Prompted by some discussion elsewhere, would it be a good idea if we include active crossover hardware equivalent to "standard range drivers" as part of the project? If people want to use what they already have or want to go with expensive packaged solutions like hypex, mini-dsp or equivalent then making the coefficients/transfer functions available would support this. It was more for people that are unfamiliar with active crossovers and wishing to get going with a modest cost approach.
For example, a raspberry pi with an 8 channel DAC hat would cost around £100 (plus bits and bobs). This is off the top of my head and may not be an optimum solution given the wide range of systems on a chip, multichannel DACs and the like that are around. Thoughts and suggestions?
For example, a raspberry pi with an 8 channel DAC hat would cost around £100 (plus bits and bobs). This is off the top of my head and may not be an optimum solution given the wide range of systems on a chip, multichannel DACs and the like that are around. Thoughts and suggestions?
It is not too early to start thinking about DSP options. There are recording engineers and other professionals on this forum who have informed opinions about various DAC and DSP options, and the performance implications of selecting one technology over another.
Does this mean you think it is a good idea to include specific modest cost active crossover hardware as part of the speaker design?
What about amplifier channels? Should we treat those like amplifiers for passive speakers and not include in the design? Personally I would want amplifier hardware to be separate or at least separable from the speakers for flexibility and possible thermal considerations.
It seems popular though I have never used it. There is plenty of software around for convolution and DSP in general. Indeed writing a plugin for a particular audio system is likely to be only a few lines of code but is hopefully avoidable.
I guess the first thing to sort out is the platform. Is the raspberry pi 5 a good solution for modest cost multichannel audio of the kind we seek? I have never played with one and am vaguely aware of some criticism of the audio side of things a few years ago. Not paid any attention since. It has a lot of users and hardware which is attractive. Is anyone aware of a reason not to use a raspberry pi 5? Or know of a better option at a similarish price or perhaps a bit more but not a lot more?
The measurements here suggest the raspberry pi 5 plus 8 channel hat have no significant issues but not read the thread yet though. Not state-of-the-art numbers obviously but nothing that looks likely to be audible in use. Higher spec external DACs can be used instead if desired.
Looks like an option to short list and explore a bit further. Others?
Looks like an option to short list and explore a bit further. Others?
I have successufuly used adau1452 DSP with cs42448 codec "hat" baught on Ali at about 50eu.
3 stéréo inputs, 8 chanel outputs, spdif and toslink input.
Not a highend output stuff but ok. There is a thread around here where some have tweek it to have better thd didn't have the time to dig this but i wanted to make a second hat with better dac chip.
Raspberry would have the possibility to make a steamer integrated.
Never worked with camilla dsp so I can't compare to a dsp.with sigma studio (wich has clearly some bugs but when dsp is programmed it is very stable)
3 stéréo inputs, 8 chanel outputs, spdif and toslink input.
Not a highend output stuff but ok. There is a thread around here where some have tweek it to have better thd didn't have the time to dig this but i wanted to make a second hat with better dac chip.
Raspberry would have the possibility to make a steamer integrated.
Never worked with camilla dsp so I can't compare to a dsp.with sigma studio (wich has clearly some bugs but when dsp is programmed it is very stable)
If you look a bit further in that thread and some others you will find that most are not that happy with the connectors and physical sturdiness of that Hifiberry Hat. I don't know of any other 8 channel hats, most use pro interfaces. Maybe someone else will make one now the Pi5 can do 8 channel i2s output. This is still something of a black hole where there is plenty of cheap and ordinary or expensive and high quality but not so much in the middle.
The cheapest 8 channel pro ish interface is the ESI Gigaport Ex, or Behringer UMC1820, there are a few others before the price and quality jumps up to the Audient EVO 16 or MOTU Ultralte Mk5 which are very capable interfaces that can do an awful lot, but don't look very hifi in the equipment rack.
When it comes to the electronics I think highend should be optional for those that value it and are happy to pay for it. The requirement to be something more like audibly neutral in use. But cheap audio hardware that could/should be audibly neutral in use sometimes isn't.
A widely used CPU like an ARM is likely to be more flexible, modular and easier to use than a DSP processor. Unless something pushes in the direction of a DSP processor like insufficient performance (I haven't checked) I think a CPU is perhaps a better place to start but I am not speaking from experience of current hardware and solutions. I will also confess to wanting to mess about with something like a raspberry pi and it's eco system though not necessarily a raspberry pi.
Is sigma studio cross platform or windows only? How open/closed/proprietary is the software required to work with the DSP processor?
Thanks, that is useful information to put in the against column.
The price of the cheapest interfaces with a CPU like a raspberry pi is perhaps still OK. Which raises the question what is a reasonable budget? What is a reasonable budget for a passive crossover?
Not quite sur to understand. Studio sigma is a proprietary but free tool to make your "program" (I would say a GUI plus a compiler.) once the dsp is flashed, you dont need it anymore.
I was talking about this because 2x chiper than a raspberry solution.
When talking about active digital crossover, i think that hardware implémentation is a complete other project than a speaker. All you need is a vituix project with digital filters you then export biquads value and use Ir in any platform you want...
You wanted chiper, I gave you chiper !
Pretty much any hardware processor loses out if there is a desire to use filtering that needs a lot of taps. Besides that almost any DSP has enough horsepower for run of the mill crossover, peak and shelf filters. One benefit of using FIR capable hardware is that the filters can used as is by anyone as they are self describing. Not all manufacturers use the same Q definition on their IIR filter algorithm's and this can make transferring the filters hit and miss. The Raspberry Pi is a good platform as it is easy to get up and running and is a known quantity. Lots of other boards have great potential but are let down at the software level with incompatibilities. CamillaDSP is great software and is realatively easy to get running by following mdsimon's tutorials at ASR or using it from within Moode. It will still be too much command line for some people though.
Windows only Native Windows or VM needed to use it properly.
For anyone on a budget good pro interfaces can be found cheaply on the used market and for the really low budget you can give this a go
https://diyelectromusic.com/2024/05/27/rpi-5-quad-stereo-sound-with-pcm5102a/
Good passive crossovers for 3 way designs aren't normally cheap, a lot depends on how well behaved the drivers are in their cabinet and whether they are 4 or 8 ohm for the capacitor values needed.
Major costs come from the quantity of copper>power required>coil wire gauge that is approximately 50$ for a 5 mH 1.2mm
But in a vision to be played with a x-amp, all the pages about "digital" are meaningless, other than the natural disposition of a driver to be played individually and when integrated into a system, like Andy remarked a post ago
But in a vision to be played with a x-amp, all the pages about "digital" are meaningless, other than the natural disposition of a driver to be played individually and when integrated into a system, like Andy remarked a post ago
Oh, sorry! That was the budget thread ( not checked lately...err...today) but that was leaning towards a prefabricated baffle housing the holes for the mids and tweets, but no, just for them, not the woofer. Which was aimed at also here, no?
Sigma Studio Auda and most of those cheap interfaces (at least the ones i know) don't do it for me, it still sounds digital because they are not that low noise high quality on the conversions. I work with dsp's since the late 90's for pro audio so i know what is arround and so, but for hifi I need the level of MiniDSP Flex (that is not cheap, i know). I know many think like that, so keep that in mind. If you limit it to a lower rated dsp platform like the Auda, it will turn of a part of the builders i'm affraid.
I would at least also just supply the curves you need with dsp, and let people choose their own dsp platform, maybe with a ready made solution also, but also the raw data to do it yourself with a higher end dsp than what you suggest. But making it fit for dsp also, is certainly what i would do also.
I would at least also just supply the curves you need with dsp, and let people choose their own dsp platform, maybe with a ready made solution also, but also the raw data to do it yourself with a higher end dsp than what you suggest. But making it fit for dsp also, is certainly what i would do also.
Yes that is normally how it would be done given people familiar with active crossovers are likely to already have hardware and know how they want to go about things. The suggestion to include a hardware design is for people that want to build the speakers but are not familiar with active crossovers and want a set of plans to follow that costs something like a passive crossover but delivers the advantages of an active crossover. This sets aside costs associated with amplifier channels.
Cheapness is definitely good but so is flexibility, ease of use, software that runs on non-Windows computers, non proprietary software/hardware for future fiddling about, reliability,...
Yes this might be a concern depending on how we go about things.
I would hope to go with something a bit more interesting than a handful of biquads but it is a fallback.
Many PCs are an option to some extent particularly with external hardware. The attraction of raspberry pi hardware (or equivalent) is that specific new hardware can be specified as part of a design to deliver a known performance. Obviously this rests on the performance and other relevant attributes being sufficient which has yet to be determined.
If a DIYer new to active crossovers baulks at following a few instructions on typing a few numbers into config files and/or running a few commands in a terminal in order set things up then I am comfortable with the project not being suitable for them. On the other hand, requiring a DIY speaker builder to author some code is likely too much in my view.
Indeed but an aggressively low budget is not part of the theme of the project. We are seeking to maximise technical performance using standard range drivers rather than budget range drivers or prestige range drivers. I think the active crossover should fit with this. Decent prosumer hardware would be an example though not necessarily in a prosumer package.
Yes the idea was that this might set a reasonable budget for an active crossover. If achieved the higher performance, ease of tweaking, future use in other projects, etc... may widen the appeal of the project. It is not unreasonable to look at the price of hypex plate amps, mini-dsp hardware and conclude that active crossovers are expensive.
There is no intention to dissuade people from using whatever active crossover hardware they want particularly if they already own it making it a zero cost option. The question is about whether or not to provide a ready made "standard range driver" recipe for those new to active crossovers. If the cost is around that of a passive crossover and delivers the advantages of an active crossover it is likely to be an easy decision for DIYer considering whether or not to go with an active crossover. This is likely to be less the case for a solution based on a MiniDSP Flex because the cost is significantly higher than that of a passive crossover. Would the advantages over a passive crossover be worth the higher cost?
Of course the recipe needs to be audibly neutral in use, reasonably reliable and open to use with other audio hardware. This has yet to be established.
The intention was to make the method of deriving coefficients part of the design. A fixed set of coefficients isn't going to correct for driver-to-driver variations, integrate with distributed subs, include possible placement corrections, etc... but having one or two sets to get going might be a good idea.
A 2nd order filter that uses good components costs about 120€ (per speaker). Add 15/20€ circa per notch needed.
And for good components I mean:
- Iron core and NP cap on the W low pass
- A big jantzen cross cap (MKP 400V) / Iron core coil on the HP of the mid (the big cap on the mid is the most espensive part, 30-35€)
- Air core and jantzen cross cap on the LP of the mid
- Jantzen cross cap and air core coil on the HP of the tweeter
- Mox resistors
For people like me who would like their own dsp, it would be good to share the raw measurments of the drivers in the box next to the coefficients. If i would not receive them, i would remeasure the speaker myself and work from there. Not all dsp's work the same. If you provide the clean measurements that you made yourself, it's probally gonna be better measurements than most would make, so bigger chance on success. It's no extra cost and little extra work for you to do that i think.
But making a ready made receipe is not bad at all, just don't limit the options to that.
Unless somenoe can provide some solid information that a raspberry pi with dac hat isn't going to provide adequate performance I think I will have a play. A little bit of reading around has been inconclusive. Some seem to get close to spec in use whereas others have found issues involving things like noise and robustness.
On a similar theme, should we include a microphone hardware recipe for builders that don't have a microphone and would have little use for one outside setting up and maintaining the speaker system? There's a fair chance such people won't be keen to spend significant sums on a pro sound card and decent calibrated measurement mic to sit in a cupboard for long periods. Given the presence of the raspberry pi is anyone aware of an adequate and modest cost solution based around the hardware?
On a similar theme, should we include a microphone hardware recipe for builders that don't have a microphone and would have little use for one outside setting up and maintaining the speaker system? There's a fair chance such people won't be keen to spend significant sums on a pro sound card and decent calibrated measurement mic to sit in a cupboard for long periods. Given the presence of the raspberry pi is anyone aware of an adequate and modest cost solution based around the hardware?
Last edited:
The ECM8000 sells for less than €25. While unreliable above 5kHz, it’s just fine for EQ-ing. Someone has to come up with a cheap mic pre plus phantom power. No rocket science I think and apart from the casing the parts shouldn’t cost more than €20. Rod Elliott has published quite a few options, including a DIY microphone, we only should source the mic capsules somewhere.
Yes adequate hardware along these lines is likely doable at modest cost but it leaves the less than straightforward task of how to perform a reasonable DIY calibration for little to no cost? Using the manufacturer's published tweeter response would likely get within a few dB but it would be good to come up with something more reliable. Thoughts?
To my humble opinion you’d only have to be certain in the range of the crossover. Tweeter response above that is predefined, changes one would make for taste could be evaluated with any reasonable microphone. And as a matter of fact the DIY calibration with a known tweeter is pretty reliable as long as you use a baffle.
Everybody is free to invest a further €50 for a Sonarworks mike or a separately performed ‘calibration’. Us DIY diehards do such things, but we are exceptions.
Everybody is free to invest a further €50 for a Sonarworks mike or a separately performed ‘calibration’. Us DIY diehards do such things, but we are exceptions.