• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). Come here to have fun, be ready to be teased and not take online life too seriously. We now measure and review equipment for free! Click here for details.

Review and Measurements of Benchmark DAC3

John_Siau

Technical Expert
Technical Expert
Industry Insider
Joined
Jul 17, 2018
Messages
115
Likes
854
Location
Syracuse, NY USA
OK, finally, finally we get to test the statement Benchmark has made about superiority of their headphone amplifiers. For those of you who have not tracked that discussion, it starts with this paper from Benchmark: https://benchmarkmedia.com/blogs/application_notes/12838141-headphone-amplifiers-part-1

In a nutshell, they say that real headphone loads cause far more distortion than dummy (resistive) loads always used for headphone reviews and measurements. To wit, they show these two measurements, the first being with a dummy load:

View attachment 13329

The Benchmark is in navy color at the bottom showing much better performance than the other two competing amps in pink and green. But that is no their point. Their point is that once you replace the dummy load with a real Sony MDR-V6 headphone, their distortion figures do not change but the competitors do:

View attachment 13330

I confirmed that both my Topping headphone amplifier and RME ADI-2 DAC (or was it the Pro?) are sensitive to headphone loads. And indeed their THD+N in above graph shows similar frequency dependence to competitors of Benchmark above.

What was left was verifying that the Benchmark indeed performed as well as they say it does using my instrumentation.

For that, I tried to set my APx555 analyzer very closely to what they have above. I did not have a 60 ohm dummy load but used a 50 ohm which is close enough. Levels were matched nearly the same at 0.5 watt per above graphs. Bandwidth of the analyzer is set to 90 kHz as opposed to 80 kHz but again, close enough.

Here are the results comparing the RME ADI-2 Pro output versus DAC3 using a 50 ohm dummy load:
View attachment 13331

We get pretty nice graphs with the RME ADI-2 Pro now beating the Benchmark DAC3 by good margin (4 to 7 dB).

Now let's replace our dummy load with the real Sony MDR-V6 headphone:

View attachment 13332

Ah, that ain't good! While the RME ADI-2 Pro output changes as I had measured with my other analyzer, so does the Benchmark DAC3!!!

Reading through the paper from Benchmark, it is from DAC1, not DAC3. Benchmark has newer papers on DAC3 with similar claims.

Not shown but I tested with an IEM at much lower level and it too caused a variation in response that was identical between the two amps.

Summary
It is good of Benchmark to raise awareness of measuring with real headphone loads. But the notion that their headphone amplifiers have distortion profiles that are headphone independent, do not seem true. My measurements show similar susceptibility to other high-performance amplifiers.

Sure, if you have a much higher output impedance than what Benchmark has, the effect will be exaggerated. But the core problem remains in all implementations including that of Benchmark DAC3. Bummer!

P.S. I measured the output impedance of Benchmark DAC3 at 0.7 ohms.
View attachment 13333
The RME vs. DAC3 measurements into a headphone load should match very closely (as you have shown). Both have very low output impedances and both have vanishingly low distortion when driving resistive loads.

At the time the Benchmark paper was written, it was common practice to build headphone amplifiers with a 30-Ohm output impedance. The Hifiman EF2A is a carry-over from these older designs.

The plots we published compared the DAC1 headphone amplifier to two headphone amplifiers that had 30-Ohm output impedances. Some of the distortion was produced by the high output impedance and some was created by the fact that neither of these amplifiers could cleanly drive a 60-Ohm resistive load. Here is the 60-Ohm resistive load plot for the same three headphone amplifiers:

1590692305586.png

See: https://benchmarkmedia.com/blogs/application_notes/12838141-headphone-amplifiers-part-1

It is also important to note that we made the measurements in a very quiet room with the headphones on a dummy head. If you leave the headphones open on the bench while running the test, room noise will give you much higher THD+N. You could use a THD only test, or replicate our procedure to get more accurate results.

Nevertheless, the purpose of the paper was to demonstrate the distortion-related advantages of having a low output impedance. A good follow up would be to use a headphone measurement system to determine how much the distortion changes at the acoustic output of the headphone (where it counts).

Here is a plot that compared the DAC1 headphone amplifier to itself after the insertion of a 30-Ohm series resistor:

1590692468020.png

See: https://benchmarkmedia.com/blogs/application_notes/12982989-headphone-amplifiers-part-2

Please note that a 30-Ohm amplifier driving 60-Ohm headphones gives a damping factor of 60/30 = 2. For the purpose of damping, we would like to see a damping factor of at least 10 (see http://www.collinsaudio.com/Prosound_Workshop/Damping_Factor.pdf ).

This means that the lower impedance headphone amplifiers may produce less distortion at the acoustic output. Amplifiers with an output impedance of less than 6 ohms would provide sufficient damping when driving 60-Ohm headphones, but amplifiers with lower output impedances should offer no additional significant improvements in the THD at the acoustic output.

In the output-impedance range below 6-Ohms, audible improvements will be related to achieving a predictable and repeatable frequency response. Like loudspeakers, the impedance of any set of headphones will vary with frequency. The MDR-V6 has a nominal impedance of 60 Ohms, but this changes substantially over the audio band. The output impedance of the amp and the input impedance of the driver form a voltage divider. This voltage divider is frequency dependent because the headphone impedance changes with frequency.

If you do the math, you will find that a damping factor of 100 will keep the voltage-divider frequency response flat to about +0/-0.5 dB. This assumes that there are frequencies where the impedance is 1/3 of the nominal impedance.

It would take a damping factor of 350 to keep this variation below 0.1 dB. This would require an output impedance of 0.17 Ohms when driving 60-Ohm headphones.

Lets assume we need a damping factor of 100. With 60 Ohm headphones this means that we would need an output impedance of 60/100 = 6 Ohms.

Amplifiers with an impedance higher than 6 Ohms may produce audible differences in the frequency response of a given set of 60-Ohm headphones. This implies that listeners may be able to detect the frequency response changes in a level-matched ABX test between two headphone amplifiers if one achieves a damping factor of 100 and the other does not.

On the basis of frequency response alone, the Hifiman EF2A should sound different than the amplifiers with output impedances of less than 1 Ohm, when driving most headphones. This assumes that the amplifiers are perfectly flat into resistive loads. The audible difference will be produced by the unintended voltage divider that is formed by the output impedance of the amplifier and the input impedance of the headphone. The headphone impedance is frequency dependent and this makes the voltage divider frequency dependent.

Look for my upcoming paper on damping factor here:

https://benchmarkmedia.com/blogs/application_notes

It will appear before the end of May 2020.
 

sprellemannen

Active Member
Forum Donor
Joined
Jul 21, 2018
Messages
255
Likes
302
Location
Europe
Joined
Dec 15, 2018
Messages
77
Likes
271
Location
Fremont, CA
Would it be possible for someone with a DAC3 post its spectral output of white noise at 44.1 kHz? I'd like to see the behavior of its digital filter.
 
Joined
Apr 27, 2020
Messages
22
Likes
3
Location
Sydney
Gain setting question.

My digital source has -25 dB gain (meaning a reduction of 25dB to the digital signal sent to the DAC) to provide headroom for room EQ and parametric filters etc.

My downstream Benchmark DAC is set as per factory which presumably is on the 0dB jumper / pad.

My Benchmark AHB2's have to be set on the 14.2dBu / 4 Vrms to reach comfortable listening levels with the volume at 11 or 12 o'clock, and it is certainly impossible to drive the speakers to clipping as is suggested by the Benchmark manual/ application notes.

Setting the amps to 9.8Vrms makes the speaker output almost inaudible on most listening material unless the volume is adjusted to maximum.

Basically the DAC output level is reduced by 25dB so I need to add at least 15 to 20dB between the DAC2 and the speakers - just not sure what the best option is and looking for some guidance.



So, my question is which of the following gives me the least distortion and noise for the highest gain;
  1. if I set the DAC XLR pads to plus +10 dB (or can I do plus 20dB??)
  2. if I set the AHB2 to 2Vrms (max position increases everything - not just level)
  3. change my unbalanced RCA to RCA (DAC2 to AHB2) to a balanced XLR to RCA cable
  4. something else
 

Vasr

Senior Member
Joined
Jun 27, 2020
Messages
320
Likes
309
My digital source has -25 dB gain (meaning a reduction of 25dB to the digital signal sent to the DAC) to provide headroom for room EQ and parametric filters etc.
What are you using for your room EQ and who is doing the -25db setting? If you did this manually, it may be unnecessary or too conservative. Dirac, for example, reserves -12.5db for the headroom. I don't know of any other correction system that are more aggressive in boost than that. If you are using a roomEQ enabled device then they typically do that headroom reserves themselves before they apply their filters so the upstream device can send it with no attenuation.
 
Joined
Apr 27, 2020
Messages
22
Likes
3
Location
Sydney
What are you using for your room EQ and who is doing the -25db setting? If you did this manually, it may be unnecessary or too conservative. Dirac, for example, reserves -12.5db for the headroom. I don't know of any other correction system that are more aggressive in boost than that. If you are using a roomEQ enabled device then they typically do that headroom reserves themselves before they apply their filters so the upstream device can send it with no attenuation.
So I measure using MMM method and REW over 3 positions for each channel.

Then I build a PEQ filter for each channel in REW which corrects the worst peaks and non-room mode minimums (which needs plus 14dB of headroom on one channel and plus 10dB on the other due to a very asymmetrical room), and averages the 3 MMM measurements per each channel.

Then I measure the room again with DIRAC Live 3.0 and the PEQ filters loaded, and because the PEQ is doing the heavy lifting the DIRAC correction sounds and measures much better (DIRAC can only do +/- 10dB max, so I allow 10dB of headroom).

So with the maximum corrections of +14 and +10 dB = +24 dB therefore reducing the digital signal in the DSP by -25dB gives the required headroom to avoid clipping.

Unfortunately this means I have to increase the gain downstream of the DSP - I'm just not sure where and how is the best place to do so without creating extra noise or distortion.
 

waynel

Active Member
Forum Donor
Joined
Nov 14, 2019
Messages
203
Likes
194
So I measure using MMM method and REW over 3 positions for each channel.

Then I build a PEQ filter for each channel in REW which corrects the worst peaks and non-room mode minimums (which needs plus 14dB of headroom on one channel and plus 10dB on the other due to a very asymmetrical room), and averages the 3 MMM measurements per each channel.

Then I measure the room again with DIRAC Live 3.0 and the PEQ filters loaded, and because the PEQ is doing the heavy lifting the DIRAC correction sounds and measures much better (DIRAC can only do +/- 10dB max, so I allow 10dB of headroom).

So with the maximum corrections of +14 and +10 dB = +24 dB therefore reducing the digital signal in the DSP by -25dB gives the required headroom to avoid clipping.

Unfortunately this means I have to increase the gain downstream of the DSP - I'm just not sure where and how is the best place to do so without creating extra noise or distortion.
Not a great idea to allow for 24dB or gain boost, you are just likely feeding a null. What frequency is the maximum boost applied?
 
Joined
Apr 27, 2020
Messages
22
Likes
3
Location
Sydney
Not a great idea to allow for 24dB or gain boost, you are just likely feeding a null. What frequency is the maximum boost applied?
Agreed.

Not feeding a null, lifting one side/ channel of the room between 25 and 80 Hz, to match the other side which is in a corner, by 14db. The TOTAL correction with PEQ is 14dB MAX! The additional 10dB is to allow the correct headroom for DIRAC to function without clipping. Have tested without any headroom and it clips.

So 25dB of headroom, with 14dB of correction thru PEQ.

Graphs below are after the PEQ (+14dB) for each channel.

29JUL_WIDE LOUNGE_NEW PEQ12_DV1_SENT TO P1-CORRECTION CH1.png
29JUL_WIDE LOUNGE_NEW PEQ12_DV1_SENT TO P1-CORRECTION CH2.png


Image below shows the before and after of the room using the REW PEQ.


MMM4_BOTH CH AV1_NEW PEQ12_P2_CORRECTED v ORIGINAL.jpg


The DSP configuration is up to about version 15 and has been optimised muchly, so I am happy with the headroom / filter setting for now.

But I guess my original question is where to stage the gain.

Accepting the 25dB of headroom as fixed for now - where can I get an extra 20dB!!!

In the DAC I could use the +20dB pad and get most of it back, but probably adds some nasties.

Also I can run the 2Vrms setting on my AHB2 (Mono) and probably get some gain there also, but also adds distortion and noise.
 

waynel

Active Member
Forum Donor
Joined
Nov 14, 2019
Messages
203
Likes
194
Agreed.

Not feeding a null, lifting one side/ channel of the room between 25 and 80 Hz, to match the other side which is in a corner, by 14db. The TOTAL correction with PEQ is 14dB MAX! The additional 10dB is to allow the correct headroom for DIRAC to function without clipping. Have tested without any headroom and it clips.

So 25dB of headroom, with 14dB of correction thru PEQ.

Graphs below are after the PEQ (+14dB) for each channel.

View attachment 75795 View attachment 75796

Image below shows the before and after of the room using the REW PEQ.


View attachment 75797

The DSP configuration is up to about version 15 and has been optimised muchly, so I am happy with the headroom / filter setting for now.

But I guess my original question is where to stage the gain.

Accepting the 25dB of headroom as fixed for now - where can I get an extra 20dB!!!

In the DAC I could use the +20dB pad and get most of it back, but probably adds some nasties.

Also I can run the 2Vrms setting on my AHB2 (Mono) and probably get some gain there also, but also adds distortion and noise.
my advise is stop using the PEQ first and let Dirac do what it can , also suggest moving your sub so it’s not in a null .
 
Joined
Apr 27, 2020
Messages
22
Likes
3
Location
Sydney
my advise is stop using the PEQ first and let Dirac do what it can , also suggest moving your sub so it’s not in a null .
Did that at version 5 or 6 - in other words ages ago - ended up here because of a long process of testing and optimisation.

And not using a sub - full range speakers with 13inch carbon fibre woofer - so position is fixed / optimized so to speak.

DIRAC cannot manage big cuts or lifts as well as the PEQ -the result both measures and sounds worse - see yellow spread below.

PRESET 4-1.png


So, my question is which of the following gives me the least distortion and noise for the highest gain;
  1. if I set the DAC XLR pads to plus +10 dB (or can I do plus 20dB??)
  2. if I set the AHB2 to 2Vrms (max position increases everything - not just level)
  3. change my unbalanced RCA to RCA (DAC2 to AHB2) to a balanced XLR to RCA cable
  4. something else
 
Last edited:

waynel

Active Member
Forum Donor
Joined
Nov 14, 2019
Messages
203
Likes
194
Did that at version 5 or 6 - in other words ages ago - ended up here because of a long process of testing and optimisation.

DIRAC cannot manage big cuts or lifts as well as the PEQ -the result both measures and sounds worse.



So, my question is which of the following gives me the least distortion and noise for the highest gain;
  1. if I set the DAC XLR pads to plus +10 dB (or can I do plus 20dB??)
  2. if I set the AHB2 to 2Vrms (max position increases everything - not just level)
  3. change my unbalanced RCA to RCA (DAC2 to AHB2) to a balanced XLR to RCA cable
  4. something else
And not using a sub - full range speakers with 13inch carbon fibre woofer - so position is fixed so to speak.
4). Something else , Dirac should not be trying to correct a 14dB dip. By trying to correct this you are causing problems for yourself. Either move your speakers, add 2 or more subs and make sure your mains are high pass filtered at 80 hz, or live with a dip.
Also, I can’t imagine why you would use an unbalanced cable between a dac2 and an ahb2.
 
Joined
Apr 27, 2020
Messages
22
Likes
3
Location
Sydney
4). Something else , Dirac should not be trying to correct a 14dB dip. By trying to correct this you are causing problems for yourself. Either move your speakers, add 2 or more subs and make sure your mains are high pass filtered at 80 hz, or live with a dip.
Also, I can’t imagine why you would use an unbalanced cable between a dac2 and an ahb2.
Typo re DAC 2 AHB2 connection - it is indeed a Benchmark interconnect and fully balanced 75 ohm etc etc!

I was actually meaning the DSP to DAC connection, which is unbalanced because the DSP uses XLR out balanced and the DAC uses RCA in balanced.

Long term a couple of subs will indeed be added, but I am happy having a full range three way speaker which can get down into the 20Hz area for now.

7153477A-FDFA-4BDB-B730-AF36F6770289.jpeg


Anyway my friend I thank you for your time and your advice. I will of course revert to your advice and try another round of measurements and filters should this current course of action produce something which sounds and measures worse.

However, after a few months I am VERY happy with the current results, in a room which is asymmetrical and bright, with speakers located a best as they can be, changing between HD650 cans and speakers has the balance, tonality, and space all very close. It is closer to reference than I would have believed possible given the issues with the room. And measuring with PEQ loaded has DEFINITELY allowed DIRAC to refine/ correct more. Particularly the impulse problems one has with a lopsided open room.

Edit[ Ultimately nulls cannot be filled, so the before and after measurements show where the nulls are clearly]

So, let's let some of the other experts contribute, because while I love your advice I think you are not adding anything new at this point, or answering my original post.
 
Last edited:

waynel

Active Member
Forum Donor
Joined
Nov 14, 2019
Messages
203
Likes
194
Typo re DAC 2 AHB2 connection - it is indeed a Benchmark interconnect and fully balanced 75 ohm etc etc!

I was actually meaning the DSP to DAC connection, which is unbalanced because the DSP uses XLR out balanced and the DAC uses RCA in balanced.

Long term a couple of subs will indeed be added, but I am happy having a full range three way speaker which can get down into the 20Hz area for now.



Anyway my friend I thank you for your time and your advice. I will of course revert to your advice and try another round of measurements and filters should this current course of action produce something which sounds and measures worse.

However, after a few months I am VERY happy with the current results, in a room which is asymmetrical and bright, with speakers located a best as they can be, changing between HD650 cans and speakers has the balance, tonality, and space all very close. It is closer to reference than I would have believed possible given the issues with the room. And measuring with PEQ loaded has DEFINITELY allowed DIRAC to refine/ correct more. Particularly the impulse problems one has with a lopsided open room.

So, let's let some of the other experts contribute, because while I love your advice I think you are not adding anything new at this point, or answering my original post.
Do you realize that trying to correct a 14 dB dip requires your amps to put out 25X more power and for your speakers to handle 25 times more power ?
 
Last edited:

Sal1950

Major Contributor
The Chicago Crusher
Forum Donor
Joined
Mar 1, 2016
Messages
7,535
Likes
5,960
Location
Central Fl
Do you realize that trying to correct a 14 dB dip requires your amps to put out 25X more power and for your speakers to handle 25 times more power ?
Scary and dangerous prospect. :eek:
 

waynel

Active Member
Forum Donor
Joined
Nov 14, 2019
Messages
203
Likes
194
Scary and dangerous prospect. :eek:
Right, and I should mention that his dip is centered at 45-65 Hz, precisely where rock music has the most energy and the amps and speakers are likely to be the most strained. This dip is better ignored or solved with 2 or more subs.
 
Joined
Apr 27, 2020
Messages
22
Likes
3
Location
Sydney
Do you realize that trying to correct a 14 dB dip requires your amps to put out 25X more power and for your speakers to handle 25 times more power ?
Scary and dangerous prospect. :eek:
Right, and I should mention that his dip is centered at 45-65 Hz, precisely where rock music has the most energy and the amps and speakers are likely to be the most strained. This dip is better ignored or solved with 2 or more subs.
This is now off-topic re DAC3 and gain staging, but presumably our moderators can move if needed.

I am prepared to be wrong - I'm sure you both feel the same - as rationally learning any new information seems predicated on this principle!

And I am certain I can also learn something new, but I have watched waynel jumping to conclusions on other threads which are simply not borne out by the facts. For example presuming my comparisons with the bass response between AHB2 and other amps were performed in my home where I am using the DSP mentioned previously in this thread - wrong was actually off-site in a studio!


My room modes/ non-DSP frequency response below. You can clearly see each channel (RCH = right channel LCH = left) and the respective measurement position (LS=left seat, CS = centre, RS = right) against the room itself. The REW room modes show at the bottom assume symmetry and therefore are largely useless.
CHANNELS COMPARED.jpg



The central assertions you are making seem to be;
  1. I am using a 14dB PK filter in REW to fill a room mode null
  2. I am over driving the amp and or speakers by doing the above
What I am basing my approach on;
  1. If I place extra energy into a room mode or null, the result is actually nothing = that's why it is a null. If you take the time and effort to look at the images of frequency responses and ensure you are looking at the correct channels you may notice that the null occurs at different places on each channel - this is because it is a width mode (my room is 7.2m wide X 4.7m long X 2.65m high) and my listeners right speaker is 1450mm from the side wall and 1100mm from the front wall (essentially in a corner). On the opposite side my listeners left speaker is 1100mm from the wall but has 3200mm or so to it's side wall, and the room is open forwards and backwards on that side (see CAD image). So by EQing the respective peaks and leaving the nulls I actually gain a listening spot frequency response which is reasonably flat, and as low frequency is not directional the ears don't seem to notice which channel it is coming from, but rather that it is filled in rather than missing. I am going to make another set of measurements, and build a PEQ filter which cuts more and requires less gain. Both to test your idea, and to see if it solves my gain problem. Only issue I can see is this requires large adjustments in the tone controls on the speakers and full frequency PEQ as the highs and mids become too bright if the bottom end frequencies are used as a base for the target slope.
  2. Sadly (I really want to test this AHB2 functionality) despite driving my speakers with 750W amps, or even my two mono AHB2's at 480W (into 6 ohm nominal speakers 125W continuous / 250W peak music) I have NEVER been able to generate clipping at the speaker. I have generated digital clipping by failing to provide enough headroom for DSP/ DIRAC, but even driving the speakers through the DAC / AHB2 directly, which obviously produces 25dB greater volume/ power evenly across all frequencies, the amp has not clipped. Switching between DSP and DAC while playing 20/30/40/50/60/70/80/90 Hz tones at the same level results in very very similar cone extension (less than max generated while using DAC to AHB2 only) despite your claims of 25 times the power. Basically the attenuated frequency now plays at similar volume/ power as the other un-attenuated frequencies did prior to adding DSP/ DIRAC correction. Simply put, plus 14dB minus 25dB = 11dB of which it seems DIRAC uses up to either 10 or 12.5 depending on opinion. Net result is clearly that the gain limitation (-25dB everywhere) prevents the amp from needing or indeed being able to produce any more power at that frequency that it did prior to DSP.

On the AHB2 thread John Siau wrote a great comment working through the actual power the speaker will see at different levels/ volumes. I wonder if you read it? Seems quite different to your 25 times the power claim.

ROOM LAYOUT.jpg
 
Last edited:

waynel

Active Member
Forum Donor
Joined
Nov 14, 2019
Messages
203
Likes
194
This is now off-topic re DAC3 and gain staging, but presumably our moderators can move if needed.

I am prepared to be wrong - I'm sure you both feel the same - as rationally learning any new information seems predicated on this principle!

And I am certain I can also learn something new, but I have watched waynel jumping to conclusions on other threads which are simply not borne out by the facts. For example presuming my comparisons with the bass response between AHB2 and other amps were performed in my home (actually off-site in a studio) where I am using the DSP mentioned previously in this thread - wrong!


The central assertions seem to be;
  1. I am using a 14dB PK filter in REW to fill a room mode null
  2. I am over driving the amp and or speakers by doing the above
What I am basing my approach on;
  1. If I place extra energy into a room mode or null, the result is actually nothing = that's why it is a null. If you take the time and effort to look at the images of frequency responses and ensure you are looking at the correct channels you may notice that the null occurs at different places on each channel - this is because it is a width mode (my room is 7.2m wide X 4.7m long X 2.65m high) and my listeners right speaker is 1450mm from the side wall and 1100mm from the front wall (essentially in a corner). On the opposite side my listeners left speaker is 1100mm from the wall but has 3200mm or so to it's side wall, and the room is open forwards and backwards on that side (see CAD image). So by EQing the respective peaks and leaving the nulls I actually gain a listening spot frequency response which is reasonably flat, and as low frequency is not directional the ears don't seem to notice which channel it is coming from, but rather that it is filled in rather than missing. I am going to make another set of measurements, and build a PEQ filter which cuts more and requires less gain. Both to test your idea, and to see if it solves my gain problem. Only issue I can see is this requires large adjustments in the tone controls on the speakers and full frequency PEQ as the highs and mids become too bright if the bottom end frequencies are used as a base for the target slope.
  2. Sadly (I really want to test this functionality) despite driving my speakers with 750W amps, or even my two mono AHB2's at 480W (into 6 ohm nominal speakers 125W continuous / 250W peak music) I have NEVER been able to generate clipping at the speaker. I have generated digital clipping by failing to provide enough headroom for DSP/ DIRAC, but even driving the speakers through the DAC / AHB2 directly, which obviously produces 25dB greater volume/ power evenly across all frequencies, the amp has not clipped. Switching between DSP and DAC while playing 20/30/40/50/60/70/80/90 Hz tones at the same level results in very very similar cone extension (less than max generated while using DAC to AHB2 only) despite your claims of 25 times the power. Basically the attenuated frequency now plays at similar volume/ power as the other un-attenuated frequencies did prior to adding DSP/ DIRAC correction. Simply put, plus 14dB minus 25dB = 11dB of which it seems DIRAC uses up to either 10 or 12.5 depending on opinion. Net result is clearly that the gain limitation (-25dB everywhere) prevents the amp from needing or indeed being able to produce any more power at that frequency that it did prior to DSP.

On the AHB2 thread John Siau wrote a great comment working through the actual power the speaker will see at different levels/ volumes. I wonder if you read it? Seems quite different to your 25 times the power claim.

View attachment 75924
14dB boost = 25 times the power, that’s not my idea , it’s math. 10logbase10(25) =14. I didn’t jump to any conclusion on the other thread that was proven wrong I asked if you were using this peq with 14 dB boost when you where unhappy with the AHB2. You said no. Fine. The 14dB dip between 45-65Hz may still have something to do with your preference for the other amp over the AHB2 but now I’m thinking it’s more likely in your mind.

anyway let me comment more bluntly on your quest to add a 14 dB boost then run DIRAC on top of that: You’re doing it wrong!
 
Last edited:

Vasr

Senior Member
Joined
Jun 27, 2020
Messages
320
Likes
309
So I measure using MMM method and REW over 3 positions for each channel.

Then I build a PEQ filter for each channel in REW which corrects the worst peaks and non-room mode minimums (which needs plus 14dB of headroom on one channel and plus 10dB on the other due to a very asymmetrical room), and averages the 3 MMM measurements per each channel.

Then I measure the room again with DIRAC Live 3.0 and the PEQ filters loaded, and because the PEQ is doing the heavy lifting the DIRAC correction sounds and measures much better (DIRAC can only do +/- 10dB max, so I allow 10dB of headroom).

So with the maximum corrections of +14 and +10 dB = +24 dB therefore reducing the digital signal in the DSP by -25dB gives the required headroom to avoid clipping.

Unfortunately this means I have to increase the gain downstream of the DSP - I'm just not sure where and how is the best place to do so without creating extra noise or distortion.
Don't want to get into the ongoing spat with others or tell you what to do or not to do but I am just curious about understanding what is going on here with your system if you don't mind.

In terms of your original question, any gains you do in the digital domain can lead to digital clipping from the digital processing in the path. This is why attenuating it by that much was necessary to avoid digital clipping. But you can't necessarily get it back if you boost it back up before the digital handling that caused the clipping which could be in the DAC itself. On the other hand, in the analog domain, many amps can go above reference volume (THX seems to require it for the LFE portion and why LFE is recorded -10db attenuated in the standard encodings), so any gain needs to come downstream of the last digital processing that caused the digital clipping while staying well below amp clipping. As long as that happens you should be ok.

But in terms of making sure what is happening that is causing this situation...

Do you know what the crossover points are in your speakers (assuming it is multi-way)?

I am wondering if you might be correcting an issue with the speaker linearity/phase related (or not) to the crossover rather than just the room mode. Have you tried switching the left and right speakers and measuring to see what happens to isolate the room modes from any speaker issues? If so, what were the results? The 4db difference between them could be the room mode but the common 10db might be a speaker issue. A much rarer situation might be an issue with the measuring mic itself or an incorrect calibration file. Would be good to eliminate these as possibilities in case you have not.

If the PEQ is downstream of Dirac, then you should never see a cumulative +24 db since that area if corrected properly should measure without any troughs in the same area for Dirac to boost additionally in the same area. So, it would be useful to investigate why so much attenuation is needed to avoid clipping which would suggest Dirac is correcting cumulatively in the same area as the PEQ. Looking at the Dirac results should expose those.
 
Joined
Apr 27, 2020
Messages
22
Likes
3
Location
Sydney
Don't want to get into the ongoing spat with others or tell you what to do or not to do but I am just curious about understanding what is going on here with your system if you don't mind.

In terms of your original question, any gains you do in the digital domain can lead to digital clipping from the digital processing in the path. This is why attenuating it by that much was necessary to avoid digital clipping. But you can't necessarily get it back if you boost it back up before the digital handling that caused the clipping which could be in the DAC itself. On the other hand, in the analog domain, many amps can go above reference volume (THX seems to require it for the LFE portion and why LFE is recorded -10db attenuated in the standard encodings), so any gain needs to come downstream of the last digital processing that caused the digital clipping while staying well below amp clipping. As long as that happens you should be ok.

But in terms of making sure what is happening that is causing this situation...

Do you know what the crossover points are in your speakers (assuming it is multi-way)?

I am wondering if you might be correcting an issue with the speaker linearity/phase related (or not) to the crossover rather than just the room mode. Have you tried switching the left and right speakers and measuring to see what happens to isolate the room modes from any speaker issues? If so, what were the results? The 4db difference between them could be the room mode but the common 10db might be a speaker issue. A much rarer situation might be an issue with the measuring mic itself or an incorrect calibration file. Would be good to eliminate these as possibilities in case you have not.

If the PEQ is downstream of Dirac, then you should never see a cumulative +24 db since that area if corrected properly should measure without any troughs in the same area for Dirac to boost additionally in the same area. So, it would be useful to investigate why so much attenuation is needed to avoid clipping which would suggest Dirac is correcting cumulatively in the same area as the PEQ. Looking at the Dirac results should expose those.
Thanks for your response and no I do not mind.


Speakers are 3 way. Crossover points are 500Hz and 6kHz at -12 dB per octave.

I wondered myself regarding the phase/ linearity.

Can try swapping speakers while I re-measure - but I suspect corner v open wall is the culprit.

Mic has calibration file and I use the 90 deg version (not that it seems to make much difference with MMM).

My method is to measure, create a PEQ filter in REW. Then mount that filter in the respective channel of the miniDSP SHD Studio, and run the DIRAC measuring process. So presumably the PEQ is upstream from DIRAC because the measured responses DIRAC generates are much different with PEQ then without. It seems by popular opinion and measurement that this gives the best results with DIRAC if your room is quite bad. A valid argument has been made that excessive correction places demands on other areas of the system which may be net negative. Unfortunately all I can do is try the alternative method and see what the difference - if any - actually is.....
 
Top Bottom