Confused

[Omid]

I was fiddling around with some measurements that I needed to do, and found it'd be useful to have a perfectly flat Audiolense measurement file (an ideal speaker). So I routed (internally in my ADC/DAC) the output signal of all speakers to the return ADC channel (the input channel in the measurement window in AL). The idea was to feed an Audiolense sweep directly back to the input channel, creating a perfectly flat frequency response, perfect step response etc. To my surprise, this is the frequency response I got:

The purple is the sub, the green the main speaker. Instead of a flat line, there is a ~20 dB boost in low frequencies. Above 300Hz the graph is totally flat. The same happens with a 2.0 setup (no Sub).
Is there something wrong in my setup or am I missing something?

PS I have made sure there was no weird eq in my DAC.

 

[juicehifi]

I don’t know what’s causing this, but normally you should get the result you expected to get.

 

[Omid]

Thank you for the response. I will continue to do some troubleshooting to figure out what’s wrong in my setup.

 

[jjazdk]

Your result looks like a target curve

 

[Omid]

Your result looks like a target curve

I know it looks like it, but it's not. I fixed it so it's flat now. It is a measurement. Here it is (in a simple 2 channel version) in case someone needs it at some point:

 

[dathzo]

what was the issue you were having before? Curious to know...

 

[juicehifi]

This looks all good. The group delay oscillation happens while the sweep is fading in. And if you zoom in on the frequency response you will see traces of it there too. But nothing to worry about here.

 

[Omid]

This looks all good. The group delay oscillation happens while the sweep is fading in. And if you zoom in on the frequency response you will see traces of it there too. But nothing to worry about here.

Thank you for confirming. I’m happy with it.

 

[Omid]

what was the issue you were having before? Curious to know...

I use a Motu dac. There was a shelf eq in a channel that wasn’t supposed to have one. My routing table and EQ table are a bit out of control (too big) !

The reason I had a shelf EQ in another channel is that my DAC is used as my phono preamp as well. I mix the right and left channels below 150 Hz and keep them stereo above 150 Hz (thus the shelf filter). This creates a vertical rumble filter. Then, in AL, using the target, I implement a sharp 20 Hz filter for very low frequencies (which avoids the phase shift problem that eRIAA has), a notch filter at 60 Hz for any 60 Hz bleed-through from the powerline, and a RIAA correction. So in essence, I 've recreated a digital version of a phono preamp and KUB rumble filter (https://www.analogueseduction.net/phono-stages-phono-boards/KABRF1.html).
The funny part is that I seldom listen to my LPs ! But when I do (and the disc is a new disc), it sounds great. Just liked the challenge of having everything setup perfectly.

The flat measurement was needed to measure the THD+N of my virtual phono preamp. Unfortunately I realized that to do this properly, I could not use my DAC as the source for a 1kHz signal. I would have to located a 1 kHz generator that has lower distortion than the device under test. In the meantime, I figured I'd measure the real life THD+N by using a test LP. The measured noise floor was so high that I realized instrument grade measurements were irrelevant.

The flat measurement file will also serve to measure the 'filter insertion loss'. I realized that talking about loss is not very meaningful when the user has a target that's down-sloping and de facto creating a forced loss. So a speaker/room correction should be compared to a correction applied to this ideal speaker measurement with the same target. From what I can see though there is only a loss of ~1dB between the tallest correction peak and the 0dB line. The correction peaks are dependent on your speaker/room and target curve. I was under the impression that a mismatch in volume between sub and speaker (or woofer and tweeter) would increase the loss, but it does not (the louder speaker just gets attenuated to a greater degree). It still makes sense to have roughly the same volume so your microphone measurement stays in the green zone for both speakers.

I realize this is all kind of esoteric, and useless to most people...

 

[dathzo]

I use a Motu dac. There was a shelf eq in a channel that wasn’t supposed to have one. My routing table and EQ table are a bit out of control (too big) !

The reason I had a shelf EQ in another channel is that my DAC is used as my phono preamp as well. I mix the right and left channels below 150 Hz and keep them stereo above 150 Hz (thus the shelf filter). This creates a vertical rumble filter. Then, in AL, using the target, I implement a sharp 20 Hz filter for very low frequencies (which avoids the phase shift problem that eRIAA has), a notch filter at 60 Hz for any 60 Hz bleed-through from the powerline, and a RIAA correction. So in essence, I 've recreated a digital version of a phono preamp and KUB rumble filter (https://www.analogueseduction.net/phono-stages-phono-boards/KABRF1.html).
The funny part is that I seldom listen to my LPs ! But when I do (and the disc is a new disc), it sounds great. Just liked the challenge of having everything setup perfectly.

The flat measurement was needed to measure the THD+N of my virtual phono preamp. Unfortunately I realized that to do this properly, I could not use my DAC as the source for a 1kHz signal. I would have to located a 1 kHz generator that has lower distortion than the device under test. In the meantime, I figured I'd measure the real life THD+N by using a test LP. The measured noise floor was so high that I realized instrument grade measurements were irrelevant.

The flat measurement file will also serve to measure the 'filter insertion loss'. I realized that talking about loss is not very meaningful when the user has a target that's down-sloping and de facto creating a forced loss. So a speaker/room correction should be compared to a correction applied to this ideal speaker measurement with the same target. From what I can see though there is only a loss of ~1dB between the tallest correction peak and the 0dB line. The correction peaks are dependent on your speaker/room and target curve. I was under the impression that a mismatch in volume between sub and speaker (or woofer and tweeter) would increase the loss, but it does not (the louder speaker just gets attenuated to a greater degree). It still makes sense to have roughly the same volume so your microphone measurement stays in the green zone for both speakers.

I realize this is all kind of esoteric, and useless to most people...

Thanks for the detailed explanation! And very educational as well... First time I hear about rumble filters