To DSP or not to DSP

with my new speakers in place I’ve been doing some REW measurements and the results aren’t great. A -4dB drop around 1.5kHz, and steep roll off after 12kHz. And let’s not speak of the modes in the bottom end of the spectrum. This is not dissimilar to the measurements with my old ones, so it must be the room (or the Uniti :grimacing:).

Now the options for treatment are limited to non-existent, so I was thinking would adding something like a miniDSP 2x4 HD make sense? Would the benefit offset the addition of an ADC and DAC in the chain?

I know @Xanthe uses it for sub integration, but I don’t recall anyone mentioning it for general use.

The measurement with my previous set of B&W can be found here:

1 Like

Its possible that your speakers roll off faster off-axis. Luckily with HiFi this doesn’t have to be an issue, Wavelength from about 1.5k up is really rather narrow so its possible the position of the mic (or speakers) could be causing this. If you have the mic set up where you head would normally be then I’d suggest slightly repositioning your speakers. Also, it might be worth taking a measurement of one speaker at a time. This would allow you to overlay the plots and ascertain if you are getting any cancellation.

1 Like

That’s what the above graph is.

I can try repositioning a bit, but the old speakers weren’t in the exact same place and exhibited the same issues. Wouldn’t you agree that points to other causes? Room, Uniti or Mic (UMIK-1).

The room is for sure the culprit of the peaks and troughs up to 300ish, however I doubt its having an impact on the top end. If you can take a measurement of the speaker (or particularly the tweeter) on axis (at say 1m or slightly nearer to somewhat mitigate the room) then it would be useful to help identify if the environment/positioning is a factor on the top end.

Also, with the Umik - I understand they provide a calibration file. Has this been obtained and loaded into REW prior to measurement?

It is.

I’ll try your other suggestions in the coming days. Thanks.

1 Like

No worries. I’ve always though it would be a huge help if HiFi companies provided response plots with their speakers like the individual driver suppliers do. But I’ve yet to see it.

Just to add (and to respond to the original question), if you’re happy with positioning and measurement then DSP would be a good move BUT I would equalise the lower room modes first then add a low shelf below 1.5k as opposed to boosting above. I wouldn’t bother trying to eq above 12k or so either

DSP will not be as good as effective speaker and listener placement and acoustic treatment, however it is likely to be far better than nothing where there is no remaining possibility of physical changes but with significant peaks in response remaining, when I would expect the benefit to be greater than the disbenefit of the processing. Also my understanding is that where ADC and DAC are matched, it can be extremely hard for ears to detect that the signal has been digitised, even with relatively inexpensive converters.

But whatever you do, never try to boost out cancellations as that is a very rapid route to speaker destruction - and with boosting of even minor dips remember every 3dB boost doubles the amp power in use, and power fed to the speaker, and reduces headroom by a factor of 2.

I hope one day to try inserting a DSP processor in the digital chain before my DAC - and some people using Roon already do that to good effect using one of it’s sdd on capabilities (the one thing that might one day tempt me to try Roon again).


They used to decades ago!

1 Like

Ah, the golden age! We need to get back to those days imo

Left speaker only.

Red: Mic on tweeter axis, at 1m.
Blue: Listening position after slightly adjusting speaker position. Listening axis is mid, not tweeter, at approx. 3.5m.

Did you get anywhere with this? Interesting looking plots. Taking into account the difference in measurement distance (offsetting the difference in db) the tweeter appears to perform about the same in both positions. Also, it’s only down about 10db between 12k and 20k which is pretty common tbh. Does the manufacturer of the speakers specify upper frequency range and tolerance (±3, ±10db etc.)?

The plot from the listening position looks superb from 1.5k to 12k (±2.5db). The peak at 1k could be something to do with distance of speaker from rear wall or perhaps mic from wall behind you depending on distances (wavelength, 1/2, .1/4 wavelength etc.). Wavelength of 1k is 340mm. If either are relatively close to the wall you could try moving them further from the boundary or placing a big pillow/cushion behind and retaking the measurement to see if tames things slightly.

Not really as I broke my elbow shortly after this. I still plan to pick this back up.

Regarding the speakers, they’re a DIY build. Troels Gravesen 3WC-15. The previous speakers I mentioned were B&W 702s2.

Regarding the positioning, the speakers aren’t close to the wall, but the listening position is. For domestic reasons that won’t change either. I’ll try your cushion idea.

Firstly I hope your elbow heals quickly!

Do you mean your listening position is close to a wall in your post above?

Indeed. Corrected.

1 Like

Ouch, sorry to hear that. Hope it gets better soon! Those speakers look cool af! I love a good DIY build. It looks like Gravesen has supplied some measurements on his site which may be useful… Faital-3WC-15

They are not especially well labelled but we can see he is using 1/12 octave smoothing on the individual driver plot and 1/6 octave smoothing on the combined driver plots which you should be able to set this in REW in order to compare more closely.

Here on the plot of the drivers without crossover we can see that the mid and low drivers have the same response around 900hz-1khzish. If both drivers are in phase with each other at this frequency it could be causing the boost that you are seeing in your plot.

And indeed his combined response plot does show a slight bump roughly in that region…

The HF also appears to roll off almost identically to your plot.

I would say that you should first take measurements using 1/6 octave smoothing as Gravesen has done then decide whether you need to do anything further

1 Like