• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Is Audio Science Review going about it all wrong? Or partly wrong? Or all right?

JohnYang1997

Major Contributor
Joined
Dec 28, 2018
Messages
2,559
Likes
2,315
Location
China
There are more issues with cheap electret mics that they have very big variation in production. So individual calibration is essential.

Another thing is that cardioid mics have vaired frequency response from distance. It's always better to use omni mics.
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
8,146
Likes
10,096
There is nt1 2014 version. But no measurements of those microphone except from rode themselves. Form rode it's much flatter than nt1a.
I can't say of course all maker supplied curves are accurate. The consensus has been they are fairy tales. I suspect they use considerable smoothing. I have been surprised however. I use a UMIK 1 as a reference. Obviously not an international standard quality reference. When I measured a speaker positioning all the mikes I had and could borrow at the same place (as much as possible) I was surprised to find if I used those published curves for calibration everything fell almost right on the UMIK response. If they aren't perfect, and even if smoothed apparently they are basically honest response curves.

You are correct that omnis are the best as cardioids and even more so figure 8 mikes have proximity effect for the lowest 200 hz at least. Here is a Umik 1, Avantone CK-1 omni, and CAD M179 omni. Other than the 4500 hz dip in the M179 these are pretty close. 1/6th octave smoothing. They aren't close at all without the calibration from the published graphs.

1563955998361.png


Here are two CAD M179's that differ in age by 6 months. The upper pair. The lower pair are Avantone CK-1s. QC in modern condenser microphones looks to be pretty good. I would think much of what difference you see is me not positioning them with exactitude. I've equally similar results from 3 KSM32's that span about 12 years in age.
1563956492398.png
 
Last edited:
Joined
Jun 1, 2019
Messages
46
Likes
16
A lot of these distortion graphs are unreliable since they mix the distortion from the microphone with that of the headphone. Measurements mikes are rated to produce 1 to 3% distortion over their full SPL. I also find that the resonances caused by the cavities in the measurement gear increases the levels and with it, the amount of distortion created by the microphone.
Yeah there quite of few reviews where the IEM, Headphone are tainted by whatever gear they use. Grado is a on ear open back so there mic distortion, room/headphone resonances and its own. Even tyll's distortion graphs had weird glitches on few he tested.

I highly doubt the THD+N of the ER4XR(1%) SR225e(2%) are even remotely audible with music and real ears beyond the few golden ears. But then again people still refuse believe the ER4 can reach 20 to 17,000Hz with only one BA. Despite there being tones of charts showing that the ER4 are that wide bandwidth.

I think Etymotic even admitted by one of there engineers that they never gotten any proof that the THD+N on there IEM's is audible with music, That a 0.2% DD is only good if your in areas where 85 to 120db SPL is needed. lol
 
Joined
Jul 8, 2019
Messages
28
Likes
16
I think ASR - as a forum - IS GREAT. The participants, almost all, seem well grounded, educated and forward their perceptions and observations with a minimum of venom and angst. I hear (read) very little drama in these proceedings. Such being the way rational folks should engage in discourse.

Being a trained engineer I prefer to establish my preferences and discernment on a foundation of rational thought. I don't believe in power cords, weird footers to support gear, magic tweaks, e.g. Sean Mook "dots", or any of the delusional claptrap of audiophilia. I do believe that equipment does sound different - some of it is rather startling in its ability to present recorded sound. I do believe that amps sound different as do digital players of all stripes. Not sure WHY these electronic devices do sound different - but I know they do.

I like vinyl playback - even though it is rather clear that such devices are of much lower fidelity - even though the net result is often times more emotionally engaging than a digital representation of the same music. Odd that. I like tubes - and I know that the coupling transformers ring like a bell, slow the signal, round it off and generally add a heavy signature to the recording - but egad!! Those colorations can be very engaging from a sonic standpoint. Though my main system is solid state. The tubes - old ST-70s that I rebuilt - are in my secondary system.

In short - I think ASR is the best forum on which I have ever participated. Keep it up fellows - and Amir.
 

Snarfie

Senior Member
Forum Donor
Joined
Apr 30, 2018
Messages
471
Likes
222
Location
Netherlands
I can't say of course all maker supplied curves are accurate. The consensus has been they are fairy tales. I suspect they use considerable smoothing. I have been surprised however. I use a UMIK 1 as a reference. Obviously not an international standard quality reference. When I measured a speaker positioning all the mikes I had and could borrow at the same place (as much as possible) I was surprised to find if I used those published curves for calibration everything fell almost right on the UMIK response. If they aren't perfect, and even if smoothed apparently they are basically honest response curves.

You are correct that omnis are the best as cardioids and even more so figure 8 mikes have proximity effect for the lowest 200 hz at least. Here is a Umik 1, Avantone CK-1 omni, and CAD M179 omni. Other than the 4500 hz dip in the M179 these are pretty close. 1/6th octave smoothing. They aren't close at all without the calibration from the published graphs.

View attachment 29953

Here are two CAD M179's that differ in age by 6 months. The upper pair. The lower pair are Avantone CK-1s. QC in modern condenser microphones looks to be pretty good. I would think much of what difference you see is me not positioning them with exactitude. I've equally similar results from 3 KSM32's that span about 12 years in age.
View attachment 29956
My first choice was to buy the UMIK 1 but i was advised not to do so because if i would try to do some measurments that took several meters i had to extend the USB cable which could create problems causing a degrading frequency curve at certain (thought higher) frequencies. What is your opinion/experience on this. Using now a Yamaha MG102C which has a phantom source to do measurments with the Superlux making use of XLR cables that i can extend to several meter.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
2,766
Location
Zg, Cro
My first choice was to buy the UMIK 1 but i was advised not to do so because if i would try to do some measurments that took several meters i had to extend the USB cable which could create problems causing a degrading frequency curve at certain (thought higher) frequencies.
I'm pretty certain that is not what would happen. When you extend USB cable you can expect max bitrate to drop. For example, with DAC connected to USB port to your PC with cable up to 2m you can expect to reach 384kbps without problems but if you connect it with 5m cable 96kbps may be maximum bitrate you would be able to achieve. Now UMIK-1 probably works works with fixed bitrate and assuming that bitrate is 44.1 or 48kbps what you will experience once you reach the maximum cable length is it would simply stop working, but as long as it works you will have full frequency range with your measurements.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
1,053
Likes
2,501
Location
Hampshire
When you extend USB cable you can expect max bitrate to drop. For example, with DAC connected to USB port to your PC with cable up to 2m you can expect to reach 384kbps without problems but if you connect it with 5m cable 96kbps may be maximum bitrate you would be able to achieve.
That is not how USB works. If the cable meets spec, you get the full 480 Mbps rate regardless of length (although there is a max length of ~5 m). If the cable doesn't meet spec, nothing is guaranteed. A poor cable might still pass the high-speed handshake, only to be unreliable during data transmission. For audio devices, this is likely to manifest as crackling, noise bursts, or dropouts. There is no mechanism for falling back to a lower speed.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
2,766
Location
Zg, Cro
That is not how USB works. If the cable meets spec, you get the full 480 Mbps rate regardless of length (although there is a max length of ~5 m). If the cable doesn't meet spec, nothing is guaranteed. A poor cable might still pass the high-speed handshake, only to be unreliable during data transmission. For audio devices, this is likely to manifest as crackling, noise bursts, or dropouts. There is no mechanism for falling back to a lower speed.
Well, thai is how it worked when I tested USB connection with Topping D10 and my notebook with cables of various lengths. I also tested with active USB extender which uses 10m Cat 7 cable and also was not able to get more than 96kbps (DAC bitrate, not USB speed!) from Topping, as was the case with 5m USB passive cable.

Anyway, coming back to UMIK-1 question, as it supports only one bitrate it will simply stop working once max cable length is exceeded but as long as it works it will not cut HF when taking measurements.
 

mansr

Major Contributor
Joined
Oct 5, 2018
Messages
1,053
Likes
2,501
Location
Hampshire
Well, thai is how it worked when I tested USB connection with Topping D10 and my notebook with cables of various lengths. I also tested with active USB extender which uses 10m Cat 7 cable and also was not able to get more than 96kbps (DAC bitrate, not USB speed!) from Topping, as was the case with 5m USB passive cable.
In that case the initial high-speed handshake failed, and the connection stayed at full-speed (12 Mbps). In a sense, you were "lucky" that the cable was bad enough that the handshake failed. A slightly better/shorter cable could have worked more or less intermittently at high-speed.

Also, please say sample rate when that is what you mean.

Anyway, coming back to UMIK-1 question, as it supports only one bitrate it will simply stop working once max cable length is exceeded but as long as it works it will not cut HF when taking measurements.
Looking at the UMIK-1 spec sheet, it appears to be a full-speed only device. As such, it is (probably) more tolerant of poor cable quality. However, the same kind of thing can happen here. Even when the device enumeration succeeds, data transfer can be unreliable if the cable is too poor. I have seen this happen in my own testing using deliberately awful cables.

You are of course correct in that if the USB connection does work properly, the length of the cable has no influence on the sound whatsoever.
 

Krunok

Major Contributor
Joined
Mar 25, 2018
Messages
4,600
Likes
2,766
Location
Zg, Cro
In that case the initial high-speed handshake failed, and the connection stayed at full-speed (12 Mbps).
Exactly.

You are of course correct in that if the USB connection does work properly, the length of the cable has no influence on the sound whatsoever.
His main concern was that mic won't be able to measure HF in a linear fashion with a very long cable but that of course cannot happen in any scenario.
 

Snarfie

Senior Member
Forum Donor
Joined
Apr 30, 2018
Messages
471
Likes
222
Location
Netherlands
Yes, but there is an easy way vs just eyeballing if that is what you are doing.

https://automeris.io/WebPlotDigitizer/

Here is a good tutorial.

Web plot digitizer lets you drop images of graphs into it and spit out among other things a .csv file which REW will use. Play around with it some, and you'll get the hang of it. Export the result as .csv when done with the browser based version. I've done this with many microphones, and surprisingly those published graphs appear to be accurate. You can let it automatically place dots on the graph lines, parse out the color and it will do it fairly automatically. The biggie is setting the axis precisely and in logarithmic form.

TIP: if you can only find a black and white version, some photo editing software to air brush the curving graph a different color is enough this software will select the sample points correctly.

Once you have used it a bit, won't take long to grab a FR graph and spit out a cal file for your microphone.
Did find a site with a easy to use instructions how to create a frequency data set/calibration file with Webplotdigittizer. https://medium.com/@jaakkopasanen/make-your-headphones-sound-supreme-1cbd567832a9
 

Blumlein 88

Major Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
8,146
Likes
10,096
My first choice was to buy the UMIK 1 but i was advised not to do so because if i would try to do some measurments that took several meters i had to extend the USB cable which could create problems causing a degrading frequency curve at certain (thought higher) frequencies. What is your opinion/experience on this. Using now a Yamaha MG102C which has a phantom source to do measurments with the Superlux making use of XLR cables that i can extend to several meter.
Use those active USB cable extenders. I've used them as long as 78 feet. They work fine even at USB 2 speed. I've one at 33 feet I use daily for audio. They have some that are USB 3 which I have not used, but apparently they work as well.

https://www.amazon.com/Tripp-Lite-Hi-Speed-Extension-U026-016/dp/B008VOPEQW/ref=sr_1_10?keywords=active+usb+cable+extender&qid=1564422506&s=gateway&sr=8-10&th=1

https://www.amazon.com/AmazonBasics...able+extender&qid=1564422428&s=gateway&sr=8-3

Of course nothing wrong with your Superlux. If you don't have an interface and phantom power the Umik is a simpler solution. Certainly have a recording interface is more versatile.
 
Last edited:
Joined
Jul 2, 2019
Messages
45
Likes
68
Location
Boston area, Cambridge, MA
No-one uses a blindfold. And that's a very poor way to control an experiment.

If two things sound different, they will still sound different if you don't peek.
I'd run the study this way:
- Experimenter A sets the two equipment chains up, levels them, and obscures them with screens.
- A pseudo-random digital gadget runs a switch between the audio gear being tested.
- Experimenter B briefs each subject with a quick script.
- Subjects enter the room one at a time and sit in a designated chair.
- Subjects record the results of a series of n random listening sequences on a sheet of paper.
- Experimenter B asks for impressions not captured by the in-room questionaire, post listening.
- some means is provided to detect Subjects that peek behind the curtain or engage in other disqualifying behavior.

Ok, so you still have to pick listening material and how that is handled as a cross-variable.
Impression data is correlated post-study with the known pseudo-random sequence for a particular subject.

This seems to meet criteria for a double blind study. I'd be really careful not to leak information about what equipment is studied,
what the experimenter's opinions are, etc. I'd also design the approach to the study (the physical approach) to be as neutral as
possible, (i.e. no 'priming' audio equipment in sight, logos, etc.).

This is all pretty standard behavioral study stuff, avoiding "Clever Hans" effects, "priming", information influences subject to subject and
ensuring it really is double blind.
 

SIY

Technical Expert
Technical Expert
Joined
Apr 6, 2018
Messages
3,880
Likes
7,328
Location
Phoenix, AZ
Last edited:
Joined
Jul 2, 2019
Messages
45
Likes
68
Location
Boston area, Cambridge, MA
@BostonJack Specific experimental protocols depend on the specific question being asked; trying to use one protocol across the board is something I call "the Procrustean Fallacy." I've pimped this article before and I'll do it again.
Ok, I've read the article before, thanks. Skimmed it now, will read it carefully again when I have time. He does call out the need for double-blind. I guess my main point was that 'clever Hans' can creep in in very subtle ways, as can peer-to-peer influence, and subtle biases based on experimenter behavior. Having a magician present is a great idea.

Haha! no control group here. Maybe the control group is a set of subjects where the switching is mocked and the same system plays each time.

I guess that my little improvised protocol addresses the question: "can subjects reliably detect the difference between device X and device Y".
My underlying thesis is that mainly they can't very reliably and we are engaged in a behavioral test.

People will generally prefer things that are in line with their innate assumptions and previous experience.
 

SIY

Technical Expert
Technical Expert
Joined
Apr 6, 2018
Messages
3,880
Likes
7,328
Location
Phoenix, AZ
Ok, I've read the article before, thanks. Skimmed it now, will read it carefully again when I have time. He does call out the need for double-blind. I guess my main point was that 'clever Hans' can creep in in very subtle ways, as can peer-to-peer influence, and subtle biases based on experimenter behavior. Having a magician present is a great idea.
Indeed, and I used Clever Hans in that article as a way of potentially explaining the classic "wife in the kitchen could hear the difference" anecdote- and how I tested that idea experimentally!

Everyone involved in sensory testing should have at least a basic grasp of conjuring, and I learned a ton about experimental controls from some excellent magicians.
 
Joined
Jul 2, 2019
Messages
45
Likes
68
Location
Boston area, Cambridge, MA
Indeed, and I used Clever Hans in that article as a way of potentially explaining the classic "wife in the kitchen could hear the difference" anecdote- and how I tested that idea experimentally!

Everyone involved in sensory testing should have at least a basic grasp of conjuring, and I learned a ton about experimental controls from some excellent magicians.
Doh! I feel a little stupid. I didn't connect the dots that you are the paper's author. Great paper! My exposure to experimental design is a BS in Biology taken late in life plus a lot of exposure to statisticians and life sciences types.

I subsequently learned that discussing studies comparing human toddlers, german shepherds and chimpanzees for their varying reliance on social cues isn't great date talk. %^)
 

Zerimas

Active Member
Joined
Jan 9, 2019
Messages
146
Likes
96
Doh! I feel a little stupid. I didn't connect the dots that you are the paper's author. Great paper! My exposure to experimental design is a BS in Biology taken late in life plus a lot of exposure to statisticians and life sciences types.

I subsequently learned that discussing studies comparing human toddlers, german shepherds and chimpanzees for their varying reliance on social cues isn't great date talk. %^)

I don't have the same educational background as you, but I can definitely understand where you're coming from with respect to "date talk". I wish I had a list of stupid things I've said I thought would be interesting to the other party, but ended up making me look like a crazy asshole. Although, if modern psychology has any validity there's a good chance that I may just be a crazy asshole.

Geez, those psychologist always giving me tests with high degrees of statistical validity and then telling me how I feel. It's not like "measurements reveal personality". ;)

Whoops, sorry. I slipped into "audiophile" mode.
 

Zerimas

Active Member
Joined
Jan 9, 2019
Messages
146
Likes
96
Wow. This has become a dating advice thread.
Sure, why not? I am also pretty sure that talking about audiophile hobbies on a date is a bad idea. Firstly because no one is interested in audiophilia to begin with. Secondly, even if your date were an audiophile, I would say that the chances of disagreement on something are quite high.
 
Top Bottom