Analyzer?

Another issue that could be affecting the final result is whether the mic is calibrated and what position in the room it is placed. A flat room is a starting point for your setup not the end result for listening.
 
Seems if someone is doing anything beyond setting a system or room for a flat response it no longer can be refered to as being flat ?

I see in the diagram posted the pink noise has a slope ?
 
I see in the diagram posted the pink noise has a slope ?

This is correct.

If a perfect speaker (totally flat) were to be measured in an anechoic chamber with perfect test equipment, its response when fed pink noise would look like the pink noise graph. If it was tested with white noise the response would look like the graph of white noise.

If one then used an EQ to make the measured response flat using pink noise, the actual response would have a rising characteristic with frequency and the actual final response of the output of the speaker would no longer be flat.

Of course nothing is perfect.
 
Pink noise does not trace the sensitivity of the human ear, across the audio spectrum, as is reflected in the Fletcher-Munson curves. Hence flat by pink noise would not be flat per the response of the human ear. Equalizing to pink noise, without considering the response of the human ear could result in irritating sound

A bit more research and experimentation just might be in order here.

If you had a 10-band equalizer, you could adjust your system, using this test record, and your ears. Then you could run your analysis and visualize what your ears are looking for.

R-6542811-1423086416-8405.jpeg.jpg


Good luck,
Rich P
 
Last edited:
I need a picture of what a "flat response" would look like on a proper analyzer.

As already mentioned in the "edit" section of my post (#13), fed with pink noise the analyser should show columns of equal height.


If one uses pink noise as their source and then tries to make the display on their analyzer look flat, the actual response of their system will not be flat as previously mentioned.

Not really. As mentioned above, pink noise on a conventional frequency response plot would have that 3 db/octave drop. However, a proper "certain fraction of an octave per band" analyser should basically show the energy per band - which should be constant for pink noise, so that the analyser should should columns of equal height. Or in other words: The drop in the conventional frequency response basically reflects that every next octave covers twice as wide a frequency range as the previous octave.


Greetings from Munich!

Manfred / lini
 
Yes you are correct, is all about how the test equipment is set up.

Here is the same pink noise signal that I generated for my previous post, but with different display parameters.

upload_2018-7-12_21-42-11.png

Or change the display parameters and this is what is shown for the same signal.

upload_2018-7-12_21-56-23.png

Thanks for pointing that out. It is important for those that might be using apps on a smart phone to understand the display parameters and the impact that they have on the displayed results

The input to the measuring device (whatever that might be) stays the same.

And if the analyzer is set up as you say this is what you get. Again, I am generating the same pink and white noise signals.

View attachment 1233468


upload_2018-7-12_22-32-30.png
 
Last edited:
Pink in this case means filtered to maintain a flat response over the audio spectrum using random noise so that analyzers will display a flat curve.
 
1/3 octave is 16 hz to 16k hz ? With pink noise I should have a total of 3db slope across 1/3 octave or between frequencies ?
 
When I 'play around' with EQing I use my Yamaha EQ-1100 w/stock Mic with Yamaha C-85 & M4 (both recap by Avionic)

Allow the EQ to auto adjust using white noise... Play for an hour or two but feel like I'm missing something. It's not bad but the PRaT is gone.


164055068_yamaha-eq-1100-high-end-raumkorrektur-graphic-equalizer-.jpg
 
Absolutely flat causes fatigue in most listeners. We (as humans) don't find it the most pleasing -it sounds shrill. A slight roll-off is preferred. Look up the Fletcher-Munson curve. You're probably tuning it, by ear, in accordance.

But albums have dynamic range. Flat rooms + a properly mastered/mixed album and you get what the tech and/or artist wanted. If you get fatigue by listening to an album in a flat room there's an issue with the mix.
 
The pink/white noise and analyzer thing is confusing as hell. Start with the analyzer. There are two kinds. We've traditionally used 1/N octave analyzers, which are called constant percentage bandwidth (CPB) instruments or relative bandwidth instruments. 1/3 octave made sense because you didn't have to build too many filter circuits. 1/10 octave was also popular if the unit didn't have to work in real time, though with opamps even those are easy now. If you analyze pink noise with a CPB analyzer, it displays flat response.

When FFT analyzers were developed they were constant bandwidth devices. If you analyze pink noise with an FFT analyzer, it will show a falling response because the signal power at a constant bandwidth falls off at 3 dB per octave. So, with an FFT analyzer you need white noise to display a flat response. Of course every FFT analyzer is based on a processor, so many of them can convert from constant bandwidth to constant percentage bandwidth.

I hope I got that right, because it's still confusing. No doubt somebody will fix it if I didn't! Bottom line is either analyzer will tell you what your system response is, but you have to use the right signal for the type of analyzer you're using. Flat electronics/system will be correctly shown to be flat with either type of analyzer, if you use the right noise source. If you just listen to the noise sources, pink noise is softer and more pleasing than white noise.

IMO, 1/3 octave analysis is way better for looking at speakers and rooms because it simplifies things and ignores narrow resonances and dips.
 
Last edited:
The pink/white noise and analyzer thing is confusing as hell. Start with the analyzer. There are two kinds. We've traditionally used 1/N octave analyzers, which are called constant percentage bandwidth (CPB) instruments or relative bandwidth instruments. 1/3 octave made sense because you didn't have to build too many filter circuits. 1/10 octave was also popular if the unit didn't have to work in real time, though with opamps even those are easy now. If you analyze pink noise with a CPB analyzer, it displays flat response.

When FFT analyzers were developed they were constant bandwidth devices. If you analyze pink noise with an FFT analyzer, it will show a falling response because the signal power at a constant bandwidth falls off at 3 dB per octave. So, with an FFT analyzer you need white noise to display a flat response. Of course every FFT analyzer is based on a processor, so many of them can convert from constant bandwidth to constant percentage bandwidth.

I hope I got that right, because it's still confusing. No doubt somebody will fix it if I didn't! Bottom line is either analyzer will tell you what your system response is, but you have to use the right signal for the type of analyzer you're using. Flat electronics/system will be correctly shown to be flat with either type of analyzer, if you use the right noise source. If you just listen to the noise sources, pink noise is softer and more pleasing than white noise.

IMO, 1/3 octave analysis is way better for looking at speakers and rooms because it simplifies things and ignores narrow resonances and dips.

So hows do these analyses pertain to the various designs such as planars/ESLs compared to dynamic drivers? Are there inherent characteristics exclusive to each that influence how a flat response affects your ears?
 
Doesn't matter which you use, the answer is (conceptually) the same. The shape of the response is the shape of the response. I just find FFT plots harder to deal with because they show way too much. Other people seem to like them and find them revealing. I suppose you could say the averaging of the 1/3 octave analyzer and pink noise is closer to the way you hear sound, but now we're getting beyond my pay grade. The problem of measuring speakers in a room is quite difficult because it's a 4-dimensional problem. IMO, differences between ESLs and dynamic drivers are way beyond the tools any of us have available, other than ears! There's a lot of info at the B&K site, though you might have to make a free account to see it all- https://www.bksv.com/en
 
As a start, this is what I'm shooting for ?

upload_2018-7-12_16-40-17.png

On this graph it looks like 20 hz is at -30 db and 20k hz is close to -60. How is that considered a 3 db slope ? Looks more like 30 db.
 
That screen shot is of an FFT type analyzer, so flat with white noise. Pink noise is -3 dB per octave. An octave is a doubling or halving of frequency, so 20-20k is about 10 octaves (I think, somebody check that!) 30/10=3
 
20 40 -3db
40 80 -3db
80 160 -3db
160 320 -3db
320 640 -3db
640 1280 -3db
1280 2560 -3db
2560 5120 -3db
5120 10240 -3db
10240 20480 -3db

-3db on each doubling is -30 db total ?
 
I've messed with this for a bit, obviously need more money in equipment than my system for this to work.
 
Back
Top Bottom