Garden mix room near Nottingham, UK

#256

Great! Thanks for this, will give it a go. Had a very quick go with the midi file yesterday and was quite surprised how loud some notes were vs others.

As an aside, I’d been wondering if the 200Hz bump was created by my desk, so will at some point move that behind the mic and take the measurement again.

Having sold some gear I’m thinking of making a smaller desk anyway…

#257

you could also set the midi note velocity to say 100 for all notes and see the differences as well

#258

Morning
I took a few measurements yesterday. The first was to see whether the desk was contributing to or causing the 200Hz ish bump, so I redid the measurement then moved the desk to back of the room (well, in front of the sofa) so it wasn't between the mic and speakers.

The results suggest the desk isn't causing that lump, which on the one hand is a good thing:

Having put the desk back, I then moved the desk and mic quite far forward towards the speakers, and started working backwards, moving the desk and mic around 1" at a time. As the purpose was not to find the best position for the desk I didn't measure this perfectly.

I did 21 measurements which covered about 2'. For reference, all my previous readings during the build were taken from "-13 original mic pos".

I then put the desk and mic back to the "-10" position and did a few side to side movements - each movement was about 3". Here are those results:

So finally, I created average readings (thanks Endorka) and they are below:

And finally, a 1/12 smoothing version of the above:

I'm going to try and use the EQ to drop the wide peak that has a centre frequency of around 220Hz, and also to reduce the peak centred on 34Hz.

I will re-do the listening exercise with the MIDI notes and report back on that, to see how it compares to the measurements.

As the above graphs aren't very clear, here are the REW files.

#259

The REW EQ settings pane is really interesting.. sat on the sofa rather than in the studio I've played with filter settings compared with the front/back average measurement(s). I started with the room modes as identified by the Bob Golds calculator:

31.3
62.4
93.5
111
152.5
155.5
164.1
178.5
187
215
272
313

Setting a target line on the graph at an 'average' point, I followed Gulfo's advice of several filters doing 2-3dB, so here's what I've got. I haven't touched the Q setting, not sure whether I should be - figure as these are filters based on room modes they need to be quite narrow filters?

The predicted response is quite good around the 200Hz peak, but there are a lot of filter moves so not sure if there'll be phase problems?

Finally, there is only one room mode at 31.3Hz, yet it looks a lot broader than that in the measured room. It'll be interesting to see how effective a narrow filter at 31.3Hz is at reducing that broader peak. The predicted response above suggests that it won't be very effective!

So, looking at the waterfall plot from my new mix position, there are ringing resonances at: 23.9, 27.4, 31.6, 36.4, 39.9, 44.5, 47.2 - the ones in bold are half (or near to) the frequencies identified by the Bob Golds calculator.

So, I tried using a few more filters on some of those, and I broadened the bandwidth a touch on three of them:

#260

And in reality, the filter predictions are pretty good at the mic position. I tweaked them to make them a bit broader and here are the current EQ settings (after a bit of tweaking):

and impact on the readings at the position where my head actually is:

My previous sofa readings had been in front of the sofa, so I put the mic where my head would be and here's what we end up with (pic to show mic position for reference):

Any thoughts on those spiky nulls at 246 and 322?!

#261

It's like an addiction this is! I ended up keeping on going and did some work on the top end - a dip around 3K and a boost (my first one) around 6K.

As others have said, the closer to flat the harsher it sounds, so I opened some filter settings Stuart had done for me in my previous studio and figured out which 3 bands were doing the B&K house curve HF roll off.

I applied those, but found it a bit too dead, so I came back 1.5dB from there.

I'm sure there's a lot more that could be done but I don't have the expertise and something tells me to avoid boosting (even though looking at my previous settings Stuart had done). I've also done my settings the same for L&R.

I'm going to leave these settings on and listen to / work on some stuff and see how it feels.

Any thoughts or corrections gratefully received!

Final settings:

Comparison between EQ on and off:

Finally, with psychoacoustic smoothing on (looks way better!!):

#262

I agree, this sort of thing does get very addictive! I've never done the electronic room correction properly, so please take my comments with a pinch of salt.

If you have time and inclination I'd be interested in seeing the results of measuring you mix position with the correction filters in use, but with the average of several positions around it.

For a small room Rod Gervais recommended measuring over a cubic area of 2x2'. Measurements ~1' apart. By my quick ready reckoning that's 27 measurements! A lot of work to do this with precision. When it's time to do this I reckon I'll build a jig. Thinking of a base built from some plywood or OSB with 9 holes in a square, each 1' apart. The holes are far slotting the pole of a mic stand in to. The mic stand is a telescopic type with 3 marks on the moving section 1' apart. This should allow the 27 measurements to be done quickly and accurately.

Marks on the floor e.g. masking tape to allow precise overall positions measurement to be repeated from before / after tests.

I know this seems like a lot of work, but if I've spent years doing physical labour of making the room acoustics as good as possible, I reckon it'll be worth it to get the last part as accurate as possible

Here's a graph of several "ideal" room response curves I found. The B & K curve is there.

Looking at it for the B & K I noted some frequencies and the approximate SPL reduction at those points for fitting a curve to.

200 -0.5 dB
2000 -3 dB
20000 -6 dB

Cheers!
Jennifer

#263

Thanks Jennifer. Your jig idea is a good one and easy enough to do. I’ll pull out Rod’s book and see if I can find that section.

My other thought is to try one of the auto correction systems and then once that’s done it’s work replicate the settings on my FBQ - i don’t want to be relying on a plugin running and having to remember to turn it on/off, plus it wouldn’t suit my workflow.

#264

garethmetcalf wrote:Source of the post Thanks Jennifer. Your jig idea is a good one and easy enough to do. I’ll pull out Rod’s book and see if I can find that section.

My other thought is to try one of the auto correction systems and then once that’s done it’s work replicate the settings on my FBQ - i don’t want to be relying on a plugin running and having to remember to turn it on/off, plus it wouldn’t suit my workflow.

I totally understand that, and feel the same way. I currently have the B&K curve done by an EQ plugin on the DAW's monitor FX chain. It's a PITA because I have to remember to bypass it when switching to headphones. Also instruments being recorded and monitored in real time don't go into the monitor FX chain so the B & K curve isn't being applied to them. Having the FBQ between the monitor controller and speakers solves all this.

As I said, I really have no expertise in this electronic room correction, but the more I think about, the more using an average of measurements taken over a certain amount of space makes total sense. For sure it's overkill at certain stages of design and build, but for this kind of finalising it feels apt.

There must be some statistical analysis one can perform that would demonstrate the likely accuracy and error bounds of a single point measurement vs. multiple measurements over a space. So that you could say for example "if your head is within this 2' cube, you can be 95% confident the frequency response will be within however many dB of the ideal house curve".

Rod mentions the part of this dependent on frequency. But unfortunately a fuller type of this analysis is beyond my skills or time to investigate at the moment.

If this is the case, then presumably it would also be necessary to perform something similar for those commercial room correction plugins? Otherwise one might just be compounding errors.

Cheers.
Jennifer

#265

Seemingly the plugin systems such as sonarworks require multiple measurements to be taken. Some also change phase as well as using eq filters to correct (although that would introduce a time delay) and so not all of their functions could be replicated with the FBQ.

I think taking a multiple of measurements with and without my eq settings will be a good idea, with a jig as you suggest.

Got some other stuff to work on but I’ll return to this. Listening yesterday I preferred the sound at my mix position with the FBQ in, but need to experiment walking round the room.

#266

garethmetcalf wrote:Source of the post Seemingly the plugin systems such as sonarworks require multiple measurements to be taken. Some also change phase as well as using eq filters to correct (although that would introduce a time delay)

That's good to know. If the delay was too much it would make recording in the control room impossible with the correction engaged.

Cheers!
Jennifer

#267

yes, most room equalisation products want multiple (some as many as 50 or more) measurements in an attempt to use digital timing and eq'ing to "fix" the room. what you would look for - mix position first and best. secondly the couch (reasonable).

Soundman2020 - Studio Design Forum