We do. It absolutely can be done properly without a calibrated mic. The vast majority of the important work is relative, not absolute.
This prompted a bit of thought about the project given to me someone messing around with DSP controlled speakers as a hobby not being too bothered about what is going on objectively seemed odd given the objective to maximize technical performance using standard range drivers (and equivalents for other components). The speed and ease of making changes with DSP would almost certainly lead to someone interested in speakers wanting to use a reasonably accurate mic a fair bit more than with equivalent passive speakers which are a lot more expensive and time consuming to play around with.
Given the successfully completed expensive monkey coffin project, the good progress on the budget version and my wish to contribute rather than lead this standard driver version I am wondering if dropping the monkey coffin aesthetic as an objective might be worth considering. Keep a budget based on standard drivers and equivalent hardware, keep maximizing technical performance for the home, keep modular, keep 3 way mains just be more flexible about the form factor. This would increase my interest in leading and raise the build from number 3 or 4 to number 1 or 2 on my to-do list (I impulsively gave away my desktop speakers). But would that be of any significant interest as a group project? It would seem the monkey coffin form factor is of interest to a fair few and that some factors contributing to maximizing technical performance like DSP might be somewhat less so. Just a thought.
It is an interesting concept.
So the idea is a group project to design the best possible 3-way speaker, given a driver budget of $xyz. Further constraints are that it be modular and active DSP controlled. The form factor is not specified at the outset, but would be determined during the design process based on performance, construction feasibility, cost, and aesthetics.
Earlier in the thread, you suggested a driver budget of 500 - 1000 GBP (approximately $600 - $1200 USD) per pair. Is this still the range you are considering?
I think it is a great project. I vote yes.
So the idea is a group project to design the best possible 3-way speaker, given a driver budget of $xyz. Further constraints are that it be modular and active DSP controlled. The form factor is not specified at the outset, but would be determined during the design process based on performance, construction feasibility, cost, and aesthetics.
Earlier in the thread, you suggested a driver budget of 500 - 1000 GBP (approximately $600 - $1200 USD) per pair. Is this still the range you are considering?
I think it is a great project. I vote yes.
That would make it even more interesting for me - and my interest in (melamine-foam) absorptive baffle.
In my simplified understanding an (semi-) absorptive baffle could be a good compromise between wide and narrow baffle.
The modular idea might even allow gradual development to find one's optimum. Maxbe even an experimental cardioid midrange enclosure?
I would suggest 3 way mains rather than 3 way speakers given the likely role of distributed subs in achieving a high technical performance over the lowest couple of octaves in a home.
The budget would depend on the solution adopted. It is likely that controlling the sound radiation pattern down to 100 Hz or so will have a significant positive influence in raising in room performance in a home. If adopted this would mean sound cancellation of some form raising the required woofer cone area considerably if clean SPL at standard levels is still to be met. This might perhaps double the cost of the woofers.
Standard range drivers rather than budget or prestige ranges. A driver budget of roughly £500-1000 but with a contigency fund to double the woofer costs if we opt to control the low frequency radiation pattern. We might come up with one or two other things that could raise performance if the budget was raised and so long as it involves standard range hardware rather than prestige I would suggest a case should be considered.
I prefer a modular design in order to prototype and support a DIY hobby. It can be dropped to a fair extent if a builder is not interested.
Active crossovers are not a constraint given their use would tend to follow from seeking a high technical performance and being modular (if not a plate amp that's part of the rear panel). However some active hardware from hypex, mini-dsp (not all) is likely excluded because it is more prestige than standard. Which doesn't mean don't use it if you already have it or want to use it but that the design will seek something more inline with the budget.
Yes. This would involve a degree of writing down the options, performing simulations, searches and perhaps the odd measurement to gather information, assessing the pros and cons, pencilling in decisions,...
That's good.
It is certainly something to consider though the rule of thumb requirement for something like a quarter of a wavelength for absorption to be effective will be a limitation.
I would agree that both active and passive cardioid enclosures for the woofer and midrange should be assessed. 0D/1D simulation and looking at the performance of existing cardioid speakers is perhaps an effective way to work towards a ballpark design. Detailed 3D simulation and experiments to refine the design/s if things look promising.
There is very little available research material that supports this view and a decent amount of anecdotal evidence that contradicts it.
Doing this can ensure the control of peaks and dips in the lower midrange and upper bass.
Otherwise with multiple early reflections, it's difficult to get a smooth response by either absorption or positioning. It's much harder to find a complementary power response at the same time without significant compromise.
Anyone who has successfully designed a larger baffle speaker may appreciate this. While some point out that we are less sensitive to the diffraction aspect, and this is true, the problem is worth fixing and it's not something you can manage with a woollen tapestry on the walls.
I personally am somewhat happy that this project has loosened some unnecessary (for some people at least) constraints on speaker form factor and would rather focus on a 'performance-optimized' 3way speaker.
But I would still suggest not getting blindly attracted to various loudspeaker designs just based on the aspect of pattern control down to 100 Hz. For example, here are the actual measurements-based pattern control details of one of my own speakers.
The above crossover was DSP-based using all IIR filters.
Here is a VituixCAD generated, first reflections-based room impact on the PIR (predicted in-room response)
The above kind of PIR is almost exactly what I measured in my room based on nearby boundary locations & listening distance.
Now, I don't know/was not able to isolate the exact impact of that orange curve in real-world listening sessions (given the other many other sub-optimalities in my listening set up).. 🙂 .
It is also worth asking what is the impact of the excess group delay needed to achieve the above pattern and whether it is tolerable given that there are some studies that say it is best to have the excess group delay around 100Hz to less than 2.1ms. For example, in the above speaker, the below is what happened because of several "choices" I made during the design stage by just blindly focusing on pattern control.
If one has access to FIR filters with a large number of taps, at least the excess group delay problem might go away to some extent, but questions surrounding how much pattern control is enough down to 100Hz are very well worth studying. For example, in the above speaker, the best attenuation I could get beyond 120 degrees or so is in the range of 10 to 12 dB.
Maybe things would have much been better with 20dB attenuation beyond 100 degrees or so and with careful positioning (if there's flexibility) and at least some strategies to combat floor and ceiling reflections.
So, if one is starting from scratch trying to optimize a 3-way speaker to get the best performance, my suggestion would be to start with a study of common room geometries intended for final placement and identify their impact for a given polar pattern shape (say down to 100Hz). Then, keeping the associated restrictions/constraints in mind, design the speaker ground up trying to address at least some/most of these aspects..
In summary, all I am saying is it would be easy to get distracted by a nice-looking polar pattern in anechoic conditions and skip studying the room impact aspect and then realize later that it would have been better to consider that first 🙂
But I would still suggest not getting blindly attracted to various loudspeaker designs just based on the aspect of pattern control down to 100 Hz. For example, here are the actual measurements-based pattern control details of one of my own speakers.
The above crossover was DSP-based using all IIR filters.
Here is a VituixCAD generated, first reflections-based room impact on the PIR (predicted in-room response)
The above kind of PIR is almost exactly what I measured in my room based on nearby boundary locations & listening distance.
Now, I don't know/was not able to isolate the exact impact of that orange curve in real-world listening sessions (given the other many other sub-optimalities in my listening set up).. 🙂 .
It is also worth asking what is the impact of the excess group delay needed to achieve the above pattern and whether it is tolerable given that there are some studies that say it is best to have the excess group delay around 100Hz to less than 2.1ms. For example, in the above speaker, the below is what happened because of several "choices" I made during the design stage by just blindly focusing on pattern control.
If one has access to FIR filters with a large number of taps, at least the excess group delay problem might go away to some extent, but questions surrounding how much pattern control is enough down to 100Hz are very well worth studying. For example, in the above speaker, the best attenuation I could get beyond 120 degrees or so is in the range of 10 to 12 dB.
Maybe things would have much been better with 20dB attenuation beyond 100 degrees or so and with careful positioning (if there's flexibility) and at least some strategies to combat floor and ceiling reflections.
So, if one is starting from scratch trying to optimize a 3-way speaker to get the best performance, my suggestion would be to start with a study of common room geometries intended for final placement and identify their impact for a given polar pattern shape (say down to 100Hz). Then, keeping the associated restrictions/constraints in mind, design the speaker ground up trying to address at least some/most of these aspects..
In summary, all I am saying is it would be easy to get distracted by a nice-looking polar pattern in anechoic conditions and skip studying the room impact aspect and then realize later that it would have been better to consider that first 🙂
Last edited:
Good to see there is some interest in revising the spec and some good suggestions on how to go about things. I guess the next task is to pull together suggestions for what the revised spec should be.
floor standing or monitor?
position relative to front wall:
- against?
- near?
- out in the room?
clean output at standard levels (80-85 dB average) at 3-4m with low frequency extension to:
- 50-60 Hz? - requires a sub system
- 30-40 Hz? - forms part of a sub system or standalone with limitations
- 20 Hz? - is there any case for this with 3 way mains?
well controlled radiation pattern with what shape:
- omni below 80 Hz? - any case for something else?
- omni or narrowed from 80 hz to a few hundred Hz?
- beam width above a few hundred Hz?
- degree of narrowing with increasing frequency?
target audience and degree of complexity of build:
- experienced DIYer?
- inexperienced DIYer that intends future builds?
- inexperienced kit builder?
standard range drivers and components
modular:
- separate woofer and mid/tweeter cabinets?
- external crossover and amplifiers?
- separable crossover and/or amplifiers in own chassis hung off the back?
versions:
- just one?
- wide and narrow beams with different woofer and mid/tweeter modules?
- against wall and out in room woofer module?
- others?
method of controlling radiation pattern:
- waveguides
- coaxial
- delayed rear radiation
- assess pros and cons as a task?
active crossover:
- DSP FIR
- case for IIR?
- case for analogue?
- PC
- Raspberry pi or equivalent?
- DSP boards?
- Video chips?
- GPU?
- DSP with amp channels?
- other hardware?
passive crossover:
- none?
- see what can be done with an unconstrained active design?
- include passive crossovers constraints in design
???
floor standing or monitor?
position relative to front wall:
- against?
- near?
- out in the room?
clean output at standard levels (80-85 dB average) at 3-4m with low frequency extension to:
- 50-60 Hz? - requires a sub system
- 30-40 Hz? - forms part of a sub system or standalone with limitations
- 20 Hz? - is there any case for this with 3 way mains?
well controlled radiation pattern with what shape:
- omni below 80 Hz? - any case for something else?
- omni or narrowed from 80 hz to a few hundred Hz?
- beam width above a few hundred Hz?
- degree of narrowing with increasing frequency?
target audience and degree of complexity of build:
- experienced DIYer?
- inexperienced DIYer that intends future builds?
- inexperienced kit builder?
standard range drivers and components
modular:
- separate woofer and mid/tweeter cabinets?
- external crossover and amplifiers?
- separable crossover and/or amplifiers in own chassis hung off the back?
versions:
- just one?
- wide and narrow beams with different woofer and mid/tweeter modules?
- against wall and out in room woofer module?
- others?
method of controlling radiation pattern:
- waveguides
- coaxial
- delayed rear radiation
- assess pros and cons as a task?
active crossover:
- DSP FIR
- case for IIR?
- case for analogue?
- PC
- Raspberry pi or equivalent?
- DSP boards?
- Video chips?
- GPU?
- DSP with amp channels?
- other hardware?
passive crossover:
- none?
- see what can be done with an unconstrained active design?
- include passive crossovers constraints in design
???
To initiate further thought process and brainstorming along the lines of speakers with good amount of pattern control to lower frequencies and to understand expectations in terms of form factor, here is an example of a speaker with a decent amount of pattern control; here is another one of my speakers having a smaller form factor.
This is a speaker I use for my PC along with a low-cost DSP amp like this:
https://zoudio.com/
This is a 3way speaker with a SICA 5.5inch coax (in passive cardioid configuration) + Wavecor 8.5inch subwoofer in sealed box
Measurements with full 3way DSP active crossover
Impact of first reflections (Please note that the below 1kHz region is of more interest in all these pics since taming the above 1kHz region is relatively easier)
Adding another speaker to form an MTM configuration and possibly changing into a configuration like this with the coax still in passive cardioid mode (picture copied from MrStitcha's project)
This will extend pattern control down to something like below (Please note that I am just indicating possibilities w.r.t form factor and extrapolating results obtainable based on the 3way speaker shown in first pic. In a final speaker, the crossover may be completely different)
Impact of first reflections
Now going a step further and adding side/back drivers one may be able to extend the directivity control to even lower frequencies. An example could be something like the ascilab speakers
It seems to have measurements like this
The speakers shown above also has the advantage of having a modular cabinet, relatively high SPL capabilities very good pattern control to 100Hz etc..
If one is interested in just the top module, a nice speaker can be made out of that as well.
(Since speaker stands are most of the time a waste of space and SPL capability/pattern control capabilities I mostly like floor standing versions).
Maybe we should create another poll regarding SPL requirements, in-room positioning restrictions, form factor restrictions, budget restrictions etc so that one may be able to gauge the interest/opinions and choose design requirements appropriately..
My personal preference is for a relatively small, modular speaker with an add-on bass module, coax mid tweeter/waveguided tweeter, with good pattern control down to 100ish Hz, as long as it helps to some extent with taming the deep nulls that occur in the 200 to 300Hz region. In my personal listening situation, I sit about 2.2m away from speakers, the room is L-shaped, max SPL i hear songs at an average of 80dB SPL from MLP. I prefer having a relatively wide, smooth beamwidth of about 60 degrees or so above 1kHz. I prefer to keep a control on excess group delay to approx less than 2.1ms at 100Hz. I prefer having less clutter in terms of wires running around. So a plateamp with DSP capabilities would help. I prefer having relatively shallow cabinet speakers than deep cabinets etc.
This is a speaker I use for my PC along with a low-cost DSP amp like this:
https://zoudio.com/
This is a 3way speaker with a SICA 5.5inch coax (in passive cardioid configuration) + Wavecor 8.5inch subwoofer in sealed box
Measurements with full 3way DSP active crossover
Impact of first reflections (Please note that the below 1kHz region is of more interest in all these pics since taming the above 1kHz region is relatively easier)
Adding another speaker to form an MTM configuration and possibly changing into a configuration like this with the coax still in passive cardioid mode (picture copied from MrStitcha's project)
This will extend pattern control down to something like below (Please note that I am just indicating possibilities w.r.t form factor and extrapolating results obtainable based on the 3way speaker shown in first pic. In a final speaker, the crossover may be completely different)
Impact of first reflections
Now going a step further and adding side/back drivers one may be able to extend the directivity control to even lower frequencies. An example could be something like the ascilab speakers
It seems to have measurements like this
The speakers shown above also has the advantage of having a modular cabinet, relatively high SPL capabilities very good pattern control to 100Hz etc..
If one is interested in just the top module, a nice speaker can be made out of that as well.
(Since speaker stands are most of the time a waste of space and SPL capability/pattern control capabilities I mostly like floor standing versions).
Maybe we should create another poll regarding SPL requirements, in-room positioning restrictions, form factor restrictions, budget restrictions etc so that one may be able to gauge the interest/opinions and choose design requirements appropriately..
My personal preference is for a relatively small, modular speaker with an add-on bass module, coax mid tweeter/waveguided tweeter, with good pattern control down to 100ish Hz, as long as it helps to some extent with taming the deep nulls that occur in the 200 to 300Hz region. In my personal listening situation, I sit about 2.2m away from speakers, the room is L-shaped, max SPL i hear songs at an average of 80dB SPL from MLP. I prefer having a relatively wide, smooth beamwidth of about 60 degrees or so above 1kHz. I prefer to keep a control on excess group delay to approx less than 2.1ms at 100Hz. I prefer having less clutter in terms of wires running around. So a plateamp with DSP capabilities would help. I prefer having relatively shallow cabinet speakers than deep cabinets etc.
Attachments
Last edited:
I've volunteered myself to do some work on a "standard driver" active crossover using raspberry pi bits and bobs. I have no experience with this type of hardware, have not done much research and so there is a fair chance it will not work sufficiently well. If it doesn't what it can and can't do should be useful to know anyway. Gone for 8 channels initially so stereo active cardioid 3 way should be doable though not with subs.
Raspberry Pi + 6/8 channel DAC is a capable FIR DSP as long as one gets a good enough DAC.
Even in the early days of CamillaDSP on Pi (3-4 years back), I had experimented with full Bandwidth FIR filters (16k taps per channel) and found it to be working fine except for a slight delay (which I think will go away when the FIR filters are properly/appropriately designed) in music playback using Moode on Pi. But at the time I had used an Allo Boss DAC hat for Pi and only 2 channels of FIR.
So, experimenting and identifying the limits of Pi in a full 3-way stereo system might be very useful, I think.
But I doubt that eventually, it will come down to the ease of use/amplification/DAC choices for the active crossover and people's choice of music sources rather than the DSP capabilities of the Pi itself.
Even in the early days of CamillaDSP on Pi (3-4 years back), I had experimented with full Bandwidth FIR filters (16k taps per channel) and found it to be working fine except for a slight delay (which I think will go away when the FIR filters are properly/appropriately designed) in music playback using Moode on Pi. But at the time I had used an Allo Boss DAC hat for Pi and only 2 channels of FIR.
So, experimenting and identifying the limits of Pi in a full 3-way stereo system might be very useful, I think.
But I doubt that eventually, it will come down to the ease of use/amplification/DAC choices for the active crossover and people's choice of music sources rather than the DSP capabilities of the Pi itself.
Here is a hypothetical speaker which is cardioid/cardioidish down to 20Hz . Above 1kHz, directivity is even higher.
This is created using my earlier cardioid SICA coax + a bass module similar to the one in the above above ascilab speaker pic with a woofer facing front and another at the backside 200mm away from the front woofer and facing the wall.
MLP is 2.2m away from speakers. I assume minimal to no room treatment. Speakers are toed-in 10 degrees
scenario-1: I push the speaker closer to front wall
Here is the impact on PIR
We can see the minimal impact of front wall reflections... Looks like cardioid works as advertised 🙂
Scenario-2: Now I activate the front wall (400mm away) + left wall (assuming 1.5m away)
impact on PIR
Looks ok still, I think.. 🙂
Scenario-3: Now I activate the front wall (400mm away) + left wall (assuming 1.5m away) +ceiling (1.7m up from reference axis)
impact on PIR
Maybe even this is ok.. 🙂
Scenario-4: Now I activate the front wall (400mm away) + left wall (assuming 1.5m away) + ceiling (1.7m up from speaker centre) + floor (1m below the speaker reference axis, with maybe a minimal carpet on the floor)
impact on PIR
Does it look ok now?🙂
This is created using my earlier cardioid SICA coax + a bass module similar to the one in the above above ascilab speaker pic with a woofer facing front and another at the backside 200mm away from the front woofer and facing the wall.
MLP is 2.2m away from speakers. I assume minimal to no room treatment. Speakers are toed-in 10 degrees
scenario-1: I push the speaker closer to front wall
Here is the impact on PIR
We can see the minimal impact of front wall reflections... Looks like cardioid works as advertised 🙂
Scenario-2: Now I activate the front wall (400mm away) + left wall (assuming 1.5m away)
impact on PIR
Looks ok still, I think.. 🙂
Scenario-3: Now I activate the front wall (400mm away) + left wall (assuming 1.5m away) +ceiling (1.7m up from reference axis)
impact on PIR
Maybe even this is ok.. 🙂
Scenario-4: Now I activate the front wall (400mm away) + left wall (assuming 1.5m away) + ceiling (1.7m up from speaker centre) + floor (1m below the speaker reference axis, with maybe a minimal carpet on the floor)
impact on PIR
Does it look ok now?🙂
Attachments
Scenario-5: I activate the front wall (400mm away) + left wall (assuming 1.5m away) + ceiling (1.7m up from speaker centre) + floor (1m below the speaker reference axis) + more absorption on the floor, and ceiling
Is this better?
I request those who know better about acoustics & speakers to comment about these aspects... I am a noob in this.
Is this better?
I request those who know better about acoustics & speakers to comment about these aspects... I am a noob in this.
Now you're moving to a very different speaker than intended, so maybe change the title or start a new tread. Now it's confusing i think (without reading the whole tread).
I have changed the speaker overall response slightly
Scenario-6: All major reflections as above activated but the speaker is moved closer to the floor down from 1m to 700mm and a (probably) heavily absorbing carpet on the floor
Looks better than before?
Maybe there is hope with this kind of speaker.. 🙂
Scenario-6: All major reflections as above activated but the speaker is moved closer to the floor down from 1m to 700mm and a (probably) heavily absorbing carpet on the floor
Looks better than before?
Maybe there is hope with this kind of speaker.. 🙂
Attachments
Scenario-7: A 3way traditional dual woofer + (mid + tweeter) in normal 'ish' cabinet setup under below shown conditions
How much better or worse are we compared to the cardioid speaker..? 🙂
How much better or worse are we compared to the cardioid speaker..? 🙂
I dont know if all the above simulations are valid since the room tab in vituixcad looks to show(/model?) the speaker as a single driver with the polar pattern got for the whole speaker. Maybe when sufficiently far away from the speaker (like 2+m away) the simulations and the single sources assumption are still valid for even multiway speakers with relatively well spaced out drivers. I don't know exactly which distrances away from the speaker these assumptions are valid/when they breakdown.
But if they are valid especially for the below 1kHz region, I dont see the cardioid offering significant advantages over a relatively non cardioid speaker..
But if they are valid especially for the below 1kHz region, I dont see the cardioid offering significant advantages over a relatively non cardioid speaker..
The idea was to assess the standard raspberry pi approach because of the large amount of hardware and support available for those that may be interested in speakers but unfamiliar with and not that interested in the electronics (e.g. me). So I bought the current pi and the standard berry 8 channel DAC hat with tiny jack plugs (?). The measured performance by the camilladsp author was about 20 dB down on decent pro hardware but this should be adequate if it doesn't degrade in use due to interference and the like which seems to be fairly common with cheap consumer audio hardware. Using standard pi power supply. Held off buying a complete system with ADC, screen, case, etc... until things check out if they do.
That's encouraging. This was presumably an earlier pi that doesn't get hot in use? The current one does under load and the add-on hardware appears to get in the way of cooling. One of the things that will need checking out.
I too expect the DSP capabiliites of a current pi to be sufficient for the number of channels typical of current 2 or 3 way DIY main speakers. If it proves so I would look to increase the number of channels for a sub system and then onto beam forming in order to get a variable radiation pattern. This will introduce a bunch of complexities but should hopefully keep the budget down at hobbyist levels which using pro audio hardware normally doesn't.
I am not familiar with the software you are using and the modelling assumptions it makes. Is it vituixcad on a windows computer?
The suggestion was to make only a single change to the project objectives by dropping the constraint of a monkey coffin configuration in favour of whatever maximises technical performance. The other objectives have been retained. For me the retro visuals was of modest importance compared to other objectives like maximising technical performance using modestly priced standard hardware. For others the enthusiasm and progress in the budget monkey coffin thread rather suggests the reverse. Indeed I didn't even understand some comments like the visual proportions of the drivers.
I agree the title needs changing to remove misleading words like monkey and to replace it with whatever the name of the project might be. Not that the project ever had a name. Suggestions? The OP needs modifying to remove the monkey coffin configuration and how that affects one or two other objectives. Perhaps it would be better in a new thread?
@andy19191: I had used a Raspberry Pi Model 3B/B+ for some months, at the time I think. I couldn't find any significant heat issues at the time.. Infact, I hadn't explicitly focussed on this issue so it. may even be possible that I have missed noticing it..
Yes I use Vituixcad on a windows PC.
Yes I use Vituixcad on a windows PC.