HOLMImpulse: Measuring Frequency & Impulse Response

Well there is not really 'anything' to learn as HOLMImpulse is a very easy program to setup and use. I suggest you give it a try. For the ease of comparing measurements if nothing else.
Did I mention it was easy to use?

ARTA is not a bad program and does a lot more than HOLM does at the moment. But for measurements and ease of use HOLM wins.
IMHO that is.


"There is no such thing as an impedance impulse, because the impedance is strictly a frequency domain concept."

If I put it this way. Is there a practical way to remove the signal going into the speaker and effectively using it as a "mic" by looking at the back EMF or other signals generated. Getting time domain info like any mic would.
Backwave, internal resonance, pipe resonances and other type of effects seen by the cone should show up.
The biggest problem now would be the speaker actually acting like a big mic picking up a lot of noise. Especially LF.
So removing that part would not be easy. Although a long or multiple sweeps might fix that.

Not sure if it is of any use when it comes to the concept of HOLMImpulse buy I just can't get it out of my head.
 
Administrator
Joined 2004
Paid Member
Hey Patrick I use the Dayton WT3 and Arta. Arta is very nice, but I found HOMImpulse easier to master. Not sure why. Of course you can export to Arta if you need further charting (I do).

HOLMIpulse is easy to lean and use. And it's free. I like being able to save the files - unlike the free version of Arta.
 
I've been trying for days to get a decent measurement out of speaker workshop.. finally downloaded HOLMImpulse and I think I'm finally looking at reasonable numbers for the first time. This is a lifesaver! Crossover design here I come!

Also, I was posting about my problem in the Software Tools forum, which is new, and is a little barren. This thread would be perfect for that forum. Perhaps an admin can move it over?
 
I also agree - Personally I use HOLMIMpulse for a lot besides multiway:

* Electrical measurements
* Soundcard validation
* Fullrange systems
* General validation of the relation between Impulse <-> Frequency

I also started the thread HOLMImpulse: Measurements in practice so that measurements of loudspeakers in general could be and that thread.

Bugs, feature requests, technical discussions should be in this thread

I'll think about it and contact a moderator
 
"..... impedance is strictly a frequency domain concept."


Just browsing this thread. I don't know where the above quote was initiated but it is not entirely correct. Impedance is defined for any wave form that can be expressed by an exponential such as V = Vo exp(st), I = Io exp(st), s = a +jb. This leads to the well known relationships,

Zr = R,

Zc = 1/(sC)

and ZL= sL

s need not be limited to s = jw.
 
HOLMImpulse > ARTA

What I find very smart in HOLMImpulse, is that you have 50 measurements in a collection where you can compare up to 3 at the same time
In ARTA you can only compare two impulse-response in the same graph using overlays.

With the new fileformat it is very easy to import the recorded impulse responses to ARTA:

1. HOLMImpulse > File > Save Measurements (WAV)
(This will create a zip-file with all the measurments)

2. Unzip the measurement wave-file you want to import in ARTA

3. ARTA > File > Import .Wav file. Now make eg. your CSD etc.. Which I have not yet implemented in HOLMImpulse
 
Administrator
Joined 2004
Paid Member
multi-pass

New version seems to work great. The file saving in FLAC or WAV is nice.
Keeping response uscaled by default is also good.

One thing that went away. The average of multiple passes. Can we have it back please? I always used 3 or 4 passes for my speaker measurements.
 
Problem with stiching phase

I'd like to get people's opinions on this. I'm suprised noone's posted about it yet, I searched the thread and found nothing about stitching or phase problems.


I'm doing near field and gated far field measurements to get a complete picture of the driver's response in its enclosure. Standard operating procedure. After getting decent measurements in each configuration, I used the modify menu to stich the two together into a new complete response graph.

As the stitching tool mentions that it should both match amplitude AND phase before stitching, I'm assuming that I should get a nice resulting phase graph that blends from one measurement to the next, just as with the blending of the amplitude.

This seems not to be the case. Whenever I do a stich between two measurements, the phase in the resulting measurement EXPLODES IN A WRATHFUL BALL OF DOTTED LINES.

*ahem*

Let me simply demonstrate with a screenshot! (The built-in image capture feature is awesome btw)

The first image is the two measurements, with both amplitude and phase displayed, before the stitch

The second image is all three measurements; the near field, far field, and resulting stich; displayed. Obviously one can't make out much as the phase on the stitch is all over the place. The amplitude, btw, is a perfect blend between the two at 800hz with a 200hz overlap. In this instance I checked "match amplitude" and "match phase" before performing the stitch, but I have tried this with all combinations of different frequencies and different settings for the match, and everything behaves normally, but the phase always explodes as shown.
 

Attachments

  • a5-prestich.png
    a5-prestich.png
    61.6 KB · Views: 371
  • a5-postich.png
    a5-postich.png
    163.6 KB · Views: 360
More phase stuff

I'm building an MTM so my workflow for generating a response for my woofers is as follows:

1. Nearfield measurement of woofer 1
2. Nearfield measurement of woofer 2
3. Sum measurements of woofers 1 & 2
4. Farfield measurement of both woofers together
5. Stich the Summed measurement from step 3 and gated far-field from 4.

Ignoring phase, this is still fine and good from a frequency standpoint, until I get to the final step. It appears that the stitch function also refers to the phase calculation when blending the amplitude from one measurement to another, and creates a frequency null if the drivers are out of phase, as would happen in a crossover situation.

The problem is that when using the "Sum" function to add the two nearfield measurements, the resulting phase graph is all over the place, so when doing the final stitch, unless I use a stich width of "1" I get a huge dip in response, which I'm guessing comes from bizarre phase information generated from the Sum function.

Not sure if this is how it should behave or not. Either way, here's a screenshot of what I'm seeing. The first two graphs (blue and red) are nearfield measurements of both drivers in the box, which are appropriately similar in both amplitude and phase (they are not normalized in the measurement options). The third measurement (green) is the Summed measurement of the first two, and the resulting phase.
 

Attachments

  • da175-sum-phase.png
    da175-sum-phase.png
    71.4 KB · Views: 356
New version with Bug fixes 1.3.0.4

This seems not to be the case. Whenever I do a stich between two measurements, the phase in the resulting measurement EXPLODES IN A WRATHFUL BALL OF DOTTED LINES.

It was a bug. - Stitching now works again

Version 1.3.0.4 (2009-08-28)

Features/Changes:
* Lowpass/Highpass variable length
* Old autosaved file (MeasAutosaved.hlm) is moved to AutosavedObsolete
* QBox fit - Automatic plot
* Options - Sample offset increment
* Logarithmic frequency axis - Grid improved
* Export - Overwrite confirm dialog (Removed from browse)

Bugfixes:
* Stitch Measurement C = A -> B
* Impulse domain: Auto Zoom
* Options: Clear measurement resets values
* Clear measurements and close does not overwrite autosaved measurements
* Memory leaks
 
New Release 1.3.0.6 - Import of DUT response

And with the complete import option as you plan it one could contercheck if the processing/DUT is really consistent with the linear superposition principle, that is if the chains
a) signal --> response --> user filter
b) signal --> user filter --> response
are identical in the basic response (but not in the S/N and in the distortion, that is obvoius). Which then yould be a nice check for the LTI-quality of the system under test, etc

- Klaus

Version 1.3.0.6 (2009-09-01)

Features/Changes:
* Import of DUT response Import > 'Signal & Recording'


------------------------------------------------------------

Attached picture shows the response of 30sec mp3 encoded music
 

Attachments

  • ScreenShot005.png
    ScreenShot005.png
    60.4 KB · Views: 1,254
Last edited:
Music as measurement signal

What I need next is to play music through my speakers and correlate with a recording.

Another attachment shows the difference between 128 kbs, 320 kbs mp3 -
Does anybody know where a new thread regarding frequency response of encoded files would belong? Or If it is old news - where can I read about it.
 

Attachments

  • pic1.png
    pic1.png
    57 KB · Views: 1,262
Hmm I dunno exactly where to read about it. First thing you'll maybe want to do is find the er "best sounding" mp3 algorithm or at least the most popular ones (LAME for instance) and test them. But a friend and I (more him than me) have done a little experimentation. There could be a couple of commonly overlooked variables imo when using mp3s. I think some of them clip the signal if you don't drop the gain a little before encoding. Something to do with floating point - please feel free to tell him if I am wrong everybody because this is more what I have seen people report and I think I did notice once or twice back when I dealt more with lossy.

But a lot of it will have to do with perceptual encoding. I like testing every step of the chain in a recording but I have not been brave enough to do it with the common mastering techniques combined with lossy encoding. I pretty much predict it will be VERY messy.
 
Last edited: