1) I look for used home theater amplifiers on Craigslist/Facebook/Offerup etc that have multi-channel inputs. Each DSP channel goes into one multi-channel input then out one of the amplified channels. That's my go-to option for rapid development. Rapid development is just when I see someone talking about something on this forum and I decide to try it to see what they're talking about.
2) I think there are widely known good reasons to abandon analog, engineering + psychoacoustic + streaming services that let you experience many different types of music. In addition, I think there's going to be another amazing reason to go straight digital. In the near future software will write song lyrics and virtualize vocalists. You'll be able to listen to Mick Jagger sing a song he never sang in real life. You'll get a personal live performance and it will be better than if he had sung it in real life due to better recording studio production. The software will become the recoding studio as it creates the music and it won't make mistakes unless you tell it to make mistakes.
Within the decade we won't listen to recorded music anymore. Instead, software will go back and re-create recorded music, eliminating studio recording errors. Same thing will happen with movies. All the bad movie sound mixes will be re-created from scratch.
Amazon is full tilt into artificial intelligence and they have the world's entire (almost) music and movie catalog on their servers. They'll take old music then re-create it from the ground up and separate it into multi-channel. Our voice assistant microphones will handle the room correction for us by communicating with active speakers. Audio will be at an entirely different level of amazingness -- and absurdly low cost relative to today's standards -- by the end of this decade. It's wild to think about it. We will be able to say, "Alexa, I really like Led Zeppelin's Bron-Y-Aur but it's too short. Make me a 5 minute version. Also, make it a studio cut where Robert Plant interjects with a comment to Jimmy Page about adding lyrics."
(Regarding my comments about lyric generation, google GPT-3 text outputs.)
There's a bunch of 8-channel prosound amps.
Balanced inputs always, but easily converted to unbalanced as they are almost always euroblock input connections needing only an external pin jumper to convert.
(But who the heck in their right mind wants unbalanced to begin with, other than being hung with unbalanced sources...
)
If you stay on top of ebay for used install amps, it can get quite inexpensive.
Balanced inputs always, but easily converted to unbalanced as they are almost always euroblock input connections needing only an external pin jumper to convert.
(But who the heck in their right mind wants unbalanced to begin with, other than being hung with unbalanced sources...
)If you stay on top of ebay for used install amps, it can get quite inexpensive.
I'm an outcast and use the input PEQs for digital RIAA eq. But BSC and LF rolloff would seem good uses for these?
So for me, output EQs are about flat mag and phase in as near anechoic conditions as possible,
and input EQs are about tone control and mating the flat speaker to the room/environment.
we all have our own methods, huh ?
You are way ahead of me. I'm still at the speaker equivalent of monty pythons 'Gumby flower arranging'. I'm here to try and inform some of my insane choices
I'm an outcast and use the input PEQs for digital RIAA eq. But BSC and LF rolloff would seem good uses for these?
You are way ahead of me. I'm still at the speaker equivalent of monty pythons 'Gumby flower arranging'. I'm here to try and inform some of my insane choices
Haha, funny! And yeah, me and monty python be mates too
I think input EQ's are perfect for RIAA, baffle step correction, and LF rolloff...
When i see all the calculations and work that folks put into BSC, my eyes roll back in my mellon.
Just put some dang real-time shelving EQ's in place, twist some knobs, and get guaranteed results
I don't think it does and is questionable to oversample an FIR EQ. Dave Gamble who wrote EQuilibrium gives his view here DMG Audio : Dev Blog
My real question should have been what sample rate are you running the DAW at when comparing the filters with different taps?
No they do not, but in a lot of EQ plugins those differences have been put in on purpose rather than being an example of bad coding vs good.
I've also got EQuilibrium and Fab Filter Pro in my processing stack.
This can be an example of comparing apples to oranges and deciding that apples are better
I know you love that soap box. I wouldn't bother trying to nudge it by trial and error. I would either use an optimizer or invert the response to see the shape of the difference curve but without some practice you are right the other methods are simpler and more foolproof.
Very true an FIR does not mean phase has been manipulated and it is self describing so can be transferred between systems. As stated above IIR EQ plugins for whatever platform are all different and matching them is a total PITA.
We don't have to agree on everything
In an analogue LR4 multiway you must cascade the filters properly to get the filters to sum as intended, this does not apply in the same way for a DSP implementation. An LR4 needs time alignment to work as intended whether that be digital delay or analogue all pass, if you don't it won't combine properly. 10 minutes with REW and a few synthetic crossover curves from rephase will show that easily, which you know yourself.
The LX521 is a 4 way IIR crossover designed by Linkwitz himself. It is one of the best speakers available. I use this as an example because I had one. I now have speakers with almost perfect time response but I don't believe that has much to do with why I like them more.
My opinion is that phase rotation above 400Hz or so does not matter that much from an audibility perspective. It is very hard to hear and if heard is even harder to decide whether it is better.
Ha, It's not that easy to find a real example of where the difference between 60,000 and 200,000 taps can be shown to matterBut Wow, that's pretty extreme alright !!!
Brick wall at 15Hz ??
@Bradleypnw - Phew, I get it. Maybe "Let me hear Larry Carlton jam on Blue Steel Blues" as well. I worked at Amzn for about a year and suggested in an employee contest that they could make all tracks of an artists recording available to the consumer, to mix as they please. Dial tone in response - I think I got one "like" from another employee reading contest entries.
Maybe, in line with your prediction: "Alexa - a little less drums. And get rid of that ridiculous tambourine at the into" I should live to see the day.
Maybe, in line with your prediction: "Alexa - a little less drums. And get rid of that ridiculous tambourine at the into" I should live to see the day.
Note that what I'm about to say is my own speculation and not something I have any experience with
I personally have always had a bit of an issue with the idea of out of band flattening to allow a standard electrical order filter to produce the desired acoustic rolloff.
The reason being is that I can't imagine that getting a driver flat (especially if that means fighting it's natural rolloff) can be done without some compromises. Removing peaks like breakup peaks I have no issue with, but bringing levels up to flat is a different kettle of fish.
Maybe it is not an issue because you are attenuating it anyway and it is a means to an end, but something just does not sit well with me with this approach, possibly because I don't understand what is actually happening.
Tony.
This goes back to picking the right drivers and crossover points for the system.
You shouldn't be fighting the natural rolloff of the driver or trying to EQ breakup in a crossover region.
If you don't get the first half of the design right not amount of fiddling with the crossover will fix it.
The point of this approach is to match the acoustic slopes of the crossover to the desired target. Flattening a driver with reasonable response outside the crossover region by up to two octaves either side to ensure that the electrical slopes combine as they should is really no different to adding extra components in a passive crossover to flatten the response. When you look at the combined EQ and crossover for each side it will look just like a complicated passive crossover would.
It's an easier way of getting to the same result.
How much EQ is needed and whether that EQ should be used will depend a lot on the driver.
In a DSP system boost or cut shouldn't matter as long as you keep the gain structure in mind and the processing is done at a higher bit rate than the incoming audio. Not all of those things happen so cuts are always safer but with limited numbers of EQ slots boosts can make sense.
You shouldn't be fighting the natural rolloff of the driver or trying to EQ breakup in a crossover region.
If you don't get the first half of the design right not amount of fiddling with the crossover will fix it.
The point of this approach is to match the acoustic slopes of the crossover to the desired target. Flattening a driver with reasonable response outside the crossover region by up to two octaves either side to ensure that the electrical slopes combine as they should is really no different to adding extra components in a passive crossover to flatten the response. When you look at the combined EQ and crossover for each side it will look just like a complicated passive crossover would.
It's an easier way of getting to the same result.
How much EQ is needed and whether that EQ should be used will depend a lot on the driver.
In a DSP system boost or cut shouldn't matter as long as you keep the gain structure in mind and the processing is done at a higher bit rate than the incoming audio. Not all of those things happen so cuts are always safer but with limited numbers of EQ slots boosts can make sense.
Ro808,
I've heard a bunch of Kinoshita's Rm inspired loudspeakers (which are two ways 2x15 ( sometimes mtm) + horn loaded 2" or 1,5") processed with fir crossover and found them to be overall 'better' sounding to me than passive equivalent.
These design ask for 24db slope even in passive that may be a reason, it is easier to time align the drivers and ask a bit 'more' to the CD/horn combo too which plays a role imho.
Looking for asymetric filter slope. Are you gonna use a coax in the end?
I'm waiting to hear the last 12" BMS and could push the trigger on this one.
About boost / cut: Agree with Fluid but there is another point to avoid boost as much as possible as our brains are more likely to spot treatment* because of phase change which are more obvious/easily heard when 'boosting' ( with iir eq analog, digital passive active, whatever...).
This make another reason to look for high efficiency drivers, this way you have 'spare' db to 'sacrifice' at first. Gives a greater headroom too which can not be detrimental.
Even for linkwitz transform ( i really like the 'freedom' closed box give with LT to shape the lower part of a way) i try the 'lossy' aproach ( only cut) when there is enough efficiency to boot on, iow with pa drivers.
* i think it comes from our evolution and the fact we are used to comb filtering to evaluate our acoustic environnement. We are more used to listen cut than boost 'naturaly'.
I've heard a bunch of Kinoshita's Rm inspired loudspeakers (which are two ways 2x15 ( sometimes mtm) + horn loaded 2" or 1,5") processed with fir crossover and found them to be overall 'better' sounding to me than passive equivalent.
These design ask for 24db slope even in passive that may be a reason, it is easier to time align the drivers and ask a bit 'more' to the CD/horn combo too which plays a role imho.
Looking for asymetric filter slope. Are you gonna use a coax in the end?
I'm waiting to hear the last 12" BMS and could push the trigger on this one.
About boost / cut: Agree with Fluid but there is another point to avoid boost as much as possible as our brains are more likely to spot treatment* because of phase change which are more obvious/easily heard when 'boosting' ( with iir eq analog, digital passive active, whatever...).
This make another reason to look for high efficiency drivers, this way you have 'spare' db to 'sacrifice' at first. Gives a greater headroom too which can not be detrimental.
Even for linkwitz transform ( i really like the 'freedom' closed box give with LT to shape the lower part of a way) i try the 'lossy' aproach ( only cut) when there is enough efficiency to boot on, iow with pa drivers.
* i think it comes from our evolution and the fact we are used to comb filtering to evaluate our acoustic environnement. We are more used to listen cut than boost 'naturaly'.
Last edited:
What makes you think this it makes no sense to me at all. With IIR EQ the phase changes with the magnitude so whether you cut or boost the phase will change and the idea is to have a flat magnitude and phase curve.
I could understand it more in a linear phase filter where the magnitude might move further away from it's natural phase. This is I think why many studio people say they don't like the sound of linear phase EQ.
Magnitude, phase AND Q.
And this last one should not be forgoten.
What makes me think that? Well experience with eq in my audio engineering career: first by word spells by senior engineers when i was student and assistant ( no proof of evidence i agree but still experience).
Second by the fact that when you use eq when tracking or mixing we are not alwqys about 'technical perfection' of correction but most often about artistic choice.
In this case you usually set a freq and a gain and quickly twidle the q knob withoutgoing further than what please the ear atm ( so you induce 'error' of phase and magnitudes which weren't present at first). When yo do this and boost it is far more audible than when you try the same things in cut ( even if you need more bands to achieve same results).
For me it is phase related as i've had access to different analog desk which implement different filter topology despite being both with same overall topology - serial connected filter cells-( Ams/Neve V series and SSL 4000G/E): constant Q design for Neve ang SSL G modules, constant gain design for SSL E.
Most engineers prefer eq E than G or V eq. Why? Because they are more 'musical' to them and 'add' something to the signal despite V and G are considered more 'accurate' or 'surgical' and less obvious when used. What is the difference between them: the way the behave regarding phase.
In this case it can't be the topology ( paralell filters cell are usually considered as 'cleaner' -there is less interaction between bands- eg: ITI eq, GML eq -the one which patented parametric eq- or Sontec).
There is same thing with every eq type existing each one does have certain behavior regarding all this: depend from the type of circuit and technology... but i drift.
This is why i don't get some comment about digital eq sounding more or less the same being digital... they are based around analog circuit choice and recreation so share the same issue if not FIR. Some try to 'clone' analog circuit behavior, some doesn't. In 25 years i've had alot of them and the different generations of coding proved to be an improvement over the previous till around 2010 when for me a plateau was reached regarding quality in pro world ( some digital eq are equals or better than their analog counterparts which was and still is the benchmark).
It was the same with convolution engine and reverb: circa 2000 the fuss was about convolution reverb impulse response and all that... did it became the revolution people thoughts ? No because quickly people seen quality has a price and all this algo weren't equal in quality. And yes i had acces to a very good convolution reverb ( the biggest Sony of the time) and Waves IR1 and they didn't 'sound' the same using same impulse...
Which leads to:
Studio people don't like linear phase badly implemented. Put a Weiss eq on the table and nobody will complains. Can't be useful for everything though... sometimes a classical Pultec or a 'simple' semi/quasi/parametric is all is needed.
But it is not the same goal as the one you target about loudspeaker management system.
Last edited:
That now makes sense why you think that and your experience is no doubt correct. I think sometimes an experience in one situation can predispose you to think that way about another situation though.
In loudspeaker correction particularly DSP based if you keep the level constant and the overall response the same then whether you get there by cut or boost does not change the sound if nothing stupid has been done like clipping or very high gain high Q filters.
In DSP's where there is limited headroom then a pre gain cut has to be applied before the boost to avoid clipping. Cut too much and you lose resolution not cut enough and you get digital clipping which is horrible.
I think it all depends on the raw vs desired response as to whether cutting or boosting makes sense. Having a target response in REW and being able to adjust that up or down gives you an instant visual indication of how much cut or boost is needed and which makes more sense to use.
In loudspeaker correction particularly DSP based if you keep the level constant and the overall response the same then whether you get there by cut or boost does not change the sound if nothing stupid has been done like clipping or very high gain high Q filters.
In DSP's where there is limited headroom then a pre gain cut has to be applied before the boost to avoid clipping. Cut too much and you lose resolution not cut enough and you get digital clipping which is horrible.
I think it all depends on the raw vs desired response as to whether cutting or boosting makes sense. Having a target response in REW and being able to adjust that up or down gives you an instant visual indication of how much cut or boost is needed and which makes more sense to use.
To me a mixing desk eq is a 'musical instrument' so different rules than domestic active crossover where the aim is to get the speakers to work at their best. No problem if someone wants a pultec upstream to mess up their replay but that should be in the 'defeatable' controls part of the system no something that's always there. But the joy of DIY is everyone can pick their own poison!
Bill, it's fun you point about Pultec because most people doesn't realise it was released as a 'broadcast' processor who first use was to eq broadcasted ( radio) signals ( thus considered at the time as a 'technical tool' rather than a creative one to be used on tracking or mixing (or even mastering as there is some settings on the unit which make it sound 'magical' with almost every source material coming through it!).
Anyway my point wasn't to promote the use of a reference rather than another ( especially not a passive design with tube based makeup gain amplifier cell / with transformers and feedback! Such a sin with audiophile ).
More to point to real world example which illustrate my previous statement ( as i can't find theorical proof of what i said atm, but i know they exist) that one can try by him/herself and see his own conclusion.
Iow try it then see for yourself if it makes a difference to you or not.
The situation i described is overly simplified as to be fair other parameters should have been took into account too ( analog eq doesn't have symmetric overall shape for boost and cut especially parametric eq, is an inductor saturating and brings harmonic distortion into account,...) but it resume my thougth about it. I could happily been corrected if my observations is from something different happening.
Anyway my point wasn't to promote the use of a reference rather than another ( especially not a passive design with tube based makeup gain amplifier cell / with transformers and feedback! Such a sin with audiophile
). More to point to real world example which illustrate my previous statement ( as i can't find theorical proof of what i said atm, but i know they exist) that one can try by him/herself and see his own conclusion.
Iow try it then see for yourself if it makes a difference to you or not.
The situation i described is overly simplified as to be fair other parameters should have been took into account too ( analog eq doesn't have symmetric overall shape for boost and cut especially parametric eq, is an inductor saturating and brings harmonic distortion into account,...) but it resume my thougth about it. I could happily been corrected if my observations is from something different happening.
Last edited:
I share this view strongly.
A mixing/mastering studio desk is a very different animal than a loudspeaker management dsp, imo.
It's easy for me to see that studio desks can be built to be different and sound different...as part of their function is to be a creative tool.
But in my mind, speaker processors should all do the same thing and sound the same way.
Imho, their purpose is to simply turn a speaker into a linear playback device, and then have their settings locked up.
Ideally, there should be no variations in filter implementations between manufacturers.
i know this does not exist today, but variations appear to be diminishing as speaker processors evolve (at least in the prosound world).
I think we can all agree that high-Q PEQs mess up sound, whether from a studio screwup or us tinkering with tone adjustment EQ's on our systems.
It seems to the one area of phase audibility folks can concur on...
Naturally, the LX521 often pops up in crossover related debates.
At least 2 of the owners of the horn loaded OB system(s) - which I previously referred to, also own(ed) LX521s. This is one of the comments:
"I thought I'd chime in even though it's a little of topic with a little insight on the BD Designs vs. Linkwitz Labs. I'm friends with the original principles of Audio Artistry which included Sigried Linkwitz (one of the other two was the best man at my wedding).
I own the Beethoven Grands which were their pinnacle of open baffle(dipole) design, I now have BD Orphean horns coupled with 3 high sensitivity 15" drivers per channel, my DIY version of the Orelino 2's minus the refinement of the slick Fletcher Munson compensation curve selection and the chip amps (so read that as not quite as good but in the same family). To me frequency response and tonality is quite similar to the Linkwitz designs, but there is a dramatic difference in information retrieval and dynamics. I can't imagine going back to the Beethovens on a regular basis although I still enjoy them and they throw a huge soundstage.
I never liked the sound of horns, just like some of the other owners here. However, you would never know you're listening to BD horns without seeing them!"
Note that what I'm about to say is my own speculation and not something I have any experience with
I personally have always had a bit of an issue with the idea of out of band flattening to allow a standard electrical order filter to produce the desired acoustic rolloff.
The reason being is that I can't imagine that getting a driver flat (especially if that means fighting it's natural rolloff) can be done without some compromises. Removing peaks like breakup peaks I have no issue with, but bringing levels up to flat is a different kettle of fish.
Maybe it is not an issue because you are attenuating it anyway and it is a means to an end, but something just does not sit well with me with this approach, possibly because I don't understand what is actually happening.
Tony.
Hi Tony, you raise a lot of good points, things i had to struggle with too before coming to fully accept the out-of-band flattening technique.
The biggest question for me was, 'are natural rolloffs minimum phase'.
Because if not, the whole idea is bogus.
After learning that they predominantly are, the next question was 'but is it really safe to raise levels so much out of band, even considering xover attenuation will be in place'.
I made a bunch of electrical measurements after both flattening and xovers were applied, and saw all was well...and most importantly matched the 'nudge method' for achieving a given acoustic target.
(I'd like to stop and give credit to POS (of rephase) for his posts that got me started with the flattening technique.)
The only 'Danger Will Robinson!' is make damn sure the xover is in place
Anyway, the biggest issue i've found with the flattening technique, is it can be dang difficult to do for a full octave or two past xover frequency.
It can take a series of shelving filters gradually applied.
Which is another reason I'm fond of steep xovers....they make the out-of-band flattening region much narrower and easy to achieve.
Anyway my point wasn't to promote the use of a reference rather than another ( especially not a passive design with tube based makeup gain amplifier cell / with transformers and feedback! Such a sin with audiophile
More to point to real world example which illustrate my previous statement ( as i can't find theorical proof of what i said atm, but i know they exist) that one can try by him/herself and see his own conclusion.
I think we are on the same page and I think more people should have switchable EQ at home as long as it can be bypassed. How to work out how to get the crossover and speaker correction as neutral as possible is interesting. There are as you can see many views, all of them valid in their reference frame.
Lucky guy to be able to compare the two. It should be remembered that SL was a classical music lover first with his reference being the SF symphony so he designed speakers with imaging being a prime driver. They were not (to my knowledge) optimised for all types of rendition...