avphil,

I´ts easy to tell most stuff apart, but of course it´s possible to find stuff that sounds very similar/identic, it very much depends on the set up.

CD players and cables measures way better than amps and I have "blind picked" lots of ´em.

Some of the better amps I´ve listened to at home and at friends is;

GamuT D200
Aleph 5
ARC VT100
LC Audio Zapsolute and Patriot
Muse
Holfi Integra

those have been compared to cheap (but with good specs) stuff like Arcam, Yamaha, Luxman...

All those have very good specs and sound "extremely" different.

Answer to your question: most likely, yes!

/Peter
 
Let's measure the difference!

I have an idea that would allow everyone to 'listen' to these differences between amplifiers.

I recently purchased a soundblaster audigy sound card that has a 24 bit A/D and a greater than 100 dB signal to noise ratio. It seems to me that a standard measurement technique could be developed to allow me to record the output of any amplifier with enough precision to hear the difference, if it is there, in the recording.

I'll take a shot at it.

1) Connect the line level input of the sound card directly to the speaker terminals. ( The speaker provides a realistic load for the amplifier, of course the same speaker should be used each time )

2) Set the system to output the test music or other repeatable signal at say 1 volt p2p, ( some level that uses 90% of the input range of the sound card )

3) Record the test passage at a high sample rate into an uncompressed wave file.

4) Save the wave file for each amplifier tested.



This would allow us to easily compare many amplifiers, and have a record of how each sounded on an identical passage of music.

Also it allows the direct comparison of the individual files mathematically. For instance two files could be subtracted - to yield the difference - allowing you to listen to the difference.
 
diyAudio Senior Member
Joined 2002
SOUNDCARD.

Hi,

1) Connect the line level input of the sound card directly to the speaker terminals. ( The speaker provides a realistic load for the amplifier, of course the same speaker should be used each time )

You mean to connect the speakerterminals of and amplifier to the linelevel input of the soundcard?

How does the speaker come into play here?

Sorry, but I don't really get the setup.

Cheers,;)
 
Record the difference

Great question.

The set up would be

1) Connect the amplifier to the speakers as usual. This provides a load for the amplifier.

2) Connect an additional cable between the amplifier output terminals and the line level input on the sound card. Most sound cards have a miniature stereo headphone type ( 1/8 inch ) input. So a cable that fits that input is required. The other end could just be bare wires / or what ever is required to connect easily to your amplifier output.

Of course, one must use care to not play the amplifier too loud in this setup, as the input of the sound card could be damaged.


Does that help?
 
diyAudio Senior Member
Joined 2002
RE:Record the difference

Hi,

Of course, one must use care to not play the amplifier too loud in this setup, as the input of the sound card could be damaged.

Thanks, I thought that was what you meant.

Actually a low impedance (about 8 Ohm) L-pad high wattage attenuator could be used in series with the linelevel input to test the amp at higher power levels too.

Unfortunately these don't come cheap though.

Cheers,;)
 
diyAudio Senior Member
Joined 2002
MISMATCH?

Hi,

Why not simply use a high impedance voltage divider?

Well, I have no idea about the input impedance of the linelevel input of the average soundcard but I guess it would be around 10K.

Being a tube guy, high impedance means 100K+ to me so I doubt those values would be any good.

Feel free to clarify your POV though.

Cheers,;)
 
Level match

Yes, That sounds like a good idea. I would look at using something like 2 kohm pot. That would put an insignifigant load on the amplifier ( at 28 Volt it would pull 0.4 Watts ), and the low value of resistance would not add much noise. The thermal noise of a 2 Kohm resistor is around 0.8uV, so that's going to limit the measurement accuracy to 20 bits at 1V p2p. I think it's pretty easy to find a 1Watt potentiometer.
 
First, a 8 ohm load in paralell with speaker makes the situation worse for the outputstage of the amp.

If I get the idea right it´s about sensing the voltage at the speaker terminals, so, a high impedance divider wíll not load the outputstage of the amp and we do not have to worry about power dissipation in resistors.

A 10-100kOhm voltage divider (one series, one paralell resistor) would do the job, scaled according to the actuall level of amplitude that the amp drives the speaker with.

Even though the idea is nice, I think I´ll stick with listening to the amps directly via the speakers :)

/Peter
 
Divided

The problem with adding a divider is that the resistors can add more noise. For instance, a 2 Kohm resistor has 0.8 uV of thermal noise at 25 deg C, and a 20KHz bandwidth. If the sound card had a S/N of 106 dB, for a 1.0 volt measurement that puts the noise at 5.0 uV. So this value resistors would not degrade the measurement.

A 2Kohm pot would dissipate about .3 Watt when the amp was running at 100W output in to 8 ohms. Since 1 Watt pots are readily available this seems OK.

The point of coming up with a standard technique would be that the .wav files could be exchanged between us. It may only be practical to compare files created with identical components. I'm going to give it a try to see if I can differentiate between my Adcom standalone amp, and the cheaper amp in my Pioneer reciever.
 
Re: NOPE...

fdegrove said:
Sorry, but that won't cut it...it would render all measurement invalid.

Things aren't just that simple.

Ciao,;)
They're not? That's what the volume control does on the input of any amplifier and those are "audiophile approved", right? It's also essentially what the input attenuator does on audio test equipment including distortion analyzers and oscilloscopes. Can you explain how or why a simple passive divider network to attenuate the voltage across the speaker would "render all measurement invalid"???

I also have to point out that if the experiment ends up showing you cannot hear a difference between two amplifiers, the more esoteric audiophiles here will VERY likely blame it on your soundcard, the A/D process (not to mention your speakers and drinking habits ;) ) and various other "masking" effects. I've used a soundcard in similar ways but for somewhat different reasons. Still, it might be a worthwhile experiment and I believe it has some validity.
 
Noise

I did some quick calculations and found that a 2 K ohm potentiometer would work pretty nicely. It has thermal noise of less than 1 microVolt over a 20KHz bandwidth at room temp. For comparison 106 dB below a 1.0 volt level is about 5 microvolts. So it would NOT comprimise the sound cards performance.

With 28 volts across 2 Kohm ( read 100W at 8 ohms, really loud) it only dissipates .3 Watts. So a 1.0 Watt pot is pretty standard stuff.

I think it's a great idea, and would let you match the input to the sound card for any listening level.


The beauty of this idea is that .wav files can be easily exchanged in email. So I could remotely audition all these neat amps you folks are building.
 
diyAudio Senior Member
Joined 2002
Re:They're not?

Hi,

Is the load on the amp the same whether you connect a speaker on its' own or whether you connect a speaker + a soundcards' input impedance?

Hey, if it doesn't matter to you it sure as hell doesn't matter to me.

Just let me tell you that 90% of the measurements out are made by people not knowing what nor how to measure so why should all of that make a difference, right?

Getting the scope of it? ;)
 
Re: Re:They're not?

fdegrove said:
Hi,

Is the load on the amp the same whether you connect a speaker on its' own or whether you connect a speaker + a soundcards' input impedance?

Hey, if it doesn't matter to you it sure as hell doesn't matter to me.

Just let me tell you that 90% of the measurements out are made by people not knowing what nor how to measure so why should all of that make a difference, right?

Getting the scope of it? ;)
No I'm not getting the scope of it. Your original suggestion of using a high powered L-Pad in parallel would have seriously changed the load the amplifier sees.

The suggestion to use a higher impedance divider is a logical one as it would allow the soundcard to effectively not be there (ignoring grounding issues for the moment, as all of these schemes share that problem).

The ONLY downside I see to a higher impedance divider is not with loading the amplifier as you somehow suggest, but if you make it too high, the input circuitry to the soundcard may not maintain its intended frequency response. But any slight frequency "bias" would be equally applied to both amplifiers being tested and shouldn't get in the way of evaluating the results.

The proposed network was 100k/10k which would present an 110k ohm load to the amplifier. That will change the load the amplifier sees from 8 ohms to 7.9994 ohms or a whopping 0.007%. I fail to see an issue here?

In fact, I'd suggest a network around 100 ohms/1k ohms to keep the source impedance to the sound card reasonably low while still keeping any loading on the amplifier negligible. That will change the load to 7.94 ohms or about 0.7%. If you do the math with the output impedance of even a tube amp, that 0.06 ohm change in load is going to amount to a gnat's eyebrow of a db--way under anything that's even remotely audible.

So, please explain why the loading is an issue and why an 8 ohm L-pad is better?
 
diyAudio Senior Member
Joined 2002
DEMANDING FRANKIE...

Hi,

The ONLY downside I see to a higher impedance divider is not with loading the amplifier as you somehow suggest, but if you make it too high, the input circuitry to the soundcard may not maintain its intended frequency response. But any slight frequency "bias" would be equally applied to both amplifiers being tested and shouldn't get in the way of evaluating the results.

Think so? You'd better make sure you have all parameters under control.

Cheers,;)
 
Divided

I would look at using a 2 kohm pot.

1) Thermal noise of 2 kohms is about 0.8 uV at room temp for a 20KHz bandwidth. 106 dB down from 1.0 Volt is about 5 uV, so the 0.8 uV resistor noise doesn't comprimise the sound card's noise floor.

2) 2 Kohm will only absorb about .3 watts at 28 Volts ( the voltage for 100W into 8 ohms) Not that anybody is acutally doing much listening at a sustained 100W level. So an easy to find half watt potentiometer will work just fine.


The beauty of creating wav files is that they can be shared by everyone. It would be great to hear how all these home brew amps people are building sound.

:idea:
 
It Is Worth A Try

Recently I found a site that had downloadable wav files of the outputs of different amplifiers and effects units.
I forgot to bookmark the page - sorry.
IOW this has been done before.
I had a quick listen on my $5 PC speakers but could not pick a difference however - needs better replay system or dump to a CDR and listen on a proper system.

Eric.
 
The sound card test.

I think that you may not be able to achieve what you want.
First of all you 'must' use a speaker load as it affects the response at the output of the amp - assuming that the amp has some real output impedance.
Secondly playing back the recording which has now gone through an ADC and then a DAC and its buffer amps will "add" to whatever went in. So no matter what you do you cannot get back the exact same sound. It might be better to play both amps with the same input signal and then get the null signal and record each null signal one on each channel. Then you can compare them on screen or listen to it.Note that some losses could be there due to the bandwidth limiting on the sound card.
I think you need to come up with a better proceedure.

Just occured to me - why don't we all get null signals from our amplifiers and record it via a sound card. Then post the null signals on the forum so that we can all see what the null sounds like for different systems.
1. Null with no load.
2.Null with a real speaker.
3.Null with an 8 ohm resistor load.
all at some sensible voltage level say 10 volts peak?
 
measurement is not impossible

Sorry for the repeats, My posts apparently got delayed, so when they didn't show up, I kept trying.

Actually, I realize that the sound card A/D isn't going to be perfect, but I suspect it is more sensitive than ANYONES ears or loudspeakers. Think I'm kidding. Try this. Go find something really loud, like a car horn. Now while your standing with your ears near the grill listening to that 110 dB horn, try to also to discern what somebody is saying in a normal speaking voice 10 feet away, (at a level something like 50 dB). You can't hear what they are saying. Their voice is 1/1000 th the strength of that car horn hurting your ears. That's a 60dB difference. That's more than your ears can discriminate. Now maybe your a golden ear, so move that guy talking back another 10 feet (40 dB) you got nothin. But this situation is no challenge for a sound card with a 100 dB signal to noise ratio, and dynamic range.

So yes any audible, and many sub-audible differences will be easily measurable.

The other thing that is really promising about this idea, is that the distortion of the sound cards A/D is about 1/50 that of the best loudspeakers ( .004 % vs .2%). Shoot, most these guys claim to hear differences through speakers with .5% distortion. So these differences would have to be on the order of .5% to be audible through speakers.

The only flaw I've thought of is that I don't have a way to synchronize the recording sampling with the source. I will have to repeat the measurements a few times to see how .... repeatable they are.

I do really like the idea of recording the 'nulls' as well. That gets straight to the heart of it, but doesn't let me reproduce my own A/B from the .wav file.
 
Status
Not open for further replies.