Where comes ther difference from between my calculations and REW

[Frank_F]

Inspired by REW I started working on a little DSP programming myself. I was reading lots of theory and actually I am pretty happy so far with first results of my little app. Overall the IRs done by the app seem to look pretty close to the ones done with REW, with few differencies which I dont understand so far.
First of all my IRs so far look much more noisy without filtering then the REW ones. Only when I do 1/48 smoothing in REW I get kind of same smooth results than REW delivers without?? smoothing?
Second I have differencies at low frequencies and strangly arround 7.2kHz, where REW seems to have a much more significant drop.
I tried few things that changed the overall picture little but not completely. When I told my app to calculate the IR with 60Hz/12db lowcut, I get closer to the REW result in the bass area, but still looks like there is kind of little sine wave in that part. trying different Tukey parametrs in my app did not bring me closer, too. I did all measurments with same mic position, same preamp, same Amp. All corrections (soundcard, mic) in REW where switched to "none".
Could you please have a look and give me a hint what is going wrong?/different? here.

 

[John Mulcahy]

S/N ratio of that impulse response is very poor, but there isn't really a way to say why it isn't what you expect.

 

[Frank_F]

ok, after few headaches I found my mistake. the mistake was in my algorithm. The strange sine came from a stupid wrong cutting, was no noise.

But now I face another (for me) strange phenomena. When I just process the data FFT and extract the IR the frequency response looks pretty good from 150 to 6,5 kHz. When I shot same IR with REW the IRs a pretty close to the IRs I get from my code. This already made me pretty happy. What I dont understand is the behaviour below 150 and above 6-7 kHz. In REW the frequency resonse below 150 kHz and above 6500Hz is much lower (which is much more realistic for a guitar cab). I tried few things. When I started putting a low cut (60Hz, Q1, 6dB) all changed. Now I of course get a bass reduction and the curve is close to REW. This is what I was expecting. Strangly the behaviour above 6,5kHz changed too, so now there is that drop in frequency response too. The IR now looks actually pretty close to the REW one. But why? There is nothing done/changed then putting a low cut. Why this changes the high frequency response that much? Could this be a scaling problem, somehow? I am confused. Or does REW use an intrinsic low cut filter on automatic measurements too?

 

[John Mulcahy]

There is a strange feature in the REW IR from -1.1 ms to -0.2 ms, that is the source of increasing energy above 6.5 kHz. Not normally something one would see.



Your IRs are very short (too short) and truncated very close to the peak.

does REW use an intrinsic low cut filter on automatic measurements too?

No.

 

[Frank_F]

Hi John,

than you very much for your fast reply. The IR is used as guitar cab IR in a guitar modelling unit. Due to limited DSP power in real time and for the sake of acceptable latency those files are pretty short, 140ms is already a pretty long IR, most units use half (or even shorter).
That strange feature between-1.1 to -0.2 you found in the IR I shot with REW or in the IR I did with my app and imported to REW? Do you have any idea what this could be? Can this come from noise (clicking), sometimes I have the feeling that the chirp has some noise within (slight clicking in the beginning and end)

 

[John Mulcahy]

The IR is used as guitar cab IR in a guitar modelling unit. Due to limited DSP power in real time and for the sake of acceptable latency those files are pretty short, 140ms is already a pretty long IR, most units use half (or even shorter).

OK, but you should window the IR to the required length after you have finished all the processing to extract it, don't truncate anything before that. You should be starting out with your sweep padded to twice its length with zeroes and the response captured to twice the sweep length.

That strange feature between-1.1 to -0.2 you found in the IR I shot with REW or in the IR I did with my app and imported to REW?

With REW

Do you have any idea what this could be? Can this come from noise (clicking), sometimes I have the feeling that the chirp has some noise within (slight clicking in the beginning and end)

Clicks or pops could cause that, if there are any you should try and find the cause and eliminate it. Make sure the ASIO buffer setting is long enough, 1k samples or more.