• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. 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!

Cloning a speaker sound

gabo4au

Member
Joined
Apr 18, 2022
Messages
57
Likes
52
A bit of explanation of what I'm trying to do.

I have a studio control room (25'x17'x12') that has been treated fairly well. Many hours of work with REW, sound panels, bass traps, etc.

We also have a pretty good set of large monitors built into the wall, with dual subs, all tuned with DSP/DIRAC to the room. So the listening environment is very good.

Now I'm trying to develop a way to make the speakers/room sound like many other speakers via a VST plugin in a DAW. The reason for this is to be able to quickly audition what a mix would sound like on other speakers, or at least close. So I could have a classic "Yamaha NS10" sound, or an Auratone sound, or some home speaker sound, etc. It's an important thing in mixing to understand what your mix sounds like on lower quality devices.

So it's a bit of the reverse of speaker/room correction. So make my really good sounding speakers sound like a really crappy set of speakers!

It's also important to do this inside the DAW, so a mixing engineer could easily switch it around inside his mix workflow.

Of course, my first thought on how to do this was to simply put an EQ on the output bus, and manually build a curve that mimics the curve of whatever speaker I wanted to clone and save them as presets. Which still might be a good way to do it. Vituixcad has a really nice "SPL Tracer," where I can download frequency response graphs of speakers and easily build a txt file of the frequency response.

Of course the challenge is to get a reasonably accurate curve into an EQ, turns out it's pretty hard to get an accurate curve unless you have many EQ points and even then very difficult to get it accurate.

So I turned to another idea. Which was to see if I could use a Convolution Reverb in the DAW to implement an EQ curve. So I took REW and just pulled up any old room curve I had in there, went to the EQ filter section and set it to the Subwoofer EQ, built a filter and exported it as an impulse response. I then imported the IR into my reverb. The purpose this was to just do something dramatic to see if it worked, and it did. Putting the reverb on master bus resulted in a low pass filter being applied, so all the HF was gone. It's a bit fiddly as I have to make sure to set a few things in the reverb that don't apply, like pre-delay, etc. But at least I could successfully put an EQ filter in via an IR. I can probably find a better convolution engine VST plugin if I can get this worked out.

Next step was to find a speaker curve to see if I could get an accurate representation. There are tons of curves around for the Yamaha NS10, so I clipped one in snipping tool, moved it to the SPL trace of Vituixcad, successfully traced it and exported it to a txt file. I noticed the file had phase in it, which seems to not mean anything as there is no phase info in the frequency graph I traced, so I edited that out to have a file with just frequency and SPL.

I can open that file in REW, and it puts up a correct looking curve in the SPL&Phase chart, but nothing in the other charts. I tried exporting that as an impulse response but it complains that there isn't any data to export.

How can I create an accurate impulse response wav file from my txt frequency graph? Or maybe there is still a better way to get a frequency response txt file implemented into a DAW as an EQ.

Any thoughts would be greatly appreciated.
 

alex-z

Addicted to Fun and Learning
Joined
Feb 19, 2021
Messages
910
Likes
1,685
Location
Canada
You can change the frequency response all you want, but the radiation pattern still won't match, so it won't actually sound like the speakers you are trying to emulate.

Your best bet for emulating bad speakers is using the EQ to create a flat in-room response, which will sound bright and lacking bass impact. Or simply listen in another environment altogether, like a car.
 
Last edited:

antcollinet

Master Contributor
Joined
Sep 4, 2021
Messages
7,414
Likes
12,301
Location
UK/Cheshire
You can change the frequency response all you want, but the radiation pattern still won't match, so it won't actually sound like the speakers you are trying to emulate.

Your best bet for emulating worse speakers is simply using the EQ to create a flat in-room response, which will sound bright and lacking bass impact. Or simply listen in another environment altogether, like a car.
You also won't be able to simulate the distortions/cabinet resonances etc. At least not with any degree of accuracy.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
All the issues of directivity, resonances, distortion etc are true. However, don't write it off. I have done this a few times and you might be surprised how close you can get things to sound. It is one of those experiences that accentuate the idea that hifi is 85% frequency response.

Now I did it with a room correction unit. I measured right around one seating position. Doing separate correction to a single target curve for left and right channels. If you limit where you listen after correction it can get pretty close. I'll give a most extreme example which would have everyone claiming no way would it work.

I used a Hales Signature Two speaker on one channel. Sealed box MTM speaker of good quality. The other was a Soundlab ESL. A dipole panel speaker. Had both pair of speakers side by side in the room. Was when I first owned the Soundlabs. I had a couple friends over who were very familiar with my system. Played it this way without mentioning anything, and of course gave the visitor the prime listening position as a courtesy of course. Both times they listened and never noticed anything amiss. I didn't tell them anything. Thought they were listening to the Soundlabs. When I revealed it both found it hard to believe. If you moved around to other non-central listening positions there were tell-tale signs you could notice though even that was less obvious than you would imagine. Makes me think that first arrival FR is the most important part of the sound we hear. It amazed me how good the central imaging was this way.

Now turn off the room correction and it sounded about as messed up as you would imagine even when I left it so the relative loudness was matched. I used Tact Audio Room EQ for this so it was actually doing more than simple EQ it was also correcting for phase to some extent. I also did this with a few other speakers with similar results. Sometimes you had to adjust the target curve for the limited performance envelope of the lesser speaker.

This isn't quite what the OP is doing. I have done that too to a lesser extend. EQ'd one set of speakers based upon measurements of another speaker. That also works, but I've not tested that side by side so I am not sure it works as well.
 

FeddyLost

Addicted to Fun and Learning
Joined
May 24, 2020
Messages
752
Likes
542
I think the only way of proper simulating "some other speakers" is to rent these speakers and measure them at intended SPL with a good calibrated mic, then generate a convolution filter and apply it only at intended SPL.
Even this way will not show you some minor distortions like IMD or power compression and different directivity, but at least it will work better that "generic bookshelf EQ" made by hand according to FR graphs in audio magazines.
 
OP
G

gabo4au

Member
Joined
Apr 18, 2022
Messages
57
Likes
52
Well, first off. I know it isn't going to sound "exactly" like the other speaker. But if it gets me 85% there it's valuable. So there's that. For many years, mixing engineers have used multiple monitors in the studio. It's almost standard to have "big monitors" and "nearfield" monitors. But when getting close to the final mix, they also almost always dump mix's they can take to the car, the kitchen, the portable box, a friends house, etc. This allows the engineer to learn what things work and don't work on a variety of "consumer" type devices. All these consumer devices are in different environments and different setups. So the goal is to present the mix engineer with a sample of a number of these in the studio. So there is really no 100% accurate, what does that even mean in this situation as they are all different. But if I can simulate a bunch of them to 85%, that will be useful.

And to all those who say it's not possible, apparently you haven't been involved in the music industry! In the world of "guitar amps" using IR's to "model" other amps, speaker cabinets, and mics is a very common thing. They have gotten so realistic that a modeled guitar speaker cabinet is very difficult to tell from the real thing. Go take a look at a top notch modelling amp like a Revv D20, which has a 6 position switch that lets you select from one of 6 models you have loaded. You can download IRs for many amp/speaker cabinet models and load them into the amp so they become one of the 6 selectable models. The base amp is specifically designed to be neutral, so it doesn't impart it's own sound, just like most big studio monitors are designed.

So this thought it not original and has been done in other environments to great success. Granted, what I'm doing is sort of a simple approach, but the concept is the same. That being, produce a bunch of IRs that are models of the speaker sound I want and load those into my environment. If I can get the basic concept of how to take an frequency response curve and create and accurate IR from that curve, load it into a convolver, and make that work in my VST, then I can eventually go to the next level.

The next level would be to obtain speakers, and I could get access to many, and use tools to measure them more accurately including phase, cabinet resonances, and possibly even the rooms they are in to create IRs of those environments. The journey is just beginning.

By the way, in poking around I actually found a thread here on this forum for a product called "Hang Loose Convolver" which looks to have a VST version. This looks like a convolver VST that is made for room correction. This looks like a better VST solution than using a convolution reverb. So now I'm back to my original base question.

How do I take a frequency response graph and turn that into an IR? With SPL tracer, I can easily turn it into a txt file with 48 points per octave. But I've yet to get a good process for turning that into an IR.
 

dasdoing

Major Contributor
Joined
May 20, 2020
Messages
4,209
Likes
2,675
Location
Salvador-Bahia-Brasil
I would just set a high and lowpass at points where you feel the midrange of your sound is isolated. that would be an Auratone like setup.
in order to check translation those headphone room simulation programs are better imo. waves has a coulpe of them
 
OP
G

gabo4au

Member
Joined
Apr 18, 2022
Messages
57
Likes
52
I would just set a high and lowpass at points where you feel the midrange of your sound is isolated. that would be an Auratone like setup.
in order to check translation those headphone room simulation programs are better imo. waves has a coulpe of them

Nah, setting simple high and lowpass EQ isn't good enough.

Thanks for the pointer to the waves virtual room plugin! I didn't realize they made that, that's a start to what I'm looking for albeit it only for headphones. I would also like to model my own environments, so still want to pursue this route. But I'll definitely get the waves plugin to add to the thousands of plugins I already have :)

Update: In checking around, I did indeed find a number of "mix room environment plugins" available. For some reason I had not looked into that, but it seems it's not such a far fetched idea after all. However, every one I can find either simulates a specific listening environment, like a big Nashville studio, or has a few "built in" environments you can choose from including small popular monitors. However, I can't find any of them that allow you to load in IRs and sort of build your own. I'm all about flexibility and being able to extend the environment, so I haven't really found anything yet that kills my project.
 
Last edited:
OP
G

gabo4au

Member
Joined
Apr 18, 2022
Messages
57
Likes
52
Ah, so I figured it out in case anyone else wants to try this.

In REW, If I import the txt frequency response file, it will draw the frequency response curve. But in previous attempts, I could not get that to export to an impulse response, it would give me an error that there was no data to export.

Turns out you have to generate the minimum phase so it has some phase info. In REW in the SPL&Phase screen, you can go to controls and "generate minimum phase" for the plot.

Once you do that you can then export the IR for the plot. That produced another problem, as it didn't produce the desired results. Turns out it was due to the original SPL plot being at some high SPL level. To resolve this situation, I edited the SPL txt file to subtract the highest SPL level in the graph from every SPL level. When you then import that into REW, the highest level on the graph is at 0db. This effectively creates an IR with subtractive EQ as it's lowering every frequency except the highest one. Or at least that's what I think it's doing.

Putting my convolution reverb on the output bus and importing that IR, it indeed imparts the correct EQ curve onto my sound.

To prove this out, here are a few graphs.

First, here is the frequency response graph from REW, produced from SPL Tracing a published frequency response graph of a Yamaha NS10, then modifying that to a max SPL of 0db, then importing that graph into REW.

Yamaha NS10 Freq Response.JPG



Next up, I loaded a white noise wav file into my DAW (Studio One), and added a spectrum analyzer to the main output. Here is a capture of the spectrum analyzer while playing the white noise, showing all frequencies at the same SPL.

Yamaha NS10 IR Not Active.png



Next I added my convolution reverb to the main output and loaded the IR created from REW. Here's the spectrum analyzer capture.

Yamaha NS10 IR Active.png



Success!!
 
OP
G

gabo4au

Member
Joined
Apr 18, 2022
Messages
57
Likes
52
I do think it is.
people use NS10s and Auratones mainly to focus on the midrange. though the phase response is also a factor that you obviously can't simulate

Solved, so no worries. Thanks.
 

ctakim

Active Member
Forum Donor
Joined
Jan 3, 2021
Messages
148
Likes
195
Location
Menlo Park, CA
Interesting thread!!! I'm trying to do something similar with my relatively new set up of Dirac Live 3.3 on my MiniDSP SHD Studio feeding a 2.1 desktop system fed digitally from my computer USB out or via Roon over ethernet at my home computer desk. Speakers are Genelec 8341As and the subwoofer is a REL T7i. It is a nearfield set up and I use REW and a UMIK-1 to PEQ the speakers, set the crossover, and then I apply DL to this setup. The DL implementation on the MiniDSP has opened up new horizons because I can play with various target curves and listen to the results. The standard minimally sloping standard curve is very clean and enjoyable but I've also experimented with various Harman house curves with bass boosts of 4, 6, 8 and 10 dB. In general, I prefer the 4db bass boost with a relatively flat frequency response at the higher frequencies. I've also started to explore copying various "colored" FR curves from other speakers, such as the LS3/5a to see what differences I can discern, although I know you cannot exactly mimic the sonic characteristics of different speakers from the on axis FR alone. But the four built in presets in the MiniDSP that can easily be switched while listening in real time via remote has been great.
 

phoenixdogfan

Major Contributor
Forum Donor
Joined
Nov 6, 2018
Messages
3,297
Likes
5,079
Location
Nashville
Best bet is to do it over headphones with a software/hardware unit that can capture the way the speaker interacts with the room, and your body/ears. The gold standard here is the Smyth A16, but other programs like HeSuVi seem to have their adherents and the price is right. (FreeWare!)
 
Top Bottom