Well yes I think. Or maybe I'll try to use even just line splitter (or will not using xover kill the twitter?)
I'm a newbie in diy, just curious how to make less work, in best case just buy & connect all parts (if there is a hope that sound quality will not degrade)
If the result will not work good then i can fallback to the single amp and making original xover.
Last edited:
Line splitter will not have the same fr shaping as laboriously optimized passive filter. If you read the thread you know, somebody did a lot of work to optimize that passive crossover. Perhaps thinking that line splitter will exactly match it is doable, perhaps optimistic. If however, like you said, will create better sounding speakers, which you or us will never know, since you will not build two versions,
I have a suggestion for your new triamp speaker name: open source monkey pox.
I have a suggestion for your new triamp speaker name: open source monkey pox.
Why it is not so easy to convert a passive speaker into an active speaker: http://www.troelsgravesen.dk/passive-to-active.htm
If you do not understand this article, read it again.
If you do not understand this article, read it again.
That Troels article is besides the point. As I wrote before, an active filter would have to replicate the filter transfer functions of the passive filters in the OSMC. That's certainly possible with a DSP, and we actually did exactly that during the development of the OSMC.
However, all the DSP processing seems to detoriate the sound quality quite a bit in my experience. Maybe a super-duper high-end (and expensive) DSP is better than the miniDSP in this regard, but I don't think you'd save a lot of money and trouble by going this way. I'd therefore recommend to stick to the passive filters. Also, I'd prefer to use one good amp instead of three mediocre amps.
However, all the DSP processing seems to detoriate the sound quality quite a bit in my experience. Maybe a super-duper high-end (and expensive) DSP is better than the miniDSP in this regard, but I don't think you'd save a lot of money and trouble by going this way. I'd therefore recommend to stick to the passive filters. Also, I'd prefer to use one good amp instead of three mediocre amps.
The answer to this question is very simple: Stay with the ready-made solution. Here is why: the measured response (links of which you can find to in post #1) is a very good performance and to replicate something comparable with a DSP is a lot of work. You will have to make high quality measurements of all individual drivers mounted in the enclosure, over many angles. Not to speak about the tuning of the filters later on. The enclosure is big, and measuring a full spin is not only technically challenging, but physical work. You would probably want to build a measuring rig for this.
The frequency response could probably be improved with a DSP, as to my experience, beyond 1k even small changes effect the tonality of a speaker more than one would expect. But this is not time and work saving, quite the opposite: Searching for perfection. This is super laborious. And with Matthias’ crossover you will have a very sophisticated result already at hand by following his plans.
In my experience, it is usually rather the amps of an active solution that deteriorate the performance because of hiss. Active is often not as dark, and the Faital woofer is very efficient, which means it would transfer the hiss. If you use a miniDSP HD with an external amplifier, the gains in control over linear distortion could be interesting. But a high quality DSP is expensive and the work involved probably a bit over the top if you are just starting. If you have the means to build the passive version, my recommendation is to stay passive. You could even consider to have someone solder you the crossover for a loan and still where making yourself a favor.
Thank you (and all) for answers! Looked on dbx driveracks with software equalizer/dsp, seems it could do the job for a reasonable price, but yes seems going with original design could be better. One of the points of the idea I didn't find a good amp yet which could give many details in hi/mid like a-class/t-class and drive bass like d-class. While very good in separate freq amps are cheap. And seems amp which can combine both in one will be much more expansive. Wanted to achive two goals in one shot, but seem this is not so easy.
Sound quality is not just a question of the amplifier(s) alone. It's a question of the amplifier/speaker combination.
The whole idea of the OSMC is that you don't need a muscle amp to control it. I use either my USSA5 or my AlephJ amp with the OSMC. The result is exactly what you you are looking for: sweet Class-A and good bass
Hi,
As an information, for each passive or DSP loudspeaker crossover design, it is possible to realize an equivalent analog active filter with opamps (or tubes if wanted) in an easy way, if the voltage transfers of the existing filter are available.
You need a tool for it. In my case, I use Leap. For an active analog or digital crossover filter, you can make a design with analog functional blocks first. And then you transform that schematic to an analog active filter with opamp blocks with the calculated filter components (resistors and capacitors) or to a digital version with biquads.
So, to make a 1:1 analog active copy of an existing filter, you just need to design the schematic with analog functional blocks having the same voltage transfers and transform it to the opamp blocks.
For the Monkey box, I made the first proposal of the elliptical filter and also the digital equivalent of it with biquads for the first test of the crossover. So, I have the schematic with analog functional blocks almost ready, just need to make a little update to the last voltage transfers.
In this way, the analog active filter version is a perfect 1:1 copy of the passive filter, as the voltage transfers are exactly the same (within tenths of a dB).
As an information, for each passive or DSP loudspeaker crossover design, it is possible to realize an equivalent analog active filter with opamps (or tubes if wanted) in an easy way, if the voltage transfers of the existing filter are available.
You need a tool for it. In my case, I use Leap. For an active analog or digital crossover filter, you can make a design with analog functional blocks first. And then you transform that schematic to an analog active filter with opamp blocks with the calculated filter components (resistors and capacitors) or to a digital version with biquads.
So, to make a 1:1 analog active copy of an existing filter, you just need to design the schematic with analog functional blocks having the same voltage transfers and transform it to the opamp blocks.
For the Monkey box, I made the first proposal of the elliptical filter and also the digital equivalent of it with biquads for the first test of the crossover. So, I have the schematic with analog functional blocks almost ready, just need to make a little update to the last voltage transfers.
In this way, the analog active filter version is a perfect 1:1 copy of the passive filter, as the voltage transfers are exactly the same (within tenths of a dB).
I recently sold a pair of those Volt midranges. After buying them to make a Tetra 606 clone, and testing them myself I got results similar to those in this post. See the not so flat frequency response. I recall that the ATC 3 inch domes produce a very flat response. I suspect that something like a reflection off the pole piece that causes the peak & dip pair at around 3 KHz in the Volt dome. The Tetra 606 mockup didn't please the customer, so the drivers had to go. Certainly a FIR filter implemented in the miniDSP 2x4HD could be used to tame this reflection response, but for the money I decided to let them go. So there's no doubt in my mind that a DSP based crossover would greatly benefit this design. It is very useful in cleaning up the rough response of drivers like this dome before the crossover is applied.
Last edited:
I am not quite sure what you mean by "rough response" or "reflection response". I can't see anything unexpected with the Volt response measured in the OSMC baffle (see Fig. 2 here: https://github.com/mbrennwa/osmcdoc/blob/master/osmc_paper.pdf ).
Also, my ATC 75-150 drivers are very similar to the Volts except that their SPL curve extends a bit higher up (possibly due to the different shape of the eaveguide).
The transfer function of the DSP filters you are referring to are identical to those of the OSMC analog/passive filters, so your point about the need for DSPs is moot.
As you can see from my post I was referring to the plot in post 982 that showed a graph of the midrange response that looked rough. See it here: https://www.diyaudio.com/community/goto/post?id=6130046 . Other midrange drivers I have used had less variation in the response I've measured. An example would be the Aurasound 2" Whisper driver. I made no reference to your cross over filters when I talked about using an FIR filter to correct the response of the midrange, so I'm not sure how that is moot. As an example I created such a filter myself for the Volt driver. The figure below shows my own measurement of the Volt midrange and the response after applying DSP filtering to smooth the response and apply band limiting. I think that illustrates my point pretty well. I attached octave .m (saved as a .txt ) file used to generate the compensation FIR filter. The target response can be changed easily by editing the cutoff frequencies in the script should anyone want to use it. To run you would need to provide your own impulse response measurements for the system input to output loop back and the driver impulse response measurement as a .wav file. I think there are sufficient comments in the code to help someone get it to work. The calculations work but there is no error checking to make it robust. I was unable to attach example .wav files.
Ok, I believe I understand what you did, makes sense. However one needs to keep in mind that the OSMC was not about a "perfect" on-axis SPL curve. The off-axis response is just as important for what we hear. The diffraction effects need to be factored in by looking at the dispersion pattern and the power response. Correcting diffraction effects in the on-axis response is usually not a good idea. (See also the OSMC paper.)
Not to put too fine a point on it, but at the upper end of the frequency range covered by the driver, it is becoming directional such that it isn't illuminating the edges of the cabinet with those frequencies, so there should be little or no diffraction effects for that range. Cabinet diffractions happen at the lower end of the range of each driver in a system when they become close to omnidirectional. Placing a microphone right at the cabinet edge and taking a measurement will of course confirm of deny this for any given design as the response there dips where the drivers become directional. Comparing the driver response with it mounted in a very wide test baffle vs the smaller finished cabinet will also show the effect. You are right that basing the equalization applied based on a single measurement isn't optimal. Averaging the response from several measurements over a small listening window is best. My simple script doesn't do that as most commercial software does. For a forward controlled dispersion design, I don't worry too much about the spectrum over a wider dispersion as sound reflected off room surfaces will have large irregularities depending on the surface properties and angles of the reflecting surfaces. Of course for an omnidirectional or very wide dispersion design an even response at all frequencies is more important. By the way, very nice work on the speaker design and cabinetry. My cabinet building skills are privative and the results often look like I used and axe. It's great when people go to the trouble to document and share designs and measurements for others to use.
Hello folks,
I 'd like to run some sims as well to sketch a project with the Faital. Many thanks guys for this thread and sharing your datas.
I am a bit lost and dunno which T&S datas to pick up. Paul's with Fs around .043 and Qts 0,42 ? Zvu's (driver with no burn in ?).
Reading the thread I saw the Fb was little by little adapted by ear and I do like the smooth responss to mimic a sealed enclosure response with a little more spl in the early roll off towards the low. Well done.
Is the BR load close to a Bessel with the Fb at 35 hz and circa 67 L ? Has anyone measured again the 12PR320 after few month of ab-use ? Would like to sim it in a sligthy less width circa 33,5 cm width and heigth front baffle with a recessed stand a la loudspeaker from Sueeskind to see how looks like the baflle step loss and where to low pass it (400 hznor ?).
Sorry if too much off topic, maybe stands in the OSMT thread.
Thanks
Last edited: