I recently bought a KEF KC62 sub and measurement mic and would like to integrate the sub with my Magico S1 two-way speakers. I’m using a Devialet Expert 200 so I can individually set the low and high pass filters and add delays to the speakers.
At a high-level is this the best method to use REW:
Determine sub location: I took measurements using a 10-200Hz sweep with no low pass filter on the sub from the three “acceptable” locations in the room. I picked the location with the flattest response from 10-100Hz. I tried the “subwoofer crawl” but that didn’t seem too helpful.
Determine the crossover point: is the best way to do this to set the crossover to different points and compare measurements? I can do overlapping points in the Devialet and Magico recommends measuring the sub and speakers separately and overlapping the crossovers to sum dips. Would this be a better method?
At this point would I then follow the guide?
a) Would I want to set the EQ target type to Full Range?
b) Any reason not to use a house curve here? Or is it just personal preference?
c) Would I want to set Match Range to say 15-15,000Hz?
After I’ve created the filters in RePhase would I still need to time align the sub and the mains? What is the best way to go about that?
This thread and documentation have been super helpful. I’m looking forward to getting this setup and will post my progress here. Thanks!
1.I would keep the sub in the middle of the speakers unless you intend to buy a second sub as from the FR curve of your speakers ( I assume they’re MKIIs) the sub crossover frequency needs to be set at its upper limit (140 Hz):
(anything above 80Hz is locatable by the ear)
2. Already answered in 1.
3.a)I think this system with the gap closed below 140 Hz by the subwoofer which can go as low as 11 Hz deserves full range more than anything else.
b)house curve is about lifting the lower bass volume level as our brains hear these frequencies less loud than they really are (ie even when it’s flat in response, it’s not flat to our ears) but the SAM system in your Devialet kind of already does that. Check:
c) 7000 Hz wavelength is less than 5 cm, that means a little movement of your head in the listening position will change the response significantly so I don’t like EQ’ing high frequencies and ruin the tonal balance (and phase) of the speakers but you can try and test what you like. Personally I EQ the lower bass (up to 250) only, phase EQ the crossover frequencies and any EQ I have done in lower bass. Once the bass is sorted the higher frequencies comes to life. But your equipment is a bit more sophisticated.
4.Time align sub and speakers during the first measurements with REW and then remeasure and then start FQ and phase EQ. The delay in ms in every measurement is your key to adjust the distance equality. Then you should also check in rephase for the phase equality of the sub and the Magicos’ woofers. Remember that Devialet also does some phase alignments which I have not much info on.
Hi @OCA , so after a lot of speaker placement experimentation, this is where I have landed with my KEF R7s. Every time I EQ (I I EQ’d for the full spectrum)…I would feel it takes the emotion out of the music so I am thinking, Is there a way to just correct the phase without touching the Frequency response of my speakers. The dips you see seem to move in line with how the phase shifts are happening. I am not that technically educated but it seems if I can do the phase correction, these dips may self correct.
So here is the question, can I do ONLY the phase correction in RePhase using the original frequency response without importing the EQ file from REW and perhaps generate an impulse .wav file as per the guide here? Appreciate the process to do just the phase corrections only.
Below are the original measured responses for my speakers with a SVS SB3000 sub set at Cross over of 58Hz - all running together. Thanks
You can FILE/EXPORT/EXPORT MEASUREMENT AS TEXT from REW and MEASUREMENT/IMPORT FROM FILE from RePhase the saved .txt file. In RePhase you can definitely benefit from fixing the crossover phase shifts of your KEFs with “Filters linearization” tools (Your speakers have 2 crossover points at 400 Hz and 2900 Hz)
Advanced is right, I have been battling for a week now to figure out the basics and have finally figured out a bit. Now I have clipping in my signal after exporting into convolution. The eq settings I chose we only cuts and no boosts. More power to you all, but for me. I think I am just going to stick with mechanical changes and forget the DSP all together. I think it’s just beyond me.
Im afraid if you’ve spent a week on DSP, you’re already into the rabbit hole😂
Getting clipping is quite strange. Even if you had an overall boosted eq rePhase would warn you while generating the convolution file and actually even cut it down by some amount and report that to you with a pop up.
Still there are many ways, even directly in Roon to avoid clipping. May be share some details and screen shots and we can check.
Turn on the clipping indicator in Roon’s headroom management then gradually enter ever more negative values until no music piece triggers it any more.
Have you generated filters for all individual sample rates you ever use and zipped them into your convolution file?
If not, and just using one fixed sample rate file, Roon automatically upsamples it to match the source file rate, but that could cause inter-sample-overs, which is clipping.
The same can happen if you upsample via Roon’s DSP.
You had helped so much,
I thought replying on a post that was more focused on clipping would be better. Also, I didn’t want to bother you more.
I had turned on the clipping indicator and went down to -10 and was still getting heavy clipping. The measurements you see are from following the moving mic method in the Stickied Roon Setup Post. I moved my listening position forward and backward and took measurements. This was unfortunately the best measurement I got in a usable listening position More nearfield was better but my place feels cramped with the speakers so close.
I saved the full set, zipped them, and loaded the Zip file into Convolution in Roon. I do not believe I am using the sample rate conversion in Roon. Below is my set up in the DSP section. So unless the conversion happens by default, I think it is off.
I found the Headroom level being at -14 eliminated the clipping. At -13 I found a few tracks that occasionally blipped the clipping light and went one further. 30 to about 1k sounds MUCH better with the filter set engaged. To me, it sounds like I lost extension in all the tones above 1k.
This is causing me to question… By adding negative head room, I would think I am attenuating the entire signal, even the frequencies that have no adjustment on the filters. Is this correct or am I just psychosomatically perceiving this change? If it is the case, This would be a’okay for B&W’s or Focals, but the R700’s have such a nice smooth top end (to me) I would hate to alter the high’s I hear when not using the filter set in Convo
Is there a way or would it be a better practice to set the filters based on a lower db setting for the house curve?
You are saying you are not doing any boosting EQ so it might be that you are setting your target db level too high during EQ. Try to stay 2-3 db around REW’s calculated target level.
As per your high frequency problem:
Ideally you should avoid correcting anything above your room’s transition frequency (that’s around 200 Hz). This is where you are listening to your room sound not your speakers hence should be corrected. In that band, you must still avoid “non-minimum phase” regions. Even after that you should avoid using narrow EQ filters (high Q, high gain) as they will cause pre-ringing. Up to 4 times this transition frequency, there are still room effects in the sound so EQ can be useful as long as you correct only minimum phase frequencies. So by around 600-1000 Hz you should be done.
There usually are a couple of frequencies in the high band available for EQ’ing on the condition that you see the correct response there in the first place: Make sure you FDW (like 15 cycles) your measurements. Otherwise with the standard 125 ms left, 500 ms right windowing REW uses you will see much higher SPL in high frequencies and try to lower them with EQ resulting in lost tonality even if you manage to correct the phase shifts caused by the filters (easier said than done) and then apply psychoacoustic smoothing (this is the closest you will get to the real response in high frequencies also guarantees low Qs in the filters, don’t bother with any other).
Check that Excess Group Delay graph (my actual left speaker):
Only the flat areas are minimum phase thus can be corrected with EQ. So for example you don’t touch between 45 and 55 Hz, around 67 Hz but 30 Hz seems very suitable for EQ. You can see the spikes getting much more crowded after 850 Hz (that’s because my room’s transition frequency is 215 Hz - you can find that easily by comparing measurements of the same speaker at different mic locations - it’s where the deviations in response begin) and some careful filtering can be done in that area. Clearly after 4kHz there’s nothing but spikes, so that’s no go area. However around 17.5kHZ there’s a flat region so you can add a one-off filter there for instance.
I know this all kinda deviates from my guide a little but it’s now a bit outdated and I am planning to update it soon (it will probably be a video though as it’s too much work to write a guide with all the screenshots, etc.).
Yup, that was totally it. I also found the headroom adjustment in REW>EQ. I may play with that a bit. But the big thing was I was trying still too high of a target level. I figured I was safe as it was still a good bit below the auto-target REW chose. I changed the house curve down to an adjusted Harmon curve down to 4db instead of the 8 or 10 that I had and then lowered the target even a bit from there as I felt it still sounded a bit bass heavy for me. New eq is much smoother and no clipping at a 0 setting on the Roon headroom adjustment. I will leave the clipping monitor on for now to just make sure there isnt anything being detected by Roon or by my ears.
Thank you for this. I will set up a new curve to stop any adjustments above there. I was thinking about this hoping I could get away with adjusting only the bare minimum. My speaks sound really good to me in my room and I didn’t much care for the EQ above there anyway. I appreciate the validation.
To both you @OCA and @Marin_Weigel, I really appreciate the help. Just the nuggets I needed. The first time I heard that clipping and could not make sense of itin my mind I just freaked. I know logically it had to be me doing something wrong but I could not get my head wrapped around it.
Say I find my happy place and wanted to expand outside of Roon for the eq but didn’t want to change the rest of my rig (I dearly love my Freya+, PontusII, and Darlington MP-7. Could I do something like the MINI DSP SHD and not degrade the other components? The idea would be to dip into Dirac as well as provide the DSP benefits without hurting the audio quality of my other components. I realize this is a dangerous question as many people that would suggest the components I have probably think DSP is the devil.
Very Interesting, I was wondering what the psychoacoustic smoothing really did. What you said makes a lot of sense. Many more filter sets to test. I really like the way Convolution works, I can build a few and see how they affect the sound. I am really liking the current sound even if the mids are a bit recessed.
Hiii,
My system is: HP Workstation (i9+32gb+Nvidia RTX3070, W11… ) + HqPlayer (Crossover, filter, Eq from Rew and RePhase) + Roon → SOTM Ultra Neo + OKTO PRO8 + 4 Benchmark ABH2… 3 *2 channels + 1 sub.
To make the measurements and adjustments in Rew I have to introduce the signal to each one of the channels measuring independently, adjust levels, etc…
I would like to know if there is an easy way to inject the Rew “sweeps” directly into HqPlayer (NAA?, Content source URI?, Network Audio Demon?) in a way that makes it easier to obtain the frequency and phase measurements.
The Entrypoints solution is of low quality and generates a lot of delay.
There’s the option to export the measurement signal, including a timing impulse, from the signal generator setup dialogue.
Just import that into Roon, and Bada-Bing Bada-Boom, there you go…
Their measurement tool, Impala, is freeware so I checked it. Frankly, I didn’t think it’s much easier than REW. You do not have the option to select the frequency range you want to measure, it always measures for the whole spectrum and all other settings you need to go through are quite similar. You can also take limitless number of multiple measurements in REW. Impala can take up to 10 different positions. But from what I understood from their manual, the real exciting thing about Focus Fidelity is that measurements taken at multiple positions are used to build a time-domain accurate transfer function rather than a spatial average unlike Dirac and Audyssey. This is the only correct way to use multiple measurement points so they know what they are doing IMO. However as the actual program (FF) creating the transfer function is $250 and no free trial or refund policy, I could not see or test how it works.
It could be beneficial if we could import that final average response created with a mathematical model to REW and then continue with the inversion technique. As per their correction method, there’s not enough information in the manual to decide how well it works (inversion is mentioned though). One can only ask previous buyers.
Hi Serkan,
Impala is designed to take measurements of left and right channels at up to 10 positions for import into the filter designer. If you want to use Impala for other purposes, you can by exporting the impulse responses as wav files. However, the purpose of Impala is measurements for the filter designer application. Taking these ten measurements should be substantially easier and faster than using REW. At least, that is the intent and also the general feedback I have received. Options like measurement frequency range are specifically not selectable in Impala. If you still wish to use REW with the filter designer, you can by following the guide here https://www.focusfidelity.com/REWInstructions.pdf
As you’ve found, there is no demo version, a demo version to allow proper evaluation would, I believe, need to generate the same filters as *.wav files, which the paid version generates. There is no way to restrict the use of these filters. They could, for example, be used indefinitely. Demo versions are also prone to being “cracked”, etc. So, for now, there is no demo on offer.
Here are a couple of examples of user feedback regarding the filter designer,
Plus some feedback on the Focus Fidelity website,
Anyone who does purchase the software can get support via the support e-mail address, and this brings customers to a position where they are happy with the results. Usually, support is not required.
