REW will also show just how much boost/cut you actually need to flatten the response. My goal is to not boost narrow dips so I changed some settings in the template and look at the curves prior to using DRC to determine the amount of boost needed. Too much and it sounds forced.
Did you get more focus in the imaging? Are there some things you like about what it's doing? There are so many things you can change.
Seeing what it is doing in REW helped me a lot to relate what I'm hearing to what it is doing. The Normal template has quite long window settings from low to high frequencies. On my first try I ran all standard filters and listened to each one to see if there was something I liked. From there on I used custom settings and played with targets.
The final result you get is from the speaker in your room. That combination determines the correction needed. I had a big downwards slope prior to hanging two more damping panels in the room. It's much more flat than it used to be now. It could be that your room has something going in the mid to low frequency range that's making it sound boomy and/or boxy. If you have dip from a floor bounce that's being boosted back to flat it will never sound quite right.
Looking at REW measurements will tell you more than a blind gamble though.
Something like this chart comes in handy sometimes:
http://www.independentrecording.net/irn/resources/freqchart/main_display.htm
What I'm looking for is what my speaker does AND what my room does. If I gate measurements from REW I can get a feel for the direct speaker response and a different gating can teach you a lot about your listening room. I showed that a couple of pages back. It was the reason I expected better results with more damping panels in the room. I had way too much relatively late high frequency reflections from a very close back wall (behind listening position).
Nothing wrong with unchecking the Normalisation. Just as long as you keep watching the total level. Can be seen in JRiver -> DSP Studio -> Analyzer. Keep it under 100% peaks with all music.
Did you get more focus in the imaging? Are there some things you like about what it's doing? There are so many things you can change.
Seeing what it is doing in REW helped me a lot to relate what I'm hearing to what it is doing. The Normal template has quite long window settings from low to high frequencies. On my first try I ran all standard filters and listened to each one to see if there was something I liked. From there on I used custom settings and played with targets.
The final result you get is from the speaker in your room. That combination determines the correction needed. I had a big downwards slope prior to hanging two more damping panels in the room. It's much more flat than it used to be now. It could be that your room has something going in the mid to low frequency range that's making it sound boomy and/or boxy. If you have dip from a floor bounce that's being boosted back to flat it will never sound quite right.
Looking at REW measurements will tell you more than a blind gamble though.
Something like this chart comes in handy sometimes:
http://www.independentrecording.net/irn/resources/freqchart/main_display.htm
What I'm looking for is what my speaker does AND what my room does. If I gate measurements from REW I can get a feel for the direct speaker response and a different gating can teach you a lot about your listening room. I showed that a couple of pages back. It was the reason I expected better results with more damping panels in the room. I had way too much relatively late high frequency reflections from a very close back wall (behind listening position).
Nothing wrong with unchecking the Normalisation. Just as long as you keep watching the total level. Can be seen in JRiver -> DSP Studio -> Analyzer. Keep it under 100% peaks with all music.
Last edited:
Good suggestion! Although I have not needed to do that. The average of a multi point measurement mimics what I measure at the single sweet spot.
I have used the moving mic measurement in my Car long ago though. Very helpful as the sound changed a lot by moving only a fraction in that environment. Nothing that bothersome happening with the arrays in my room. Could be useful for single point speakers. Easy enough to try.
I have used the moving mic measurement in my Car long ago though. Very helpful as the sound changed a lot by moving only a fraction in that environment. Nothing that bothersome happening with the arrays in my room. Could be useful for single point speakers. Easy enough to try.
Last edited:
My first experience with DRC (about 8 years ago using an early DRC GUI called ACXO Player) was a complete disaster. Then I learned the importance of using the right kind of microphone 
. But even then, the initial results weren't so good and it took a bit of time to get a handle on the process.
wesayso has offered some good advice regarding IR measurements. It's crucial to get this part right since every subsequent step is based on it. One additional consideration might be the polar pattern of the "omni" mic. I use the Dayton EMM-6 which becomes increasingly directional above 2khz. Since I use short windows (aiming more at speaker correction than room correction) and listen close to the speakers, the direct sound seems to dominate and I can get good results by pointing the mic directly at each speaker during the sweep. This is more work since the mic needs to be slightly re-positioned between sweeps but it should only need to be done once.
My Advice as far as setting DRC parameters would be to start with short windowing and a flat frequency response. Begin with the "minimal" preset and edit the associated configuration file (minimal44100 with DRC Designer) to bypass the psychoacoustic target stage (set PTType to N). I bypass the ringing truncation stage as well. If the resulting filter doesn't improve at least some aspect of the sound then there is most likely a problem somewhere.
Also, as wesayso has mentioned, it's very important to have visual feedback for each step of the process. This is another topic however...
. But even then, the initial results weren't so good and it took a bit of time to get a handle on the process. wesayso has offered some good advice regarding IR measurements. It's crucial to get this part right since every subsequent step is based on it. One additional consideration might be the polar pattern of the "omni" mic. I use the Dayton EMM-6 which becomes increasingly directional above 2khz. Since I use short windows (aiming more at speaker correction than room correction) and listen close to the speakers, the direct sound seems to dominate and I can get good results by pointing the mic directly at each speaker during the sweep. This is more work since the mic needs to be slightly re-positioned between sweeps but it should only need to be done once.
My Advice as far as setting DRC parameters would be to start with short windowing and a flat frequency response. Begin with the "minimal" preset and edit the associated configuration file (minimal44100 with DRC Designer) to bypass the psychoacoustic target stage (set PTType to N). I bypass the ringing truncation stage as well. If the resulting filter doesn't improve at least some aspect of the sound then there is most likely a problem somewhere.
Also, as wesayso has mentioned, it's very important to have visual feedback for each step of the process. This is another topic however...
Last edited:
One note I might add... when you use "Generate Custom Filters" in DRC Designer it seems to base all changes on the "Soft" template. No matter which radio button is clicked, the template used will still be the "Soft" template followed by the appropriate Sampling Rate. The changes in settings are made by adding/changing variables of that template.
You can see that behaviour in the batch files created by DRC Designer. Nice to know if you start editing templates.
You can see that behaviour in the batch files created by DRC Designer. Nice to know if you start editing templates
.
for a snappy responsive computer hope it sounds as good or better as the old one without too much trouble and tweaking and thanks DRC tips.
Up and running again. Definitely more snappy than the old computer and so far it sounds good. Needed some tweaks to get everything running as this new workstation didn't have a Molex feed for the Xonar soundcard. I reluctantly used a Sata to Molex power adapter but it works. Sata power connector does not have as much current capacity as the older Molex 4 pins. No idea how much the Xonar ST is drawing though. So far so good. New processor should be 3x faster than the old one. Nvidia Quadro K2000 Graphics card is the slowest part now
.
Great you faster now and up again in past did nearly same Q6600 to Xeon E3-1270 V2, not that WEI Win7 was whoopee 7,1-7,7 but reality is very snappy.
Think that K2000 is plenty had the non pro version both GT640 and GTX650 and now a GTX750Ti, done some subjective experiments and experience my gear is the less TDP for GPU the better sound quality, at my dedicated audio computer run a ASUS GT720 with only 19w TDP if i change to GT640 GTX650 or higher TDP GPU its audio able.
in past did nearly same Q6600 to Xeon E3-1270 V2, not that WEI Win7 was whoopee 7,1-7,7 but reality is very snappy.Think that K2000 is plenty had the non pro version both GT640 and GTX650 and now a GTX750Ti, done some subjective experiments and experience my gear is the less TDP for GPU the better sound quality, at my dedicated audio computer run a ASUS GT720 with only 19w TDP if i change to GT640 GTX650 or higher TDP GPU its audio able.
Well, definitively something wrong going on.
I took a measurement with REW before (bypassing everything, and no sub inputs) shown in red.
Then, plugged in the subs and made a new sweep with DRC Designer. Spit out all the filters and input them into JRiver Convolution plugin, and listened to all of them. They all show the same behavior, in different intensity.
I used REW to make another sweep to see what was going on. this time using JRiver and the convolution filters applied to the signal.
Pretty funny.... NOT!
This is the sweep I get, in blue.
Just imagine my face when I listened to THAT!
I took a measurement with REW before (bypassing everything, and no sub inputs) shown in red.
Then, plugged in the subs and made a new sweep with DRC Designer. Spit out all the filters and input them into JRiver Convolution plugin, and listened to all of them. They all show the same behavior, in different intensity.
I used REW to make another sweep to see what was going on. this time using JRiver and the convolution filters applied to the signal.
Pretty funny.... NOT!
This is the sweep I get, in blue.
Just imagine my face when I listened to THAT!
Attachments
Well at least now you know why you had such bad results... Now we need to find out why.
For consistency I'd recommend all tests without subs. Just a basic correction of the full range driver. Can we see an impulse of the sweep prior to convolution? For the left and right channel seperately?
For consistency I'd recommend all tests without subs. Just a basic correction of the full range driver. Can we see an impulse of the sweep prior to convolution? For the left and right channel seperately?
That blue curve is the ideal in room target response based on decades of psychoacoustic research into the meaninglessness of measurements
. But seriously, we're one step closer now and I think this highlights the importance of having visual feedback during each step of the process.
Perhaps I'm missing something here (I don't consider myself an expert on this topic) but if I'm correctly interpreting what is being proposed here, (please correct me if I'm not) then I'm not in agreement with the author.
It's true that measured in-room frequency response changes with fairly small movements of microphone position. I agree that these differences (4-5 db in the midrange with 1/6th octave smoothing in my room) are not really noticed. But, I think this phenomenon can be easily explained and shouldn't be used to make statements such as "there is no correlation between measurements and listening quality". With speaker/room setup like Linkwitz recommends (which is what much of my personal experience is based on), the brain can effectively separate the direct and reflected sound and, to a large extent, "ignore" the reflections. Since the frequency deviations that have been spoken of are due to reflections, we shouldn't expect them to color the perceived timbre very much and therefore shouldn't be overly concerned about them.
Now, here's where I'm really lost:
Is the author suggesting equalizing the system based on the in-room frequency response of a steady-state signal such as noise? If so, this is the only situation where I would think that averaging the responses from multiple mics (or moving the mic during the measurement) would be helpful. But why would anyone use such a technique when impulse response correction is within the reach of anyone with a computer and a microphone?
I have not tried the MMM technique but I have compared IR correction using a multi-position weighted average to results obtained with a measurement from a single position. The most accurate and musically pleasing result I have yet achieved has been IR correction with a small frequency dependent time window based on a single position measurement.
Like I said earlier, I have used the moving mic method in my car, using noise and a RTA measurement. In the car most of it is reflections and hardly any direct sound. Eventually I EQ-ed to songs, that I'm very familiar with,over a longer period of time. Gradually getting a curve that works for most music. I also played sine signals at the frequencies available on my graphical equaliser, centring each frequency that way with small cuts/boosts (e.g. left a boost, right a similar cut) leading to excellent imaging in the end.
I think for Bass the moving mic technique could be very useful. But we are talking about correcting the speaker as well as (or even more than) the room.
I think for Bass the moving mic technique could be very useful. But we are talking about correcting the speaker as well as (or even more than) the room.
The MMM seems to me like a low-tech way of windowing the measurement (by way of averaging out the effects of reflections) and more work too (although not as much as the weighted multi-position average method).
I think it could be useful for the car since we typically don't have convolution engines built into our auto sound systems and the thought of pre-processing an entire music collection only to then want to tweak a setting.....
I think it could be useful for the car since we typically don't have convolution engines built into our auto sound systems and the thought of pre-processing an entire music collection only to then want to tweak a setting.....
Last edited:
For Perceval, here's the line up of impulses from a measurement:
Play left and right speaker with a sweep in REW (length of sweep I chose 1M)
On the impulse tab in REW 2 distinct pulses can be seen
Moving the mic a small increment to the right gave me this:
Almost there, just a little more to go
Finally done (took me longer than usual, about 5 tries):
After that, I continue with normal left and right separate measurements, knowing the mic is in the sweet spot. After exporting the separate left and right impulse and convolution I measure left + right again, here's that impulse:
Here's the distortion graph from one of the 3 measurements made with left and right channel, with convolution in JRiver active:
Don't mind the big peak at 25, it's not really there, just ambient noise
Hope this helps to get you started. Pictures are usually easier to follow than words.
This time around I had my Microphone further away from the speakers right above my couch. The previous measurements were with the mic slightly
more forward. What I did learn was the more forward mic gave me a wider sweet spot. So I'll repeat the measurements with the mic in the old position.
Glad I got everything running on my new PC, working well so far.
Play left and right speaker with a sweep in REW (length of sweep I chose 1M)
On the impulse tab in REW 2 distinct pulses can be seen
Moving the mic a small increment to the right gave me this:
Almost there, just a little more to go
Finally done (took me longer than usual, about 5 tries):
After that, I continue with normal left and right separate measurements, knowing the mic is in the sweet spot. After exporting the separate left and right impulse and convolution I measure left + right again, here's that impulse:
Here's the distortion graph from one of the 3 measurements made with left and right channel, with convolution in JRiver active:
Don't mind the big peak at 25, it's not really there, just ambient noise
Hope this helps to get you started. Pictures are usually easier to follow than words.
This time around I had my Microphone further away from the speakers right above my couch. The previous measurements were with the mic slightly
more forward. What I did learn was the more forward mic gave me a wider sweet spot. So I'll repeat the measurements with the mic in the old position.
Glad I got everything running on my new PC, working well so far.
Last edited:
Glad your new PC is working as it should.
One thing bothers me with all this....
Why the best spot chosen for the measurements wouldn't be the place I normally sit to listen?
Wouldn't that be the sweet spot I am aiming for? To get the best sound possible at the spot I would be listening?
A long time ago, I got in touch with a tech from the best theater we had (the Imperial Theater in Montreal, the first one to get THX sound in the city), and he told me that he would always choose the same exact seat in the theater to balance the sound. I would always try to sit at that seat, or close to it anyway.
Shouldn't I try to get the best sound from my seating position in my room?
One thing bothers me with all this....
Why the best spot chosen for the measurements wouldn't be the place I normally sit to listen?
Wouldn't that be the sweet spot I am aiming for? To get the best sound possible at the spot I would be listening?
A long time ago, I got in touch with a tech from the best theater we had (the Imperial Theater in Montreal, the first one to get THX sound in the city), and he told me that he would always choose the same exact seat in the theater to balance the sound. I would always try to sit at that seat, or close to it anyway.
Shouldn't I try to get the best sound from my seating position in my room?
The mic should be placed exactly where your head will be. I take it this is not what you did?