• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Understanding the state of the DSP market

Thread seems to have gone way off topic. It seems like "it's hard to take measurements" is actually "it can be hard to know how best to measure if you are measuring in and around the listening position in a room so as to enable your chosen dsp software to function effectively" (which is also commonly the reason why each piece of software has lengthy threads dedicated to it full of posts about the one true measurement pattern/approach for that software)
Take it from the simplest one: mic orientation.
0° or 90° ?
Two channel people will tell you 0° , HT people will tell you 90° .

What if you use it fr both?
 

Presumably you should use whatever is recommended by the developer of the software being used...

Not rocket science there surely?

(I wonder if it might be useful to split DIY speaker development from room correction here. Seems to be a lot of talking at cross purposes.)
 
Take it from the simplest one: mic orientation.
0° or 90° ?
Two channel people will tell you 0° , HT people will tell you 90° .

What if you use it fr both?
I don't understand the question tbh
 
Keith
 
Let's take one example fairly important: distortion.
Find a really good mic (essential as we will see) take a measurement with a sweep at one meter, another at 3 meters and then try FSAF (that's were the good mic comes) with pink/white/whatever noise and compare.

You'll get 3 different results. Which one would you trust?
I have tried to get fsaf measurements with Rew repeatable and reproducable. To begin with loopback cables. Just to understand the system and its limits. No success, so i stopped .
 
I have tried to get fsaf measurements with Rew repeatable and reproducable. To begin with loopback cables. Just to understand the system and its limits. No success, so i stopped .
Here's a quick and dirty loopback with pink noise:

FSAF.PNG

dis.PNG
 
Now that's interesting.
The above loop has an HP BU8 filter at 31Hz (visible at set up page)
I took another one without it and here's the interesting part:

GD.PNG

Red is with filter, Green without it.
Group delay is up by a fair bit.

(What's not to love at REW? )
 
It wouldn't surprise me at all to hear that someone has spent 5k on room correction tools and 1k (if that) on treating their room. Will there be day when that statement can be made in reverse? It's almost like people don't understand how soundwaves behave.
 
Take it from the simplest one: mic orientation.
0° or 90° ?
Two channel people will tell you 0° , HT people will tell you 90° .

What if you use it fr both?

45°

:cool:
 
I'd like to have my say about the measurement since I work in a metrology lab and am quite educated in measurement science.

Here we are talking about measuring the variation of acoustic pressure over time, done with a chain consisting of a pressure transducer and an AD converter.

Any measurement has an associated uncertainty that should define within what margin the obtained value deviates from the true value (which by definition is not directly measurable).

Measurement uncertainty is made up of several contributions and often there are standards and guidelines that help define them, but unfortunately I am not aware of those available for the physical quantity in question.

Making a fairly reasonable list of contributions, one could list the following (I use the quotation to avoid making the post too long)

1) MEASUREMENT ACCURACY OF INSTRUMENTS:

The transducer (microphone) and the ADC (audio interface) can provide a certain measurement accuracy, typically obtainable from the specifications, or better, from a calibration certificate traceable to the International System of Units (SI). In calibration these instruments are compared with references that are sufficiently more precise, but still have their uncertainty, and the measurement accuracy that they can provide can be roughly considered as the deviation found (error) plus the uncertainty of the reference.
It is often used to compensate the measurement with the inverse of the deviation (the use of the famous microphone calibration file) in order to improve its accuracy. This is a widespread practice in metrology, however, two things should still be considered:
a. The uncertainty of the reference from which it was calculated in deviation: if I detect an error of -5 and the associated uncertainty is +/-15, I cannot be sure that my compensation is bringing me closer to the true value. Conversely, if I have an error of -5 at an uncertainty of +/-1, then I am sure that by compensating with +5 I am improving my accuracy.
b. The accuracy defined in the specification by the instrument manufacturer: even where I have an uncertainty favorable to the application of the compensation, I should still take into account the accuracy defined by the manufacturer, because he has defined (or should) this parameter as a function of all the variables typically at play in the use of the instrument (e.g., temperature, humidity, etc.). So even if I have calibrated my instrument with an absurdly precise reference, very close to the Si standard, I should remember that this result is valid for those specific calibration conditions and could change as a function of them.
In addition to the above, other factors also affect the measurement accuracy of instruments. An analog instrument such as a microphone will have, for example, stability, noise, linearity, hysteresis, offset. While a digital instrument such as an ADC could have phase noise, resolution, rounding, truncation, interpolation. Even the software used contributes to the measurement uncertainty if it performs mathematical operations.

2) INFLUENCE OF PHYSICAL VARIABLES:

As can be deduced from the previous point, the variables involved in a measurement are multiple and it is often not possible to analytically quantify their contribution to the uncertainty. In the measurement of an audio system in a room we have variables such as temperature, humidity and noise that we cannot determine exactly how much they are influencing the physical quantity that we want to measure, nor the instruments that we are using. In these cases, to estimate the contribution of uncertainty we proceed statistically, repeating N times the measurements under the same conditions, or if desired, under varying conditions, so as to be able to then appropriately quantify the range within which my real value can be found.

3) INFLUENCE OF THE MEASUREMENT SETUP

The measurement we have to do here is really unfortunate. If I have to measure an electrical signal, I plug in the cable and measure. Few influences related to the setup. If I have to measure a signal that propagates by means of the variation of air pressure over time in a closed space... the influences related to the setup are many.
a. The room reflects the signal: the good thing is that the room effect is part of what I want to measure. The bad thing is that the room effect depends on the measurement point. So the uncertainty due to the reflections is very important, especially when our listening point is not extremely fixed at one point.
b. The orientation of the microphone: the microphone, although omnidirectional, introduces an alteration depending on the angle of reception of the signal in space. The uncertainty that this entails on the direct sound will probably be small (it is quite easy to position it at 0° or 90° with respect to the sound source), especially if I use a calibration file, but the reflected sound waves will have a different and unpredictable incidence. Therefore there will be a component of uncertainty linked to the signal depending on the incidence with which it arrived at the capsule.

As can be seen, the act of measurement is something that implies uncertainty by nature. To quantify it, it would be necessary to properly enter into the merits of each single point to determine its contribution and relevance. This is what is typically done in metrology (the so-called uncertainty budget). But for obvious reasons here we cannot/are unable to enter into the merits of everything. Partly because we do not have the information, partly because we do not have the tools and partly because it could be superfluous for our purposes.

What should emerge when talking about measurement, however, is that it is not possible to make a speech that is too general or too specific about what/how to do it, nor is it possible to ignore the uncertainty component inevitably associated with the result, if it is not sufficiently defined what I need to measure and with what precision I need to do it.

In this hobby in practice we rely on the fact that in most cases what we measure with the good practice represents quite significantly what we perceive and the associated uncertainties are probably negligible for perceptual purposes. And even where they are not, we still end up tuning our system with a target curve, or other parameters, that we seem to like by ears.
So how relevant is accuracy here?
Will the measurement accuracy determine if we will skip or not by listening to a song?

From my point of view, thinking about measurement beyond the limits of rough good practice, in this hobby, is something of relative utility. It's like weighing ingredients when you cook... accuracy itself will not determine predominantly if you like that food.
 
Last edited:
So good data:
Repeatability: if you mean reproducable then yes.
Dealing with variation as result measurement position variation: yes
Understanding multiple arrivals (always the case) : yes
Calibrated meas-system: yes
Understanding tolerances in results: yes

Where comes the so called pro into play in comparison with a diy person addressing his audio issues in home?

Good point...good data is good data, no matter who it comes from. Our carefulness and techniques in collecting the data is all that matters.

I think my frustration with home audio DSP is rooted against the idea of auto-room correction. Maybe this has become a bias I've developed attending/studying prosound speaker and room tuning techniques. Auto-room correction is pretty much considered a joke in these classes and circles.

I do admire all you guys who are doing your own in-room measuring, not relying on auto stuff.
As hard as it is for me to make valid quasi-anechoic speaker tuning measurements , for so far, valid in-room measurements have been a full order of magnitude more difficult.
 
Good point...good data is good data, no matter who it comes from. Our carefulness and techniques in collecting the data is all that matters.

I think my frustration with home audio DSP is rooted against the idea of auto-room correction. Maybe this has become a bias I've developed attending/studying prosound speaker and room tuning techniques. Auto-room correction is pretty much considered a joke in these classes and circles.

I do admire all you guys who are doing your own in-room measuring, not relying on auto stuff.
As hard as it is for me to make valid quasi-anechoic speaker tuning measurements , for so far, valid in-room measurements have been a full order of magnitude more difficult.
Thinking/reflecting about what you are doing, and what is going on is for me important. My experience with Acourate is that it results in high quality crossover and room curve aka driver response correction.
But also that small changes also become audible, changes that one would not even consider with passive filtering. So you will continue tinkering but at a much higher quality level ;-)
 
I'd like to have my say about the measurement since I work in a metrology lab and am quite educated in measurement science.

Here we are talking about measuring the variation of acoustic pressure over time, done with a chain consisting of a pressure transducer and an AD converter.

Any measurement has an associated uncertainty that should define within what margin the obtained value deviates from the true value (which by definition is not directly measurable).

Measurement uncertainty is made up of several contributions and often there are standards and guidelines that help define them, but unfortunately I am not aware of those available for the physical quantity in question.

Making a fairly reasonable list of contributions, one could list the following (I use the quotation to avoid making the post too long)



As can be seen, the act of measurement is something that implies uncertainty by nature. To quantify it, it would be necessary to properly enter into the merits of each single point to determine its contribution and relevance. This is what is typically done in metrology (the so-called uncertainty budget). But for obvious reasons here we cannot/are unable to enter into the merits of everything. Partly because we do not have the information, partly because we do not have the tools and partly because it could be superfluous for our purposes.

What should emerge when talking about measurement, however, is that it is not possible to make a speech that is too general or too specific about what/how to do it, nor is it possible to ignore the uncertainty component inevitably associated with the result, if it is not sufficiently defined what I need to measure and with what precision I need to do it.

In this hobby in practice we rely on the fact that in most cases what we measure with the good practice represents quite significantly what we perceive and the associated uncertainties are probably negligible for perceptual purposes. And even where they are not, we still end up tuning our system with a target curve, or other parameters, that we seem to like by ears.
So how relevant is accuracy here?
Will the measurement accuracy determine if we will skip or not by listening to a song?

From my point of view, thinking about measurement beyond the limits of rough good practice, in this hobby, is something of relative utility. It's like weighing ingredients when you cook... accuracy itself will not determine predominantly if you like that food.
How do you feel about in-room MMM FR measurements? For in-room FR I get results that are consistent and repeatable and they seem to correlate well with what I hear.

For me the in-room measurements that I really struggle to get any kind of confidence in are "timing" measurements taken anywhere but near field.
 
It's rhetorical, about the length of the said threads.
Still a bit lost but no matter :)

In the end

Taking measurements: not hard
Interpreting measurements: as hard as you want to make it (as recent posts show) and varies depending on what you measured
Know how/what to measure: see above

The impression given in many posts is that people get wildly varying results from one measurement to the next, sounds a bit odd to me if so.
 
How do you feel about in-room MMM FR measurements? For in-room FR I get results that are consistent and repeatable and they seem to correlate well with what I hear.

For me the in-room measurements that I really struggle to get any kind of confidence in are "timing" measurements taken anywhere but near field.

I use MMM or Pink Noise and RTA for FR response.

For 2 Ch "timing" - read delay settings - i go "by ear" - mono signal needs to be dead center and focused. For final subs alignment Trinnov gets it right usually, but to confirm I do 80Hz pulse and listen again for any phase cancellation at crossover point.

Just my loose thoughts reading this [and other threads]
1) While DRC is powerful tool, it should be what ti is - Digital ROOM Correction, sometimes I have the impression, that whatever the issue is, you can EQ or DSP your way out of it. Nope. ****** speakers will sound like ****** EQ'd speakers, if they have some inherent issues. You need to have - good high quality speakers, placed properly in at least acoustically acceptable room in order to get good sound. Then employ DSP properly and you are good to go.
2) Obsession about measurement accuracy - there will be much more difference between 2 sessions in your perception of music - be it mood, fatigue or THC/alcohol levels. Kind of get it 80% right by following basic good practice and you are 80% where you need to be. Also see 1)
3) I can see people obsessing about having 100% adherence to target curves etc - there are more important parameters that are audible than 1-2 db ripple here or there. Especially in bass region - prioritize Group delay/timing/decay times over last dip or peak.

Don't be afraid to do the things you like - you like more bass? go for it, you prefer sparkle or accentuated vocal region? Harman curve be damned. Just don't forget to have fun. Actually more powerful tools I have at my disposal re DRC/DSP more I am preferring to do as little corrections as possible. Better to under correct than over correct.
 
Back
Top Bottom