brucek
Second Unit
- Joined
- Dec 29, 1998
- Messages
- 335
I think most participants here know what a BFD is, and if you don't, then you can read all about it on this web site first.
Link Removed edit new site URL
A low frequency response test for the purpose of monitoring the effect of entering equalization data into a parametric equalizer for a lot of us involves the so called "manual method".
This involves burning a CD with a sequential set of several second 1/6th octave tones, and then playing these tones through our sub(s), and measuring and recording the resultant SPL level at the listening position.
These readings are entered into an Excel chart that compensates for the deficiencies of the Radio Shack SPL meter, and then by observing the created frequency response graph we can determine what filter adjustments need to be made to make the graph behave the way we want.
This "manual method" generally works well, and no matter how easy and inexpensive it is, it has the disadvantage of taking an enormous amount of time.
Enter the ETF5 acoustic measurement software. This is a PC based software program that allows you to enjoy instantaneous readouts of your frequency response on your PC combined with a RS SPL meter. The company allows you to download a demo version that is surprisingly complete, but you can't save any of your files, although you can do a windows screen save to preserve your results. I thought it would be interesting to test this method, since there have been so many posts on this forum praising ETF5. Their web site, for further reading and understanding this software is at:
http://www.etfacoustic.com/
When I initially downloaded this software it would not work on my computer, but a subsequent reload of my soundcard drivers fixed that problem - duh. There is a nice web site from a member on this forum that is quite good regarding the hookup and use of this software at the site below, although it would be nice if there was a little more detail. I don't think he's finished...
http://www27.brinkster.com/jmag999/
Anyway, if you think the BFD has a large learning curve, well it's a walk in the park compared to this software and its full use. There is a lot to understand with this software and I decided to restrict my use to low frequency response tests to compare the resultant graphs to my "manual method" graphs. I was also interested in comparing a set of filters created using ETF5 with my set of filters created with the manual method.
Once you've mastered the hookup and use of this software, it's a joy to adjust a BFD filter and see the frequency response curve change accordingly in pseudo real time. The software has a sequential mode, where once its all set up and parameters and levels set, you can one-shot or take a reading automatically every 1.5 seconds while you watch your frequency response in essentially real time on the computer monitor while you adjust your filters. It also allows you to walk around your room with your SPL mic with its cable attached and see the effects on your response.
The ETF5 program corrects for soundcard deficiencies because one of the stereo channels of the sound card is shorted from input to output for correction. You can read about this on their web site.
One of the debatable points about the use of ETF5 is the use of a calibration file for the Radio Shack SPL meter. There is a calibration file on the ETF site, but it only calibrates above 500Hz. Not much use for what I want to do. I've read that ETF support has said, this is because the microphones they use (and sell) possess flat response down to 20Hz and that ETF won't work with values less than 500Hz. Well, this hasn't been my experience. It works just fine with all the low frequency values.
The "brinkster" ETF site mentioned above says to use the standard RS SPL correction values we normally use and even provides the file on the site for download. I feel the file on that site is actually backwards as far as its correction values are concerned. I'll explain....
I have reprinted the rs.cal correction file from the http://www27.brinkster.com/jmag999/ site. It is listed below for reference. This matches what I normally use for the "manual method" and is quite correct for that use. The rs.cal file is tabled in the order of frequency, then correction value, then phase. For example, when using the manual method, I take a reading at 20Hz, I add +7.5dBSPL to the reading because that's how much the RS SPL meter needs to be corrected at that frequency.
*****************************
10.000 20.000 0.000
12.500 16.500 0.000
16.000 11.500 0.000
20.000 7.500 0.000
25.000 5.000 0.000
31.500 3.000 0.000
40.000 2.500 0.000
50.000 1.500 0.000
63.000 1.500 0.000
80.000 1.500 0.000
100.000 2.000 0.000
125.000 0.500 0.000
160.000 -0.500 0.000
200.000 -0.500 0.000
250.000 0.500 0.000
315.000 -0.500 0.000
400.000 0.000 0.000
****************************
But, my observation is that this file above is completely backwards for ETF use and that the ETF program uses a negative logic on the compensation value numbers. And so if I have a compensation factor at 20Hz of +7.5, the entry in the rs.cal file should be 20.000 -7.500 0.000 and not 20.000 7.500 0.000
I come to this conclusion three ways. One is by observing the low frequency response graph, (while the rs.cal is loaded), during the initial setup level check of ETF5 to correct for your sound card deficiencies with the "both sound card channels shorted". The corrected sound card graph seemed exactly backwards if I entered the values as in the chart above. The corrected "perfect sound card" with rs.cal calibration should show 7.5dB high at 20Hz to correct for the SPL meter at that frequency. It showed exactly opposite this at -7.5dB low.
The second most telling indication is through experimentation with a single value to see what happens when I change it in the rs.cal file. For example, if I set the compensation value for 20Hz at 0.0000 and take a reading and observe the resultant value on the low frequency response graph, I could say, "the value on the low frequency response graph is too low because at 20Hz the RS meter is weak by a value of 7.5dBSPL". I want to compensate and "raise" the value on the low frequency response graph by 7.5dBSPL. If I enter the value in the rs.cal graph of 20.000 -7.500 000 it does indeed raise it correctly.
A third reason to verify that this is correct thinking is that if you look at the rs.cal file on the ETF web site that starts at 500Hz, I notice that as the frequency gets higher and higher, the chart values get progressively more and more negative. That would seem reasonable to compensate for the SPL meter since the higher the frequency, the worst the response.
Anyway, below are the values I used for my rs.cal file. Seemed to work pretty well. You could just use this file alone if you only care about low frequencies or simply preface the rs.cal from the ETF web site which starts at 500Hz. The ETF program evenly extrapolates the values missing between the entered values.
***************************
correct rs.cal file I use..
***************************
10.000 -20.000 0.000
12.500 -16.500 0.000
16.000 -11.500 0.000
20.000 -7.500 0.000
25.000 -5.000 0.000
31.500 -3.000 0.000
40.000 -2.500 0.000
50.000 -1.500 0.000
63.000 -1.500 0.000
80.000 -1.500 0.000
100.000 -2.000 0.000
125.000 -0.500 0.000
160.000 0.500 0.000
200.000 0.500 0.000
250.000 -0.500 0.000
315.000 0.500 0.000
400.000 0.000 0.000
****************************
One of the problems with using ETF5 of course, (besides the number of cables and special plugs required) is hauling your computer out to your HT area and setting it up. This could be enough to stop a lot of people from even considering using this method. I also ignored the fact that perhaps my room acoustics would be affected by my computer and monitor being in the middle of the room.
I fed the output of my computer soundcard into my processor using a "Y" cable and set the system to my usual crossover of 60Hz and shut my main power amp off. The managed sub signal then feeds my BFD and then my X-30 and Servo-15 sub.
Anyway, without any filters, the low frequency response graph produced by ETF5 was quite close to my manual method graph. This was encouraging. I paid special attention to my graph scaling of both methods so that I was comparing apples and apples. There was an anomaly in my ETF graph which was a very narrow negative spike up at 83Hz that I attribute to a small consistent tick sound in the ETF MLS signal that I attribute to my challenged sound card and computer. I ignored the small tick and resultant spike. You also have to be careful about the level setting of your record and volume control settings because you can get into a regenerative feedback situation and your system will break into oscillation. Careful setup of the record and playback levels is important.
I placed the ETF5 in automatic RTA Sequential Data Acquisition mode and entered a new set of filters to tame the peaks, just as I had done when using the manual method except now in about a tenth of the amount of time to do the job. This is where ETF5 shines. While you make the adjustment to any filter parameter, the response graph changes. The interesting thing here is that it gives you "eyes" to see how a filter at one frequency affects other frequencies around it, allowing you to "try" different filter parameters to get the smoothest response. I admit the manual method doesn't give you this insight or instant feedback control. It was interesting when I was using ETF5 and entering filters, particularly at one frequency, it wasn't the frequency at the top of the peak on the response graph that was the most effective center filter frequency to use. When using the manual method I always chose the frequency at the top of a peak.
Once I had completed my filters and the response was equalized to my satisfaction, I then changed the BFD program to the filters I had been using from the manual method to see what the response looked like using the ETF5 program. Well, the frequency response indicated by ETF5 while using my manual method filters wasn't quite as good as the set of filters I had just created using the ETF5 method. Both the ETF5 and the manual method ended up needing 4 filters to produce an equalized response. When I compare the 4 filters I created using the ETF5 against the 4 filters from the manual method they are fairly close, but different enough to make the ETF5 frequency response look much better using the filters I created with EFT5.
Now here's the interesting thing. If I then do a frequency response using the manual method and use the filters I created with ETF5, the resultant Excel graph looks poor, but looks great when I use my old filters I created earlier with the manual method.
So, I guess my question is, which method is correct and what set of filters do you use. ETF5 filters look great when I use ETF and the manual method filters look great when I use the manual method, but not vise-versa for each case. I'm not saying it's that far off, but different enough to make you wonder.
So for me, there's no clear winner. The problem with ETF5 is that it has a steep learning curve, and also requires a move of your computer system into your HT room. If you can overcome these two hurdles and can come up with the price of the full version and perhaps a better quality microphone, it would be a nice thing to own. I always advocate buying software rather than getting frustrated with partial functionality. It's nice of ETF to allow people to evaluate ETF5. Certainly the program is far more capable than just doing simple low frequency response tests.
For myself, I don't do this sort of measurement enough to justify the cost. I'm fairly quick with the manual method and it seems effective enough.
brucek
Link Removed edit new site URL
A low frequency response test for the purpose of monitoring the effect of entering equalization data into a parametric equalizer for a lot of us involves the so called "manual method".
This involves burning a CD with a sequential set of several second 1/6th octave tones, and then playing these tones through our sub(s), and measuring and recording the resultant SPL level at the listening position.
These readings are entered into an Excel chart that compensates for the deficiencies of the Radio Shack SPL meter, and then by observing the created frequency response graph we can determine what filter adjustments need to be made to make the graph behave the way we want.
This "manual method" generally works well, and no matter how easy and inexpensive it is, it has the disadvantage of taking an enormous amount of time.
Enter the ETF5 acoustic measurement software. This is a PC based software program that allows you to enjoy instantaneous readouts of your frequency response on your PC combined with a RS SPL meter. The company allows you to download a demo version that is surprisingly complete, but you can't save any of your files, although you can do a windows screen save to preserve your results. I thought it would be interesting to test this method, since there have been so many posts on this forum praising ETF5. Their web site, for further reading and understanding this software is at:
http://www.etfacoustic.com/
When I initially downloaded this software it would not work on my computer, but a subsequent reload of my soundcard drivers fixed that problem - duh. There is a nice web site from a member on this forum that is quite good regarding the hookup and use of this software at the site below, although it would be nice if there was a little more detail. I don't think he's finished...
http://www27.brinkster.com/jmag999/
Anyway, if you think the BFD has a large learning curve, well it's a walk in the park compared to this software and its full use. There is a lot to understand with this software and I decided to restrict my use to low frequency response tests to compare the resultant graphs to my "manual method" graphs. I was also interested in comparing a set of filters created using ETF5 with my set of filters created with the manual method.
Once you've mastered the hookup and use of this software, it's a joy to adjust a BFD filter and see the frequency response curve change accordingly in pseudo real time. The software has a sequential mode, where once its all set up and parameters and levels set, you can one-shot or take a reading automatically every 1.5 seconds while you watch your frequency response in essentially real time on the computer monitor while you adjust your filters. It also allows you to walk around your room with your SPL mic with its cable attached and see the effects on your response.
The ETF5 program corrects for soundcard deficiencies because one of the stereo channels of the sound card is shorted from input to output for correction. You can read about this on their web site.
One of the debatable points about the use of ETF5 is the use of a calibration file for the Radio Shack SPL meter. There is a calibration file on the ETF site, but it only calibrates above 500Hz. Not much use for what I want to do. I've read that ETF support has said, this is because the microphones they use (and sell) possess flat response down to 20Hz and that ETF won't work with values less than 500Hz. Well, this hasn't been my experience. It works just fine with all the low frequency values.
The "brinkster" ETF site mentioned above says to use the standard RS SPL correction values we normally use and even provides the file on the site for download. I feel the file on that site is actually backwards as far as its correction values are concerned. I'll explain....
I have reprinted the rs.cal correction file from the http://www27.brinkster.com/jmag999/ site. It is listed below for reference. This matches what I normally use for the "manual method" and is quite correct for that use. The rs.cal file is tabled in the order of frequency, then correction value, then phase. For example, when using the manual method, I take a reading at 20Hz, I add +7.5dBSPL to the reading because that's how much the RS SPL meter needs to be corrected at that frequency.
*****************************
10.000 20.000 0.000
12.500 16.500 0.000
16.000 11.500 0.000
20.000 7.500 0.000
25.000 5.000 0.000
31.500 3.000 0.000
40.000 2.500 0.000
50.000 1.500 0.000
63.000 1.500 0.000
80.000 1.500 0.000
100.000 2.000 0.000
125.000 0.500 0.000
160.000 -0.500 0.000
200.000 -0.500 0.000
250.000 0.500 0.000
315.000 -0.500 0.000
400.000 0.000 0.000
****************************
But, my observation is that this file above is completely backwards for ETF use and that the ETF program uses a negative logic on the compensation value numbers. And so if I have a compensation factor at 20Hz of +7.5, the entry in the rs.cal file should be 20.000 -7.500 0.000 and not 20.000 7.500 0.000
I come to this conclusion three ways. One is by observing the low frequency response graph, (while the rs.cal is loaded), during the initial setup level check of ETF5 to correct for your sound card deficiencies with the "both sound card channels shorted". The corrected sound card graph seemed exactly backwards if I entered the values as in the chart above. The corrected "perfect sound card" with rs.cal calibration should show 7.5dB high at 20Hz to correct for the SPL meter at that frequency. It showed exactly opposite this at -7.5dB low.
The second most telling indication is through experimentation with a single value to see what happens when I change it in the rs.cal file. For example, if I set the compensation value for 20Hz at 0.0000 and take a reading and observe the resultant value on the low frequency response graph, I could say, "the value on the low frequency response graph is too low because at 20Hz the RS meter is weak by a value of 7.5dBSPL". I want to compensate and "raise" the value on the low frequency response graph by 7.5dBSPL. If I enter the value in the rs.cal graph of 20.000 -7.500 000 it does indeed raise it correctly.
A third reason to verify that this is correct thinking is that if you look at the rs.cal file on the ETF web site that starts at 500Hz, I notice that as the frequency gets higher and higher, the chart values get progressively more and more negative. That would seem reasonable to compensate for the SPL meter since the higher the frequency, the worst the response.
Anyway, below are the values I used for my rs.cal file. Seemed to work pretty well. You could just use this file alone if you only care about low frequencies or simply preface the rs.cal from the ETF web site which starts at 500Hz. The ETF program evenly extrapolates the values missing between the entered values.
***************************
correct rs.cal file I use..
***************************
10.000 -20.000 0.000
12.500 -16.500 0.000
16.000 -11.500 0.000
20.000 -7.500 0.000
25.000 -5.000 0.000
31.500 -3.000 0.000
40.000 -2.500 0.000
50.000 -1.500 0.000
63.000 -1.500 0.000
80.000 -1.500 0.000
100.000 -2.000 0.000
125.000 -0.500 0.000
160.000 0.500 0.000
200.000 0.500 0.000
250.000 -0.500 0.000
315.000 0.500 0.000
400.000 0.000 0.000
****************************
One of the problems with using ETF5 of course, (besides the number of cables and special plugs required) is hauling your computer out to your HT area and setting it up. This could be enough to stop a lot of people from even considering using this method. I also ignored the fact that perhaps my room acoustics would be affected by my computer and monitor being in the middle of the room.
I fed the output of my computer soundcard into my processor using a "Y" cable and set the system to my usual crossover of 60Hz and shut my main power amp off. The managed sub signal then feeds my BFD and then my X-30 and Servo-15 sub.
Anyway, without any filters, the low frequency response graph produced by ETF5 was quite close to my manual method graph. This was encouraging. I paid special attention to my graph scaling of both methods so that I was comparing apples and apples. There was an anomaly in my ETF graph which was a very narrow negative spike up at 83Hz that I attribute to a small consistent tick sound in the ETF MLS signal that I attribute to my challenged sound card and computer. I ignored the small tick and resultant spike. You also have to be careful about the level setting of your record and volume control settings because you can get into a regenerative feedback situation and your system will break into oscillation. Careful setup of the record and playback levels is important.
I placed the ETF5 in automatic RTA Sequential Data Acquisition mode and entered a new set of filters to tame the peaks, just as I had done when using the manual method except now in about a tenth of the amount of time to do the job. This is where ETF5 shines. While you make the adjustment to any filter parameter, the response graph changes. The interesting thing here is that it gives you "eyes" to see how a filter at one frequency affects other frequencies around it, allowing you to "try" different filter parameters to get the smoothest response. I admit the manual method doesn't give you this insight or instant feedback control. It was interesting when I was using ETF5 and entering filters, particularly at one frequency, it wasn't the frequency at the top of the peak on the response graph that was the most effective center filter frequency to use. When using the manual method I always chose the frequency at the top of a peak.
Once I had completed my filters and the response was equalized to my satisfaction, I then changed the BFD program to the filters I had been using from the manual method to see what the response looked like using the ETF5 program. Well, the frequency response indicated by ETF5 while using my manual method filters wasn't quite as good as the set of filters I had just created using the ETF5 method. Both the ETF5 and the manual method ended up needing 4 filters to produce an equalized response. When I compare the 4 filters I created using the ETF5 against the 4 filters from the manual method they are fairly close, but different enough to make the ETF5 frequency response look much better using the filters I created with EFT5.
Now here's the interesting thing. If I then do a frequency response using the manual method and use the filters I created with ETF5, the resultant Excel graph looks poor, but looks great when I use my old filters I created earlier with the manual method.
So, I guess my question is, which method is correct and what set of filters do you use. ETF5 filters look great when I use ETF and the manual method filters look great when I use the manual method, but not vise-versa for each case. I'm not saying it's that far off, but different enough to make you wonder.
So for me, there's no clear winner. The problem with ETF5 is that it has a steep learning curve, and also requires a move of your computer system into your HT room. If you can overcome these two hurdles and can come up with the price of the full version and perhaps a better quality microphone, it would be a nice thing to own. I always advocate buying software rather than getting frustrated with partial functionality. It's nice of ETF to allow people to evaluate ETF5. Certainly the program is far more capable than just doing simple low frequency response tests.
For myself, I don't do this sort of measurement enough to justify the cost. I'm fairly quick with the manual method and it seems effective enough.
brucek