TOPIC:

Hmmm.... I suspect the above issue is due to a (Windows) laptop setting but I can't fix it. If I have the 'Default Format' set to 24 bit, 48 kHz then I get the results as I had above. If I try changing this to either 24 bit, 96 kHz or 24 bit, 192 kHz then it's as if REW can no longer 'see' the SHD as when I click on 'Check levels' it no longer detects any signal.

I'll leave this since I believe it's a computer rather than an SHD issue.

The ability to look directly at what Dirac is doing is very useful. I've been playing with a family of target curves based on the Harman curve - one of the early versions of Dirac had a target curve that was labelled as Harman, and @Ultrasonic posted a link (in another forum) to some more that applied 4dB, 6dB, 8dB, and 10dB bass lift from 400 Hz downwards. I grafted the two curves together so I had the behaviour of the original target curve from 400 and above and the 4, 6, 8, and 10 dB lifts from 400 Hz and below. The original curve gave about a 3dB lift to the bass and a small cut to the treble. Anyhow, I loaded 4 Dirac configurations into the DDRC-24's slots using these 4 composite curves, and the result is seen in the graph:



The nice flat horizontal line at around 102.5 dB is the FR with Dirac Live "off"; the other four curves show what happens when the four target curves are applied. A couple of interesting takeaways, for me at least:

- The absolute max boost that Dirac seems to apply is about 10dB which is consistent with what others have suggested elsewhere.
- The fact that there is a max boost level means that if you have a significant dip in the FR of the speaker, Dirac may decline to boost that dip to fully meet the target curve, because it would need more than 10dB boost to do so. This can be seen in the family of curves at around 55 Hz, where the boost given with the 8 and 10 dB boost targets (blue and green) is noticeably flattened, and below about 44 Hz where the only target curve that doesn't get some degree of flattening seems to be the 4dB curve (mauve).

This is consistent with my listening tests - the 4dB target curve gave results that were similar to the original Harman curve I had been using, and the 6dB curve gave noticeably more bass extension, but applying the 8 and 10 dB boosts didn't give the additional bass boost I had been expecting in practice, because the falling bass response of the actual speakers means that Dirac would need to apply more than 10dB of boost to actually achieve the desired target performance. I need to do some more playing with this, but the obvious solution would seem to be to apply PEQ filters at the bottom end that will make the "raw" FR curve flatter in the bass region (or even to apply some degree of bass lift there), which will then leave more scope for playing with the target curves. I have already applied some PEQ-based filters at the higher frequencies, as it seemed to be a good plan to get the overall FR curve as good as possible before applying Dirac; the findings above illustrate why this may be a good idea.

What fun we all seem to be having on New Year’s Day thanks to @Tw99!!

Happy New Year guys!

@TonyJ, for context are your example curves for Dirac run after some PEQ has been applied or without, and what if any smoothing has been applied? I realise there is a very large dynamic range on your graph that suppresses differences but it still looks somewhat surprisingly smooth below 40 Hz to me. Unless perhaps the room it was measured in was small enough that the lowest frequency room mode is at a higher frequency that this?

You'll know but for others I'll just mention that one risk of applying any PEQ boosts ahead of Dirac Live is that you may end up with a combined boost of greater that 10 dB at some points, which may not be a good idea depending on the amp/speaker headroom and distortion that may result, and due to the increased digital domain headroom that will need to be left to avoid signal clipping.

For context the relatively low peak boost of just over 2 dB for my own example was using the Dirac default target curve, and with my subwoofer gain set high enough to minimise boosts below 100 Hz where they would have been much higher if I was running my main speakers full range.

Tony_J wrote:

Dirac may decline to boost that dip to fully meet the target curve, because it would need more than 10dB boost to do so.

@Tony_J are you saying Dirac will do nothing with the dip or saying will try its best up to a limit of 10db boost?

asx77 wrote:

Tony_J wrote:

Dirac may decline to boost that dip to fully meet the target curve, because it would need more than 10dB boost to do so.

@Tony_J are you saying Dirac will do nothing with the dip or saying will try its best up to a limit of 10db boost?

The latter.

Ultrasonic wrote:

@TonyJ, for context are your example curves for Dirac run after some PEQ has been applied or without, and what if any smoothing has been applied? I realise there is a very large dynamic range on your graph that suppresses differences but it still looks somewhat surprisingly smooth below 40 Hz to me. Unless perhaps the room it was measured in was small enough that the lowest frequency room mode is at a higher frequency that this?

You'll know but for others I'll just mention that one risk of applying any PEQ boosts ahead of Dirac Live is that you may end up with a combined boost of greater that 10 dB at some points, which may not be a good idea depending on the amp/speaker headroom and distortion that may result, and due to the increased digital domain headroom that will need to be left to avoid signal clipping.

For context the relatively low peak boost of just over 2 dB for my own example was using the Dirac default target curve, and with my subwoofer gain set high enough to minimise boosts below 100 Hz where they would have been much higher if I was running my main speakers full range.

@Ultrasonic this isn't a room measurement - the input to REW was from the ddrc USB output. There is no PEQ in the FR shown on the screenshot.

No smoothing either.

Tony_J wrote:

@Ultrasonic this isn't a room measurement - the input to REW was from the ddrc USB output. There is no PEQ in the FR shown on the screenshot.

No smoothing either.

I understood it was USB based but presumably it is using Dirac Live filters that were designed based on microphone measurements? If not could you explain how this works please?

Ultrasonic wrote:

Tony_J wrote:

@Ultrasonic this isn't a room measurement - the input to REW was from the ddrc USB output. There is no PEQ in the FR shown on the screenshot.

No smoothing either.

I understood it was USB based but presumably it is using Dirac Live filters that were designed based on microphone measurements? If not could you explain how this works please?

I'm using the method described by TW99 here: www.minidsp.com/forum/dirac-series-suppo...droom?start=15#52800

Only difference is that I used a single tone sweep in REW rather than a pink noise sweep. So what you see in the graph I posted is a set of 5 curves that effectively show what the DDRC-24's filters are actually doing - the "flat" curve is with the Dirac filters turned off (there is no PEQ in the DDRC within the FR shown on the graph, although there are PEQ filters at higher frequencies); the other 4 curves show the effect of the Dirac filters when using the Harman target curves with 4, 6, 8, and 10 dB of bass lift. Yes, of course, the Dirac filters are based on room measurements, as is normal; however, I'm not attempting to show what the 4 target curves actually do to the in-room response, just what they do to the internal filters, so I can see how much "headroom" they consume. As said earlier, it looks like Dirac does indeed limit its max boost to 10dB (although depending on the measured curve and the target curve what it actually applies at any given frequency may be less than that).

Tony_J wrote:

I'm using the method described by TW99 here: img.discogs.com/AosbkePsrX_a0fO3ibt8V0a5...383242-8749.jpeg.jpg

@Tony_J You’ve posted the wrong link there.

Tony_J wrote:

I'm using the method described by TW99

Yes, I understood that . I used the same.

Tony_J wrote:

Yes, of course, the Dirac filters are based on room measurements, as is normal; however, I'm not attempting to show what the 4 target curves actually do to the in-room response, just what they do to the internal filters, so I can see how much "headroom" they consume. As said earlier, it looks like Dirac does indeed limit its max boost to 10dB (although depending on the measured curve and the target curve what it actually applies at any given frequency may be less than that).

OK, I had understood correctly then, so my points above about how surprisingly smooth your corrections are hold and I'm still a bit surprised by them. If we look below 100 Hz the examples posted show Dirac is making much smaller and fewer corrections for you compared to myself and @asx77 which is why I commented. Over that range you have about a 10 dB range in filter effect and this includes the effect of a 4 dB boost as target , with no apparent bass peaks to be reduced. By contrast @asx77 has a maximum attenuation of about 16 dB for the biggest peak and range of corrections covering about 22 dB. I have a maximum attenuation of 19 dB and a range of correction covering 21 dB.

@tw99's results are closer to yours in dynamic range but still show more features compared to your smooth response.

I'm now wondering if one important factor here beyond room size differences may also be how large the area over which the measurements were made is. My own example is for points on a sphere of just over 1 m diameter whereas I think I remember you commenting before that you use a much larger area?

Out of interest what does one of your full frequency range Dirac filters look like?

asx77 wrote:

Tony_J wrote:

I'm using the method described by TW99 here: img.discogs.com/AosbkePsrX_a0fO3ibt8V0a5...383242-8749.jpeg.jpg

@Tony_J You’ve posted the wrong link there.

Thanks - fixed now.

Ultrasonic wrote:

My own example is for points on a sphere of just over 1 m diameter

Mine too.

Ultrasonic wrote:

Tony_J wrote:

I'm using the method described by TW99

Yes, I understood that . I used the same.

Tony_J wrote:

Yes, of course, the Dirac filters are based on room measurements, as is normal; however, I'm not attempting to show what the 4 target curves actually do to the in-room response, just what they do to the internal filters, so I can see how much "headroom" they consume. As said earlier, it looks like Dirac does indeed limit its max boost to 10dB (although depending on the measured curve and the target curve what it actually applies at any given frequency may be less than that).

OK, I had understood correctly then, so my points above about how surprisingly smooth your corrections are hold and I'm still a bit surprised by them. If we look below 100 Hz the examples posted show Dirac is making much smaller and fewer corrections for you compared to myself and @asx77 which is why I commented. Over that range you have about a 10 dB range in filter effect and this includes the effect of a 4 dB boost as target , with no apparent bass peaks to be reduced. By contrast @asx77 has a maximum attenuation of about 16 dB for the biggest peak and range of corrections covering about 22 dB. I have a maximum attenuation of 19 dB and a range of correction covering 21 dB.

@tw99's results are closer to yours in dynamic range but still show more features compared to your smooth response.

I'm now wondering if one important factor here beyond room size differences may also be how large the area over which the measurements were made is. My own example is for points on a sphere of just over 1 m diameter whereas I think I remember you commenting before that you use a much larger area?

Out of interest what does one of your full frequency range Dirac filters look like?

I will have a play later on and post some full range plots, but at least one difference may be that I am applying Dirac over the full FR (~20 Hz to ~20KHz), and given that there is a limit to the number of filters Dirac can fit into the DDRC's memory, that would result in less fine detail than if you apply Dirac correction over a smaller range.

With this particular system I am working in a relatively small room and the measurement volume is similar top what you are describing - measurements approx 1 metre from the central (first) measurement, and using the "chair" configuration in Dirac, the listening position about 2 metres from the speakers. It is far from being a good listening setup as the room doubles as a study and workshop!

Tony_J wrote:

Only difference is that I used a single tone sweep in REW rather than a pink noise sweep.

I realised I'd made a mistake there. Mine was also a single tone sweep. I'm not even sure what a pink noise sweep could be...