Thanks for the info
I really do appreciate your feedback and ideas.
I did not expect to get this much qualified feedback on diyaudio
Yes I have been reading
http://zone.ni.com/devzone/cda/tut/p/id/4278
The challenge when making the coherence function seems to do the averaging. I'm thinking all the way integrated in the GUI.
------------------
I will be away for two weeks kayaking along the eastern cost of Sweden. My brain needs to settle down and get some idle time fishing and paddling.
So I will be back around the start of August
I really do appreciate your feedback and ideas.
I did not expect to get this much qualified feedback on diyaudio
Yes I have been reading
http://zone.ni.com/devzone/cda/tut/p/id/4278
The challenge when making the coherence function seems to do the averaging. I'm thinking all the way integrated in the GUI.
------------------
I will be away for two weeks kayaking along the eastern cost of Sweden. My brain needs to settle down and get some idle time fishing and paddling.
So I will be back around the start of August
Ferrit37 said:
In a reverberant environment this will occur as the HF get delayed into the reverberant field of the LFs. But theoretically delay is not part of coherence. This is a real quandry in room acoustics however since if that is true, then the reverberant field should not affect the coherence and it does. This means that more than simple reflection is going on as waves move through a room.
Coherence trace is very valuable when tuning systems. It's one of the best functions of SMAART software. I rely on it a lot.
It would be a great addition to HOLMImpulse to see coherence trace next to the frequency response.
When you see the coherence trace dip, you know you should not try to correct frequency response anomalies in that area, because the measurement is not reliable there. You should move the microphone, change FFT parameters, maybe apply pre-equalization to measurement signal to get more reliable reading. It's a great guiding tool for tuners in real environments.
When it comes to anechoic or semi anechoic measurements, coherence trace is not as important, because of the nature of the measurements, but it still could provide valuable data about S/N of the measurement environment.
It would be a great addition to HOLMImpulse to see coherence trace next to the frequency response.
When you see the coherence trace dip, you know you should not try to correct frequency response anomalies in that area, because the measurement is not reliable there. You should move the microphone, change FFT parameters, maybe apply pre-equalization to measurement signal to get more reliable reading. It's a great guiding tool for tuners in real environments.
When it comes to anechoic or semi anechoic measurements, coherence trace is not as important, because of the nature of the measurements, but it still could provide valuable data about S/N of the measurement environment.
Diffraction from speaker enclosure- no. I measured speakers using SMAART in a large empty space using ground plane method, and the coherence trace doesn't dip more than a few percent.
In my experience the worst offenders are nulls due to reflections/cancellation and loud ambient noise- like powerful air conditioning rumble.
In my experience the worst offenders are nulls due to reflections/cancellation and loud ambient noise- like powerful air conditioning rumble.
How about impedance measurement?
Similar to how ARTA does it with a series resistor and two feedback channels.
Not that it's anything wrong with ARTA it's just that I want to see the impulse of the impedance.
Cabinet resonance should show up nicely and so should reflections inside the case. Especially with the coming CSD.
As it might not show up in normal measurement but indicate a hidden problem.
It would also work great as near field feedback if you have two channels + mic available.
Would this not be an interesting way to analyze a cabinet or driver?
And I also like a way to get LCR values as a curve and not just samples like in ARTA.
The coherence feature seems very interesting.
Similar to how ARTA does it with a series resistor and two feedback channels.
Not that it's anything wrong with ARTA it's just that I want to see the impulse of the impedance.
Cabinet resonance should show up nicely and so should reflections inside the case. Especially with the coming CSD.
As it might not show up in normal measurement but indicate a hidden problem.
It would also work great as near field feedback if you have two channels + mic available.
Would this not be an interesting way to analyze a cabinet or driver?
And I also like a way to get LCR values as a curve and not just samples like in ARTA.
The coherence feature seems very interesting.
There is no such thing as an impedance impulse, because the impedance is strictly a frequency domain concept.
You can simulate the impedance by assuming a pure voltage source, and measuring the Impulse response of the current through the device. The impedance is then the inverse of the resulting current spectrum.
The current is best measured with a small resistor to ground on the side of the loudspeaker away from the amps positive. Many amps will allow this, but NOT ones that float the ground. The output ground must be true ground or you can blow up the amp.
The assumption of a pure voltage source is the limiting factor as there is a series resistor and the amp has a finite output impedance. But this technique does not require a two channel system - HOLMImpulse is not a two channel system and I suspect that it would add a lot of complexity to make it two channel.
You can simulate the impedance by assuming a pure voltage source, and measuring the Impulse response of the current through the device. The impedance is then the inverse of the resulting current spectrum.
The current is best measured with a small resistor to ground on the side of the loudspeaker away from the amps positive. Many amps will allow this, but NOT ones that float the ground. The output ground must be true ground or you can blow up the amp.
The assumption of a pure voltage source is the limiting factor as there is a series resistor and the amp has a finite output impedance. But this technique does not require a two channel system - HOLMImpulse is not a two channel system and I suspect that it would add a lot of complexity to make it two channel.
Exporting the FR as ascii
I've been trying to use the FR response as an .FRD file to bring into the FRD passive crossover designer. But I'm having no luck.
The file exports OK, all the values are there. But even if I remove the comments at the top, as usually needs to be done, it still doesn't work.
Looking at the exported file I do see that the level is referenced to 0dB. So I added 90dB to everything (in Excel) and tried again. Still no go.
Anyone had any luck getting a good FRD file from the HOLMImpulse export?
I've been trying to use the FR response as an .FRD file to bring into the FRD passive crossover designer. But I'm having no luck.
The file exports OK, all the values are there. But even if I remove the comments at the top, as usually needs to be done, it still doesn't work.
Looking at the exported file I do see that the level is referenced to 0dB. So I added 90dB to everything (in Excel) and tried again. Still no go.
Anyone had any luck getting a good FRD file from the HOLMImpulse export?
New file format?
I'm working on a new fileformat to replace the .hlm files.
My Idea is to save the IR using FLAC compression and a info file describing the IR
Advantages:
Disadvantages:
Preview the file and tell me what you think
http://www.holmacoustics.com/downloads/HOLMImpulse/misc/test.zip
I'm working on a new fileformat to replace the .hlm files.
My Idea is to save the IR using FLAC compression and a info file describing the IR
Advantages:
- Measurement files will only be 1/10 of current files
- Measurement files can be inspected without using HOLMImpulse eg. by dragging the file.flac to Audacity will show the IR immediately and by inspecting the file.info with a text-editor you can see information about the measurement
- It will allow you to edit/collect a measurement list without using HOLMImpulse
- The info-file can be edited by hand
- Very easy to extend the information in the info file without currupting the binary format.
Disadvantages:
- It takes time (2-3 sec) to save measurements
- When using flac-unsupported samplerates such as 192 kHz
then the IR will be saved as a wave-file. (Flac supports 8 32 44.1 48 96 kHz) - FLAC only supports 24 bit which corresponds to the accuracy of floating points. (The saved FLAC files will of course be normalized to unity to obtain the best bit-resolution)
Preview the file and tell me what you think
http://www.holmacoustics.com/downloads/HOLMImpulse/misc/test.zip
Re: New file format?
I haven't had any trouble with FLAC and 192kHz. I have noticed that instead of 32-bit fixed or float point it will be 24-bit though with a flac file.
An alternative to FLAC is wavpack. Upside to using wavpack would be bit depths above 24, smaller file size, faster compression. Downside would be slower decompression.
askbojesen said:
I haven't had any trouble with FLAC and 192kHz. I have noticed that instead of 32-bit fixed or float point it will be 24-bit though with a flac file.
An alternative to FLAC is wavpack. Upside to using wavpack would be bit depths above 24, smaller file size, faster compression. Downside would be slower decompression.
I just don't see the point in saving in any compressed format. Its not like we are all short of hard drive space. File size is just not that big a deal anymore. I wouldn't use it if it was an option. And it needs to be optional since I don't want to have to get Audacity to look at the data. Wav files are great as far as I'm concerned.
Re: New file format?
Ok, I can easily make it configurable that you will save wav-files and not flac
What do you think about the open format instead of the closed hlm-file?
Advantages:
When forcing the use of wave-files, there is really no disadvantages compared to the current hlm-file format.
gedlee said:
Ok, I can easily make it configurable that you will save wav-files and not flac
What do you think about the open format instead of the closed hlm-file?
Advantages:
- Measurement files can be inspected without using HOLMImpulse eg. by dragging the file.flac to Audacity or other will show the IR immediately and by inspecting the file.info with a text-editor you can see information about the measurement
- It will allow you to edit/collect a measurement list without using HOLMImpulse
- The info-file can be edited by hand
- Very easy to extend the information in the info file without currupting the binary format.
When forcing the use of wave-files, there is really no disadvantages compared to the current hlm-file format.
Open format is always good. I didn't consider FLAC as open format that's why I don't like it. Wav is open format.
I've used the program a lot now. I really like it and I recommend it all the time.
The only downside: its not easy to make impedance measurements. I use impedance a lot for subs, etc. There is a lot of information in the impedance. At the very least an "inverse" function so that a current measurement can be made into an impedance measurement by assuming a voltage source. Even better would be a calibration for the source so that it didn't even have to be a voltage source.
Bug: Each new measurement requires me to turn off "from locked" in the time-window settings in "options" for each measurement. This should default to whatever the last measurement was set at. (Maybe thats not possible now that I think about it.)
Found: I found out that the pre-impulse I was getting at 200 data points was due to crosstalk when I had the other channel connected to the output of the amp (like one would do in a two channel measurement). If others see this, just unplug that channel. The crosstalk is never as good as the measurement, although it is easy to get rid of.
I've used the program a lot now. I really like it and I recommend it all the time.
The only downside: its not easy to make impedance measurements. I use impedance a lot for subs, etc. There is a lot of information in the impedance. At the very least an "inverse" function so that a current measurement can be made into an impedance measurement by assuming a voltage source. Even better would be a calibration for the source so that it didn't even have to be a voltage source.
Bug: Each new measurement requires me to turn off "from locked" in the time-window settings in "options" for each measurement. This should default to whatever the last measurement was set at. (Maybe thats not possible now that I think about it.)
Found: I found out that the pre-impulse I was getting at 200 data points was due to crosstalk when I had the other channel connected to the output of the amp (like one would do in a two channel measurement). If others see this, just unplug that channel. The crosstalk is never as good as the measurement, although it is easy to get rid of.
New Release 1.3.0.0
Long time no see...
I have been re-working my analysis library and fileformat
http://www.holmacoustics.com/downloads/HOLMImpulse/ChangeLog.txt
Version 1.3.0.0 (2009-08-25)
Features/Changes:
* Fileformat - New Open format (Zipped, flac/wav files)
* Fileformat - Old packages <file>.hlm can still be opened)
* Fileformat - Full impulse is saved
* Analysis -> Measurement Data length only affect FFT size
* THD improved - Now calculating from 20th HD instead of 10th HD
* THD calculated on-the-fly when needed from the plotting
* Time zero autodetect improved at Nyquist frequency
* Waiting information improved (Hourglass when the application is working)
* Default setting: Normalize to 0dB not default
* Memory allocation now using HeapAloc/HeapFree
Bugfixes:
* Extrapolating Import of frequency response to 0Hz
Notes:
* Using new Windows SDK v7.0 library
Comments
The main difference is that I use the Impulse as the basis - not the frequency response.
Using Zipped FLAC/Wave format when saving measurements improves the transparency.
What next?
I'll be working on import the DUT response import
Long time no see...
I have been re-working my analysis library and fileformat
http://www.holmacoustics.com/downloads/HOLMImpulse/ChangeLog.txt
Version 1.3.0.0 (2009-08-25)
Features/Changes:
* Fileformat - New Open format (Zipped, flac/wav files)
* Fileformat - Old packages <file>.hlm can still be opened)
* Fileformat - Full impulse is saved
* Analysis -> Measurement Data length only affect FFT size
* THD improved - Now calculating from 20th HD instead of 10th HD
* THD calculated on-the-fly when needed from the plotting
* Time zero autodetect improved at Nyquist frequency
* Waiting information improved (Hourglass when the application is working)
* Default setting: Normalize to 0dB not default
* Memory allocation now using HeapAloc/HeapFree
Bugfixes:
* Extrapolating Import of frequency response to 0Hz
Notes:
* Using new Windows SDK v7.0 library
Comments
The main difference is that I use the Impulse as the basis - not the frequency response.
Using Zipped FLAC/Wave format when saving measurements improves the transparency.
What next?
I'll be working on import the DUT response import
Last edited:
Right now I use a Dayton Woofer Tester II to do impedance, and Arta for acoustic measurements. Would I gain anything by going to HolmImpulse? Based on what I've read here, the noise floor of my measurements may be lower, but that isn't 100% clear.
Above the noise floor, the measurements would be identical wouldn't they?
Also, my measurements are done in a car, where there's a TON of reflections. I do twelve measurements at a time, at 4 degree increments, because it helps me see what parts of the response are due to the speaker, and what parts are due to the car. I use a relatively short gating time for the same reason.
Would HolmImpulse be any better than Arta in that respect?
Just curious, as I'm very happy with Arta, and reluctant to learn another new program.