Thanks, Steven. So, at a simplistic level, to get 110dB peak from the unit, and to make the job of the amplifier as easy as possible, a 250W into 8R or better amplifier module would do the job in an active setup, yes?
Last edited:
For transients that loud yes. But just dropping down to 107db will bring that down to 128 watts and 104db should be only 64 watts. I don't think most people seem to realize how loud 104db really is. I mean if you want to do something like a realistic gun shot that is probably about 135db I think and obviously you wouldn't want a speaker that could do that if you don't want to go deaf. A live piano at a distance of say 10 feet would be how loud, that would be a great target for any speaker if you could produce the tonality and impulse response of a real piano.
Hearing damage occurs after 8 hours of continuous exposure to 85 decibels; 4 hours at 88 decibels; 2 hours at 91 decibels; 1 hour at 94 decibels; 30 minutes at 97 decibels; 15 minutes at 100 decibels; 7.5 minutes at 103 decibels; 3.75 minutes at 106 decibels; 1.8 minutes at 109 decibels; 55 seconds at 112 decibels; 28 seconds at 115 decibels.
OSHA Daily Permissible Noise Level Exposure Hours per day Sound level 8@90dB 6@92dB 4@95dB 3@97dB 2@100dB 1.5@102dB 1@105dB 1/2@110dB .1/4 or less 115dB
Hearing damage occurs after 8 hours of continuous exposure to 85 decibels; 4 hours at 88 decibels; 2 hours at 91 decibels; 1 hour at 94 decibels; 30 minutes at 97 decibels; 15 minutes at 100 decibels; 7.5 minutes at 103 decibels; 3.75 minutes at 106 decibels; 1.8 minutes at 109 decibels; 55 seconds at 112 decibels; 28 seconds at 115 decibels.
OSHA Daily Permissible Noise Level Exposure Hours per day Sound level 8@90dB 6@92dB 4@95dB 3@97dB 2@100dB 1.5@102dB 1@105dB 1/2@110dB .1/4 or less 115dB
Last edited:
Okay, I just have in my mind that for good concert seats the max SPL hits 110dB on an orchestral climax, so a good figure to have as a target.
Yes, the piano at that distance should be well realisable - the attack of the piano note is key, and that will come down to the capabilities of the amplifier.
Yes, the piano at that distance should be well realisable - the attack of the piano note is key, and that will come down to the capabilities of the amplifier.
The intention is not to blast the listener, it's to be sure that the sound is always "effortless", and having the engineering sufficient to achieve that is key, IMO. Modern compressed recordings are impossible to take at any sort of elevated levels if reproduced well, that average level exposure you mention gets to you fairly quickly - but for recordings with decent dynamics there isn't a problem.
I would say for my speaker an upper level of 104db would be ideal With a 120 watt amplifier that would give you 3db of headroom for transients and would keep you away from amplifier clipping. don't forget that the amplifier rating at 120 watts will have less than 10ppm distortion level.
Yes, 120W would be fine ... I found clean 60W with 90dB or so sensitivity speakers to give a big, rich sound on classical - plus, it could rock out nicely when asked to do that 🙂.
Target spec, features, time frame & flexibility....
Hi All,
Here is an overview on the development task we are looking at:
(1) Thread focus:
As both Steven and I have our respective loudspeakers (drivers / cabinets / loading) pretty much finished, in this thread we are focusing on the electronics to feed / power and control them.
(2) Time frame:
Assume first 90 days or so to brainstorm and find the guys who are willing and able to contribute time and effort to the thread.
Knock the concept into some sort of shape and establish broad brush parameters like max SPL / dynamic range, frequency range, power bandwidth, time domain performance.
Out of this target spec comes a broad cost / margin / RRP figure.
(3) Components:
Assume a good 6 months to identify, source and test different components, this is the hardest part, it takes time, effort and ££!
(4) Final design choices:
Assume 90 days to build working prototypes, write basic user manual, beta test and gather feedback.
(5) Finalise the design:
Assume 60 days to build first batch for group buy and finish user manual.
(6) After sales and Support:
Those who share in the profits must be available to offer support and after sales to the customers!
So Sept 2016 is the realistic target to bring the dream system to market....
If you think that sounds too slow, just look at the time others take....Hypex DLP / Naja / DEQX etc .....Years of ongoing beta test / re design and they are only looking at half the work load we are contemplating....
Hope this is answers a few of the basic questions and gives everyone an idea of the task and time frame.
All the best
Derek.
Hi All,
Here is an overview on the development task we are looking at:
(1) Thread focus:
As both Steven and I have our respective loudspeakers (drivers / cabinets / loading) pretty much finished, in this thread we are focusing on the electronics to feed / power and control them.
(2) Time frame:
Assume first 90 days or so to brainstorm and find the guys who are willing and able to contribute time and effort to the thread.
Knock the concept into some sort of shape and establish broad brush parameters like max SPL / dynamic range, frequency range, power bandwidth, time domain performance.
Out of this target spec comes a broad cost / margin / RRP figure.
(3) Components:
Assume a good 6 months to identify, source and test different components, this is the hardest part, it takes time, effort and ££!
(4) Final design choices:
Assume 90 days to build working prototypes, write basic user manual, beta test and gather feedback.
(5) Finalise the design:
Assume 60 days to build first batch for group buy and finish user manual.
(6) After sales and Support:
Those who share in the profits must be available to offer support and after sales to the customers!
So Sept 2016 is the realistic target to bring the dream system to market....
If you think that sounds too slow, just look at the time others take....Hypex DLP / Naja / DEQX etc .....Years of ongoing beta test / re design and they are only looking at half the work load we are contemplating....
Hope this is answers a few of the basic questions and gives everyone an idea of the task and time frame.
All the best
Derek.
Binely,
Derek certainly can give you the best answer to this as he has been working with this for over 10 years. I'll give you the little I know, as I have thought about this too.
Experienced DIY speaker builders using Acourate and digital crossovers in the fashion I described recommend trying to achieve the flattest uncorrected speaker response possible, then apply the correction. Non-linearities outside the range where the driver is being used (meaning in the crossover region) is taken care off by use of very steep crossover slopes. I believe Derek mentioned elsewhere he uses 96 dB/octave, for example.
Within the range where the driver is being used anomalies are corrected down. In other words, sensitivity is brought down to the worst point within the range as the system is not boosting valleys in the frequency response but rather reducing peaks to a level you define in the target curve. If there is a really nasty valley you can choose to define your target curve above that so your overall efficiency doesn't go down the drain, and opt to live with a less than ideal response.
So starting with the closest you can to a flat response still pays off.
Moderators, is there a way to block the "This message is hidden because so-and-so is on your ignore list" - message?
By now, I am getting increasingly fed up with the same guy always causing the same diarrhea of these messages, whenever a thread becomes interesting. Even when put on the ignore list, his unrelenting cerebral incontinence continues to litter too many threads with these messages.
By now, I am getting increasingly fed up with the same guy always causing the same diarrhea of these messages, whenever a thread becomes interesting. Even when put on the ignore list, his unrelenting cerebral incontinence continues to litter too many threads with these messages.
The secret of great DSP filters & Eq
Short answer ...Yes, but not for the obvious reasons!
This is close to the heart of what separates "text book" filters (crossovers, driver Eq and or Room Eq) from the best studio grade filters and Eq.
Or put another way filters which look good on paper but dont sound natural Vs ones which look almost identical on paper but sound beautifully natural.
There are 1,000's of DSP / software designers who can all come up with correct filters on paper. FIR, IIR, combinations or hybrid, that's all easy, and in fact there are dozens of good commercial examples available.
But just like the term "linear Phase" things are not what they seem....Linear phase still has group delay...Thats what filters with slopes do, its just the degree's of delay that vary with the slope.
The magic is in how you apply the correction.
Some detail and examples:
Lets say you have a very well designed pair of monitors and great system set up in a room.
All good but you know there are room / speaker issues that you want to correct.
You can have a huge PC loaded with powerful software written by the smartest computer programmers, mathematicians & designers.
Plus a huge stack of hardware and £10,000 B&K measurement microphones.
Lets assume you now have a perfectly accurate room / speaker measurement(s).
It all counts for nothing unless you what to do with that measurement....How to apply the correction & what correction do you apply?
This is what separates the legendary sound guys from the average sound guys.
The instinctive genius of Quincy Jones & MJ from the latest "Dizzy DJ & cool dude"...!
For example, lets say your measurement shows a falling response from 10KHz which is 10dB down by 20KHz. There are also several additional narrow band peaks & troughs along the way.
If all that was required was a simple " flatten every peak and fill in every trough" algorithm we all not be having this conversation....!!
Lets look at the simplest example in more detail:
A 5dB peak at 2KHz covering 1,500 Hz to 2,500 Hz....
Do you flatten it?
Do you half flatten it?
Do you do it with one correction or use two overlapping corrections?
Should these Q's be identical, similar or different altogether?
How will the phase be affected if I overlap two different correction filters with different Q's?
How will the higher and lower correction bands affect this band?
A lot of questions that all must be answered correctly as the cause and effect of interacting filters is accumulative....
And these questions all assume you are aiming for a simple 100% flat response....As we all know that sounds very unnatural....So what about the "Fletcher Munson" curves....? Where / how do I fit them in?
What about different SPL's the "final mix" will sound different at different SPL's....Can I auto compensate?
And all of the above is just in Room Eq....
Examining driver performance gets even more complex with time domain performance overriding frequency response performance....
I spent 10 years maxing out the DEQX software with external DAC's / battery power supplies / non resonant casework / silver wiring / Bybee's / snake oil and then some.....
It took a long time and lots of money to learn that the best mathematicians and code writers (DEQX have the best of both) can not walk into a studio and mix a record.....
Its knowing how to create the correct instruction algorithms to create the right filters....Not the filters themselves!![/B]
I hope this makes sense and goes some way toward explaining why so many different studio software packages and plugins sound so different when on paper they claim to do the same thing.
Cheers
Derek.
Short answer ...Yes, but not for the obvious reasons!
This is close to the heart of what separates "text book" filters (crossovers, driver Eq and or Room Eq) from the best studio grade filters and Eq.
Or put another way filters which look good on paper but dont sound natural Vs ones which look almost identical on paper but sound beautifully natural.
There are 1,000's of DSP / software designers who can all come up with correct filters on paper. FIR, IIR, combinations or hybrid, that's all easy, and in fact there are dozens of good commercial examples available.
But just like the term "linear Phase" things are not what they seem....Linear phase still has group delay...Thats what filters with slopes do, its just the degree's of delay that vary with the slope.
The magic is in how you apply the correction.
Some detail and examples:
Lets say you have a very well designed pair of monitors and great system set up in a room.
All good but you know there are room / speaker issues that you want to correct.
You can have a huge PC loaded with powerful software written by the smartest computer programmers, mathematicians & designers.
Plus a huge stack of hardware and £10,000 B&K measurement microphones.
Lets assume you now have a perfectly accurate room / speaker measurement(s).
It all counts for nothing unless you what to do with that measurement....How to apply the correction & what correction do you apply?
This is what separates the legendary sound guys from the average sound guys.
The instinctive genius of Quincy Jones & MJ from the latest "Dizzy DJ & cool dude"...!
For example, lets say your measurement shows a falling response from 10KHz which is 10dB down by 20KHz. There are also several additional narrow band peaks & troughs along the way.
If all that was required was a simple " flatten every peak and fill in every trough" algorithm we all not be having this conversation....!!
Lets look at the simplest example in more detail:
A 5dB peak at 2KHz covering 1,500 Hz to 2,500 Hz....
Do you flatten it?
Do you half flatten it?
Do you do it with one correction or use two overlapping corrections?
Should these Q's be identical, similar or different altogether?
How will the phase be affected if I overlap two different correction filters with different Q's?
How will the higher and lower correction bands affect this band?
A lot of questions that all must be answered correctly as the cause and effect of interacting filters is accumulative....
And these questions all assume you are aiming for a simple 100% flat response....As we all know that sounds very unnatural....So what about the "Fletcher Munson" curves....? Where / how do I fit them in?
What about different SPL's the "final mix" will sound different at different SPL's....Can I auto compensate?
And all of the above is just in Room Eq....
Examining driver performance gets even more complex with time domain performance overriding frequency response performance....
I spent 10 years maxing out the DEQX software with external DAC's / battery power supplies / non resonant casework / silver wiring / Bybee's / snake oil and then some.....
It took a long time and lots of money to learn that the best mathematicians and code writers (DEQX have the best of both) can not walk into a studio and mix a record.....
Its knowing how to create the correct instruction algorithms to create the right filters....Not the filters themselves!![/B]
I hope this makes sense and goes some way toward explaining why so many different studio software packages and plugins sound so different when on paper they claim to do the same thing.
Cheers
Derek.
Last edited:
Overkill,'
I was reading through the Naja thread, I didn't finish it, I finally got to the point that he was receiving his first production in Dec. 2012, You were one of the participants in that discussion. What conclusion did you or could you draw from that exercise. I see the unit for sale in an enclosure now on his site. Was this a satisfactory approach to the active speaker control, yes I know it did not do any analysis or room adjustment, but the basic concept of speaker control, was this system level powerful enough to control whatever speakers I assume you had personally? I know he was making this a commercial product so I can't say anything about the cost but I am asking about the usage, is this the type of direction you were looking for. If not what other function would you consider necessary outside of room correction and perhaps dynamic speaker correction?
I know I could ask this more elegantly but I just woke up to the conversation, yawn.
I was reading through the Naja thread, I didn't finish it, I finally got to the point that he was receiving his first production in Dec. 2012, You were one of the participants in that discussion. What conclusion did you or could you draw from that exercise. I see the unit for sale in an enclosure now on his site. Was this a satisfactory approach to the active speaker control, yes I know it did not do any analysis or room adjustment, but the basic concept of speaker control, was this system level powerful enough to control whatever speakers I assume you had personally? I know he was making this a commercial product so I can't say anything about the cost but I am asking about the usage, is this the type of direction you were looking for. If not what other function would you consider necessary outside of room correction and perhaps dynamic speaker correction?
I know I could ask this more elegantly but I just woke up to the conversation, yawn.
Vacuphole,
I feel your pain and he was on my ignore list for a long time. I just answer the questions that have any type of real question and stay away from some of the personal weirdness. I have someone else who is still on my list, not to be named and gladly not here.
I feel your pain and he was on my ignore list for a long time. I just answer the questions that have any type of real question and stay away from some of the personal weirdness. I have someone else who is still on my list, not to be named and gladly not here.
I imagine that there are as many approaches to all this as Derek has just stated about so called identical mathematics filter types. What I think Derek would agree with is that you want the speaker as best as possible before applying any corrections. I wouldn't want to take an off the shelf commercial speaker with any serious issues and try and make all the corrections with electronics, that would be equivalent to a brute force effort to break an encryption.
I think the object is to first create the best speaker itself first and use the minimum amount of processing. As Derek said, and I see it all the time. in some of the DSP correction software discussion people trying to fix every little error that can be measured. Notch filters all over the place and equalization at both extremes besides the in-band corrections.
As I learned long ago most any filter has some residual artifact after insertion, and the interactions across decades outside the intended filter function is mostly ignored it seems. You have people mixing linear phase and minimum phase filtering, I can see so many unintended consequences to this. Now add in those who are pushing a device to the limit of its useful range and it gets even worse.
I don't agree with the concept of trying to match two devices by working in the roll-off of the devices, I would rather take two speakers that can actually overlap over a fairly wide area, they are both working in their flat passband area and use the crossover to do the combining, not use the acoustical rolloff as part of the network that way. I'm sure I will get a lot of disagreement about this, this is a personal opinion.
I think the object is to first create the best speaker itself first and use the minimum amount of processing. As Derek said, and I see it all the time. in some of the DSP correction software discussion people trying to fix every little error that can be measured. Notch filters all over the place and equalization at both extremes besides the in-band corrections.
As I learned long ago most any filter has some residual artifact after insertion, and the interactions across decades outside the intended filter function is mostly ignored it seems. You have people mixing linear phase and minimum phase filtering, I can see so many unintended consequences to this. Now add in those who are pushing a device to the limit of its useful range and it gets even worse.
I don't agree with the concept of trying to match two devices by working in the roll-off of the devices, I would rather take two speakers that can actually overlap over a fairly wide area, they are both working in their flat passband area and use the crossover to do the combining, not use the acoustical rolloff as part of the network that way. I'm sure I will get a lot of disagreement about this, this is a personal opinion.
Sorry Derek, 100 % flat response (with controlled directivity) is exactly what makes loudspeakers sound neutral and realistic. Look at the curves of the latest Genelec and the JBL M2. Dead flat. For a reason.
Fletcher-Munson comes in after sound has entered your ear. The gray blob behind it knows what a natural sound (by definition flat FR) sounds like after this natural filter has pushed the ends down.
Fletcher-Munson comes in after sound has entered your ear. The gray blob behind it knows what a natural sound (by definition flat FR) sounds like after this natural filter has pushed the ends down.
Vacuphile,
Though I haven't looked at the FR of that JBL speaker the one demo of it I heard has a oh so typical horn honk PA sound. I can't say anything about the implementation but just getting the FR flat and a highly directional sound is not all there is to it.
Fletcher Munson is alive and well, you do have to make some real changes to the FR as you change level. Do you do this actively or do you leave this up to the end user, the listener to adjust to their preferences?
Leaving out of the equation the impulse response and the decay rate can make two identical looking FR and even dispersion angles sound completely different. If it was only that easy.
Though I haven't looked at the FR of that JBL speaker the one demo of it I heard has a oh so typical horn honk PA sound. I can't say anything about the implementation but just getting the FR flat and a highly directional sound is not all there is to it.
Fletcher Munson is alive and well, you do have to make some real changes to the FR as you change level. Do you do this actively or do you leave this up to the end user, the listener to adjust to their preferences?
Leaving out of the equation the impulse response and the decay rate can make two identical looking FR and even dispersion angles sound completely different. If it was only that easy.
Start with great drivers....
Hi All,
Totally agree with Steven, starting with best drivers / cabinet / loading you can achieve is the way to go.
One key point:
Eq only works with drivers with very broad power response ie any and all drivers which "beam" or are allowed to beam through poor Xover design, can not be successfully Eq'd or corrected.
Regardless of which angle (on or off axis) you measure, the applied Eq will correct all angles (vertical and horizontal) ie the entire power bandwidth.
This is a huge point and one which almost always ignored or simply not understood.
Re Naja, Its been a long time since I checked out that thread, but I get the feeling the new half price Hypex DSP crossover / DAC unit is going to be a tough act to beat once they release the final version.
The Hypex is going to end up around £500 to £600 by the time its in a nice box with good power supply....Pretty good value I think.
That box is ideal for the vast majority of DIY guys....You'd be surprised how many high end brands OEM it and sell it as part of very expensive systems.
Personally I am on a different track, both implementation and performance.
Through our Manchester University JV we are working with top musicians as well as Professors to combine cutting edge & technically accurate DSP with natural & musical algorithms created by artists and producers...
"Science with Soul" is our company tag line....It says it all really.
This DSP will then be embedded at the heart of a modular system in which each component offers world class performance as a stand alone product, but when combined in a perfectly matched system takes the system performance to new heights....
All the best
Derek.
Hi All,
Totally agree with Steven, starting with best drivers / cabinet / loading you can achieve is the way to go.
One key point:
Eq only works with drivers with very broad power response ie any and all drivers which "beam" or are allowed to beam through poor Xover design, can not be successfully Eq'd or corrected.
Regardless of which angle (on or off axis) you measure, the applied Eq will correct all angles (vertical and horizontal) ie the entire power bandwidth.
This is a huge point and one which almost always ignored or simply not understood.
Re Naja, Its been a long time since I checked out that thread, but I get the feeling the new half price Hypex DSP crossover / DAC unit is going to be a tough act to beat once they release the final version.
The Hypex is going to end up around £500 to £600 by the time its in a nice box with good power supply....Pretty good value I think.
That box is ideal for the vast majority of DIY guys....You'd be surprised how many high end brands OEM it and sell it as part of very expensive systems.
Personally I am on a different track, both implementation and performance.
Through our Manchester University JV we are working with top musicians as well as Professors to combine cutting edge & technically accurate DSP with natural & musical algorithms created by artists and producers...
"Science with Soul" is our company tag line....It says it all really.
This DSP will then be embedded at the heart of a modular system in which each component offers world class performance as a stand alone product, but when combined in a perfectly matched system takes the system performance to new heights....
All the best
Derek.
Kindhornman,
I don't think you need to compensate for loudness. All those dedicated buttons have long disappeared from even the most tacky gear. Your brain knows what to expect at a certain SPL.
Please do not mix up controlled directivity and highly directional sound. The latter is not what you want, but the first.
I haven't heard the M2, so I can't comment, but the point was that the design aim was to have a flat FR, in which they succeeded. The reason being that research by Toole and Olive very clearly showed that this is what their customers preferred, in combination with controlled directivity over as large a listening window as possible.
I don't think you need to compensate for loudness. All those dedicated buttons have long disappeared from even the most tacky gear. Your brain knows what to expect at a certain SPL.
Please do not mix up controlled directivity and highly directional sound. The latter is not what you want, but the first.
I haven't heard the M2, so I can't comment, but the point was that the design aim was to have a flat FR, in which they succeeded. The reason being that research by Toole and Olive very clearly showed that this is what their customers preferred, in combination with controlled directivity over as large a listening window as possible.
No need to apologise for being wrong....!
Dead flat does not sound good.
Nor does it sound life like.
There has never been a Genelec speaker which sounds lifelike.
£500 or £5,000 they are fundamentally flawed. Popular due to great dealer / distributor margins and very good marketing, but no better than B&W.
The widespread use of Genelec (and similar "compact" monitors) studio monitors is one of the reason so many studios produce such terrible mixes.
Just like your low grade but popular "reference" of Mini DSP, Genelec is not an appropriate reference for this thread.
Cheers
Derek.
- Status
- Not open for further replies.