Dirac Live generated filter + 7 dBFS (for Bose 901 VI w/o 901 EQ)

[skyfly]

In a friend's living room (apatment, flat, I am not sure about the most correct word), I tried the room correction with miniDSP Flex DL (Dirac Live option included) for Bose 901 Series VI (version 1) without the 901 Series IV EQ. So Dirac Live did both room correction and Bose 901 EQ.

The finished Dirac Live filter playing -0.5 dBFS sine sweep: about +7 dBFS reached a couple of times on the output RMS meter on miniDSP Device Console

The suggestion from miniDSP retailer (not minidsp.com): It does not seem that Dirac Live has nearly enough dynamic range to do that. We can offer free upgrade: their digital Bose 901 EQ.

The corrected frequency range was 30Hz - 16.8kHz. Near the ends of the range, the target curve rapidly dropped to near uncorrected level.

Note to prevent misunderstanding about "-0.5 dBFS sine sweep" I mantioned above): This post is not about clipping during the measurements for Dirac Filter generation. During the Dirac Live measurement, no signal generator was set up near 0dBFS.

Dirac Live measurement was done with PEQ and crossover off, mixer level and gain level at zero (zero by default after installing the software) on miniDSP Device Console. I finished filter design, and filter loading to miniDSP Felx DL.

I asumed that Dirac Live software automatically adjust overall level of the filter so that no clipping happens (e.g. if there is a lot of boost at some frequencies, the overall level is reduced).

Thereafter, I tried to check whether the Dirac-Live-generated filter will occasionally clip audio files (these audio sources are given to me by record labels, movie studios, or broadcast company, and I have no control over the audio signal levels in those files or steaming). As a quick extreme test, I played REW -0.5dBFS sign sweep (It replaces playing multiple music tracks while observing occurrence of clipping. It was not to measure in-room frequency response again.).

The level display(output RMS meter) on miniDSP Device Console, of course, fluctuated. To my surprise, the level reached about +7 dBFS a couple of times.

 

[alanca3]

I’m confused, are you asking a question or sharing the fact that Dirac sent a signal at +7dBFS?

In any case, Dirac is a room correction software. The 901 functions like an active speaker in that the crossover and filtering takes place before the amp. Using Dirac without the required 901 EQ would be similar to using Dirac on a speaker without its crossover. It will apply filters, but that’s not what it’s designed for.

You should insert the 901 EQ downstream of Dirac so that the algorithm can focus on correcting your room and not your speakers.


Correction filters typically add positive gain (boost frequencies), and if you leave the processor at unity gain it is liable to output signals above full scale. A signal over full scale will be limited at some point before it reaches your speakers.

 

[skyfly]

I’m confused, are you asking a question or sharing the fact that Dirac sent a signal at +7dBFS?

In any case, Dirac is a room correction software. The 901 functions like an active speaker in that the crossover and filtering takes place before the amp. Using Dirac without the required 901 EQ would be similar to using Dirac on a speaker without its crossover. It will apply filters, but that’s not what it’s designed for.

You should insert the 901 EQ downstream of Dirac so that the algorithm can focus on correcting your room and not your speakers.


Correction filters typically add positive gain (boost frequencies), and if you leave the processor at unity gain it is liable to output signals above full scale. A signal over full scale will be limited at some point before it reaches your speakers.

In short, I did the Dirac Live measurement again with (approximate) Bose 901 EQ, but had clipping from Dirac Live processor again: + 5 or 6 dBFS on -0.5 dB sign sweep.

Instead of a Bose 901 Equalizer unit, I used the PEQ on MiniDSP Device Console. The PEQ processing is separate from Dirac Live processing.

EQ1: 21, -15, 1.9, Low Shelf
EQ2: 35, 5.9, 0.5, Peak
EQ4: 232, -6.2, 0.5, Peak
EQ5: 696, -6, 1, Peak
EQ8: 8800, 7.9, 0.5, Peak
EQ9: 16900, 7.8, 0.5, Peak
EQ10: 20000, -5, 0.7, High Shelf


During Dirac Live measurement: the PEQ on. Filter design, loading completed.

Now checking for clipping. Play -0.5 dBFS sine sweep from REW. On the RMS meter on the MiniDSP Device Console, +5 or 6 dBFS observed.

Input -> Dirac Live processor -> PEQ (on during Dirac Live measurement, off during clip checking by -0.5dBFS sine sweep) -> digital output

I want to reduce the level before the Dirac Live processor stage. The retailer or the manufacturer did not yet answer where the "Mixer" level control is located in the signal chain.

 

[DVDdoug]

I don't have DIRAC, but some general notes -

As you may know, DACs are limited to 0dBFS and will hard-clip if you "try" to go over. CDs and regular (integer) WAV files are also limited to 0dB. Some formats can go over and most DSP is done in floating-point which for all practical purposes has no upper (or lower) limits. But if it goes over 0dB it should be attenuated before it hits the DAC.

CDs and MP3s are usually "maxed out" with 0dB peaks. MP3s often hit +1 or +2dB which will clip your DAC if you play at "full digital volume". The movie formats often have some headroom. So you should assume that there is no digital headroom or what you're calling "dynamic range". (There is a LOT of dynamic range on the quiet-side and the loud side is only limited by analog gain and amplifier power.)

When EQ is applied it's not unusual to go over 0dB (this can happen as a side-effect even when cutting). So most software EQ has a "preamp" setting, which is normally used for attenuation rather than amplification, so that EQ doesn't cause clipping. I would expect DIRAC to take care of that automatically since it's supposed to an automated thing, but I don't know...

 

[skyfly]

I don't have DIRAC, but some general notes -

As you may know, DACs are limited to 0dBFS and will hard-clip if you "try" to go over. CDs and regular (integer) WAV files are also limited to 0dB. Some formats can go over and most DSP is done in floating-point which for all practical purposes has no upper (or lower) limits. But if it goes over 0dB it should be attenuated before it hits the DAC.

CDs and MP3s are usually "maxed out" with 0dB peaks. MP3s often hit +1 or +2dB which will clip your DAC if you play at "full digital volume". The movie formats often have some headroom. So you should assume that there is no digital headroom or what you're calling "dynamic range". (There is a LOT of dynamic range on the quiet-side and the loud side is only limited by analog gain and amplifier power.)

When EQ is applied it's not unusual to go over 0dB (this can happen as a side-effect even when cutting). So most software EQ has a "preamp" setting, which is normally used for attenuation rather than amplification, so that EQ doesn't cause clipping. I would expect DIRAC to take care of that automatically since it's supposed to an automated thing, but I don't know...

I heard that RME ADI-2 has signal processing stages with plenty of digital headroom above 0 dBFS (Conceptually, they add digital bits above 0dBFS and below -144 dBFS. Of course, it is like reducing the level by, say, -48 dB, and calling- 48 dBFS "0 dBFS" by another standard.).

 

[skyfly]

When EQ is applied it's not unusual to go over 0dB

Of course. For example, let us sharply cut 300Hz by EQ. When 100Hz 0dBFS square wave signal goes through the EQ with headroom, the peak output is over 0dBFS.

I also saw some forum posts saying that new clipping (not clipping in the original digital signal source) could occur in sample rate conversion.

 

[skyfly]

Impulse response and frequency response correction graphs in Dirac Live launched from MiniDSP Device Console.

- In a living room in a low ceiling apartment (The place and the Bose 901 system are not mine)
- The approximate 901 PEQ above was used

I cannot notice what was corrected in impulse response. Bose 901 is a rare crossover-less loudspeaker system.

I initially tried 1 dB/oct downward slope which I saw in Erin's corner web site, but the speech intelligibility was not so good on some source material. Thereafter, I used a less steep downward slope in mid. I put a very mild de-ess in the target curve.

A behaviour at the bottom end I did not like about Dirac Live: The R speaker had side wall about 1 meter (guess from my memory, not by measurement) from R speaker's outside edge , while the L speaker did not. Dirac Live boosted L approximately up to 6 dB in approximately 30 - 33 Hz to match the level from wall-boosted R speaker. I did not like the strain on the L speaker due to this automatic Dirac Live correction. So I did some manual bottom end modification.

Note: the measured average spectrum contains the inaccuracy (the difference from real Bose 901 Series VI Equaliser unit) of the approximate 901 PEQ I used. Perhaps, the peak at 2.6 kHz is due to this - I did not see this in the measurement with the Bose 901 Series Equaliser unit -.

Note: though the Bose 901 system was purchased new from a small hi-end audio retailer that the Bose distributor of the country recommended on the phone, there is no guarantee that the system is a genuine Bose product. The Series VI drive units I have seen have darker blue area that spans almost entire cone. The darker blue are on the Dirac Live measured 901 Series VI is much smaller.

An asserted 901 Series VI online.

Much smaller darker blue area on the measured 901 Series VI. (Fake Bose drivers? Illegally re-foamed Series IV on an asserted new 901 Series VI? Who knows?)