Amateur wants to develop a speaker based on Purifi PTT4

[Glorfindel]

Hello all, I am Gian and I am new here.

I would like to embark on my first "serious" DIY project. But before I sink a ton of money in a project that may be doomed from the start, I would appreciate some feedback from more experienced people on whether what I have in mind makes any sense. I hope it is fine for a newbie to just arrive on the scene like that and ask for your advice. In return, I would of course be happy to report on the progress and final result of the project.

Maybe two, three words about myself before I explain my plan:
I am from Switzerland and am doing a PhD in electrical engineering at ETH Zurich (not related to anything audio at all, though, so don't overestimate my expertise ). I have built several DIY kits before and even developed my own loudspeaker as my high school diploma project (which, with hindsight, thought me more about how loudspeakers should not be developed), so I am not a complete newbie. All in all, I feel like I am at a place where I can come up with some project plan on how to develop a speaker myself, but I am sure that this plan contains at least a few bad ideas. Any advice that helps me find and avoid these mistakes would be much appreciated.

Okay, let's talk business. I like small speakers (this fact is not up to discussion), and I would like to get the best possible sound from a small box. Budget constraints are secondary, though I wouldn't want stuff to be gratuitously expensive. The ability to play high volumes is also less important, but I want to have good bass extension.

What I had in mind is to build an active two-way speaker with a passive radiator (PR) and a digital crossover in a ~5 liter box. For the woofer, I had my eyes set on one of the 4-inch Purifi drivers (4Ohm or 8Ohm). For the PR, I also thought of the 4-inch PR from Purifi. To achieve good directivity, I had planned to use a 4-inch waveguide from somasonus, which I would have 3D-printed. The availability of 4-inch waveguides would limit the choice of the tweeter to a SB21 or SB26 from SB Acoustics or the CSS LD22. For the crossover, I planned to use a miniDSP flex, and amplification would be provided by something like a 4-channel Buckeye amplifier. This component list makes clear that the whole project would easily land in the $3K ballpark, so not exactly cheap.

My naïve mind would imagine the development to look something like this:
I order a Purifi woofer and PR, a tweeter (my current favorite, without any solid reasons for this, would be the CSS LD22), a miniDSP flex and a UMIK-1 measurement mic. I toy around with VituixCAD to get the enclosure dimensions based on the TSP of the Purifi woofer. I have the waveguide 3D-printed and build a box prototype based on the VituixCAD simulations. I measure the woofer and the tweeter within the box prototype and twiddle with the filters of the miniDSP Flex until I am happy. I order the remainder of the component and build the "camera-ready" speakers.

I am pretty sure that it won't work like that. The problem is that I cannot really tell where reality will deviate significantly from the above outline.
Here are several difficulties of which I am already acutely aware:

• Should I go for the 4Ohm or the 8Ohm variant of the Purifi woofer? What are the relevant criteria here?
• What are the criteria based on which to decide between the three tweeter options? Will any of these tweeters even work well in combination with the Purifi woofer?
• Will some basic TSP-based simulations in VituixCAD really give me a good enough estimate so that I only have to design one enclosure prototype and the bass response will be as expected?
  • Can I in fact just trust the TSP as published by Purifi? The reviews from Erin and from Hifi-Compass (both only for the 4Ohm variant) have measured parameters which do not match perfectly with the manufacturer's datasheet, as far as I understand. Would it make more sense to buy something like this and try to measure the TSP parameters myself? I have already toyed around with some simulations based on the manufacturer-supplied parameters of the 8Ohm woofer and the PR. A screenshot of the result is attached below.
  • What about the design of the front? How wide and high should the front be, where should the drivers be placed on it, and should the front contain any fancy edge designs instead of simple rectangular edges? (I would like to avoid trial-and-error box prototypes, if at all possible.)
  • How can I even check if the actual low-end response of the boxed speaker matches the simulations? Windowed at-home measurements won't work for low frequencies. We do have an "anechoic" chamber at our university department, and I guess that I could guess access to it if that would be helpful, but as far as I know the "anechoic" also only holds for frequencies above 100Hz or so...
• Will a 4-inch Waveguide from Somasonus even give me a matching directivity with a 4-inch woofer from Purifi, or are there other things to take into consideration of which I am completely unaware?
• What else am I missing?

Maybe it would be good to say something about my expectations before I stop. I am fully aware that, no matter what I do, I will not achieve a result that matches the quality of what a seasoned loudspeaker developer would do with similar budget and goals. I am fine with that. This alone would not be a reason for me to abandon this project. But if you feel that the end result would be destined to lose out against ainy semi-competently designed speaker kit for 200$, please stop me .

Any kind of response would be much appreciated!

Cheers,
Gian

 

[MSP0416]

Have you considered building Purifi's SPK8 speaker?

 

[alex-z]

Absolutely fine to ask for advice here, this is one of the best places on the internet for technical discussions of loudspeakers.

The Purifi 4" drivers have excellent distortion performance and bass extension for their size. In exchange, the sensitivity is quite low, so you will need fairly powerful amp modules to drive them. I personally think the Purifi passive radiators are overpriced, but their is no doubt of their quality.

With a 4" woofer and good quality tweeter, directivity is not a major concern. Especially with the steep filters that a digital crossover allows.

You could use the Seas Prestige 27TBCD/GB-DXT, which already has a small waveguide included. The Directiva R1 used this to achieve excellent measured performance. The waveguide isn't really necessary for directivity control in this application, but it will reduce baffle diffraction, so is still useful.

ASR Directiva Open Source Speaker Review

As some of you know, we have had a project to design an open-source speaker based on Purifi driver. It has been led by our member, @Rick Sykora who has named it "Directiva." You can read about it here...

www.audiosciencereview.com

The 4 and 8 Ohm versions of the Purifi 4" woofer are extremely similar. I would select the 4 Ohm version, because it will allow you to access more of the power output from the Hyper NC252MP amplifier.

Because you are using a digital crossover, the exact cabinet volume + tuning is not essential, because you can do slight adjustments digitally. It is normal that T/S parameters do not 100% match the manufacturer. The DATS v3 is an excellent tool which I do recommend, but you can also make your own much cheaper version with a pair of high precision resistors.

Measure Thiele Small parameters using free software - Audio Judgement

Learn how to measure thiele small parameters, or just a speaker impedance measurement using your sound card. Software used: Room EQ Wizard.

audiojudgement.com

A simple rectangular baffle with 3/4" roundover should be sufficient when using a waveguided tweeter. Again, I would point to the Directiva R1 as an example.

Toughest part of speaker design in my opinion is actually the measurement procedure. If you want accuracy, it is crucial that you have a free-field calibration mic, and a measurement stand which contributes no reflections itself. Initially I used a UMIK-1 + 90 degree mic clip, but this had severe accuracy problems above 7000Hz. Mounting the mic on the end of a long pole + changing calibration improved things substantially. Also, the speaker stand needs to rotate around the acoustic centre, not behind it.

You can do near-field measurements, apply baffle compensation, and then splice them to get reasonable low frequency measurements. You could also do a ground-plane measurement if you have a large outdoor space.

If you do the measurement + build process in a competent manner, there is no reason you cannot surpass the majority of existing loudspeakers. Of course, your 5 litre cabinet won't match the performance of KEF R3 or Genelec 8351, but those are much larger designs with 3 way crossovers.

Speaking of the cabinet, just follow best principles. Rigid front baffle, with internal bracing that connects all 6 faces of the cabinet. If you want to be fancy, do constrained layer damping for the brace structure, like how the KEF LS50 Meta and R3 are constructed. Lightly stuff the speaker with absorption, you are aiming to reduce standing waves without over-damping the woofer.

 

[Glorfindel]

Thanks to the both of you for the replies.

@MSP0416 Thanks for posting the link to the SPK8. Somehow, I had not found this design despite searching several hours for existing Purifi TPP4 designs. Very interesting. However, I don't think that I want to go for that. The reasons are (in descending order of importance) that I really want to develop a new speaker instead of building yet another kit, that I really want to go for an active crossover design, and that the Bliesma tweeter goes for 500 bucks per piece – I imagine that in any design where one tries to get a cutoff frequency of ~40Hz out of a 4-inch woofer, the "bottleneck" of the design will be the woofer (even if it is a Purifi). So I don't see myself spending that amount of money on a rocket-science tweeter... But very interesting nonetheless.

@alex-z
Thanks for the extensive reply. Can you elaborate a little bit why directivity should not be a (big) concern in my design? I understand that when I cross between woofer and tweeter in the ~1.5-3kHZ range, a 4-inch woofer will still have quite broad directivity at that frequencies, so there will not be a major discontinuity in terms of directivity. However, AFAIK, according to the Harman lore, one wants to get "constant directivity" (or at least "homogenous direcitivty"). Aren't waveguides also supposed to help with that? I.e., if I use a tweeter without waveguide, it will be more or less omnidirectional until several kHz but then start to become quite directional at very high frequencies; waveguide would even things out.

Thanks also for the comments on the differences between the 4Ohm and the 8Ohm variant. I get the point with increased power output. But why do they even make 8Ohm woofers then? So that you can put two in parallel?
And thanks for your comments on low-frequency measurements. By "reasonable", do you mean "good enough to use as basis for DSP corrections"? Ultimately, I anyway planned to also use Dirac on the miniDSP Flex to get a flat bass response specifically in the final listening setup, so getting perfectly accurate free field measurements might not be that important. (Of course, I would still like to end up with a speaker that can also work apart from Dirac...)

Concerning the tweeter discussion, I am aware of the Seas tweeter from the Directiva. What I don't really like about it is that I think it looks decidedly unsexy (<- Is this guy even a real engineer??). Would there be anything preferable about using the Seas over one of the DIY-waveguides from somasonus? (Apart from the easier cabinet build, of course...)

Thanks for the comments on the importance of good measuring. In that case, I really might see whether I can get access to our university anechoic chamber for the later design stages. Apart from that, I had planned to get a calibrated UMIK 1 from cross-spectrum.

To summarize, the design choices with Purifi PTT4.0X04 with one Purifi 4-inch PR in a ~5l box and miniDSP Flex crossover design seems to be "so far, so good"? (I also simulated the 4-Ohm version in VituixCAD to not get any bad surprises. The results look quite similar but not 100% identical to the 8-Ohm version, see the screenshots below...) The remaining question would be about the choice of the tweeter/waveguide.

Merry christmas to all, by the way!

 

[abdo123]

I’m waiting for the ported tweeter that Erin showcased on his channel to try and make something point-source like with this woofer.

 

[alex-z]

True constant directivity is nice, but not realistic in a small 2 way design. The woofer is going to have a wide radiation pattern in the mid-range, and the small baffle causes the baffle step transition (90 to 180 degree radiation) to occur as early as 800-900Hz.

Because the small woofer is going to be radiating so widely in the crossover region, a waveguide which narrows the tweeter substantially would be counterproductive. You are looking for a mild waveguide, like what the Genelec 8030C has. Hence my recommendation of the Seas DXT tweeter.

Yes, 8 Ohm drivers are useful for passive speakers, as many speaker amps are not stable with 2 Ohm load. For example, if you go to build a 3 way speaker with a 4 Ohm mid-range driver, that can cause a minimum impedance dip below 3 Ohms.

By reasonable, I mean enough to verify that your cabinet tuning is correct. For low frequency DSP corrections, I would always go based off the in-room response.

I honestly don't know enough about the somasonus waveguides to judge them.

If you can use a free-field calibrated mic in an anechoic chamber, that puts your measurement quality above 95% of speaker manufacturers, most don't have that. Most anechoic chambers are not truly anechoic though, so you may still need to use some gating depending on the size. And of course, the mic + speaker stand will still need to be built. Here are my crude versions, they work extremely well, I just cannot measure speakers bigger than about 20kg.

Yes, the Purifi drivers + miniDSP Flex well be a good combo.

 

[Jolly Joker]

True constant directivity is nice, but not realistic in a small 2 way design. The woofer is going to have a wide radiation pattern in the mid-range, and the small baffle causes the baffle step transition (90 to 180 degree radiation) to occur as early as 800-900Hz.

How does this compare to the Directiva R1? I'm new at this, but this seems like it could be a version made to cross over to subwoofers at 80 Hz unlike the 40 and 200 Hz of the current Directivas. Do they use any tricks to handle the baffle step beyond just having a larger woofer in R1?

 

[julbo]

Bliesma tweeter goes for 500 bucks per piece

Bought a T25B-6 pair a month ago for 420EU from sound_imports, I'm diverging from my Directiva R1 adventures with different tweeters and possibly waveguides. I'm pretty new to this as well and a lot yet to learn, one of my realizations is that a speaker that measures great may not sound so great to my ears.

 

[Glorfindel]

I want to give a small update. The project is happening (or so it seems at the moment ). I have decided to take the risk and go for the SB26ADC with 4 inch waveguide from somasonus. (I could still switch if acoustical measurements reveal this to be a total trainwreck driver combo.) Besides that, I am using the Purifi PT4.0X04 and the Purifi 4-inch passive radiator, the miniDSP Flex, and the Buckeye HypexNC252MP 4-channel amp. Much of this stuff has already arrived, but I am still waiting on the waveguide to arrive from some 3d-printer shop as well as on a calibrated UMIK-1 mic. (The Buckeye amp will take a while to arrive.)

The plan would be to simulate the enclosure in the next week in order to then build a prototype box that I can take acoustical measurements with, and to take it from there. I will keep you posted. Thanks for all the interest and guidance so far.

@julbo I had realized in the meantime that the 500 bucks were the price per pair, not per piece, so that checks out. Nevertheless, I didn't see myself spending that kind of money on a tweeter.

 

[alex-z]

How does this compare to the Directiva R1? I'm new at this, but this seems like it could be a version made to cross over to subwoofers at 80 Hz unlike the 40 and 200 Hz of the current Directivas. Do they use any tricks to handle the baffle step beyond just having a larger woofer in R1?

With a 4" woofer, crossing at 80Hz would be pretty reasonable if maintaining high SPL was a goal. With any active speaker setup, you can choose to extend the bass response at the cost of peak output. The Directiva R2 could technically play lower than 200Hz, but is being designed for use with bass modules, rather than being standalone.

For the Directiva R1, baffle step was left natural, as you can see in the raw measurements contained in this thread:

Post Directiva r1 Active Crossovers here

Here is v20 as it is the minidsp HD one in Amir's review.... The Woofer peak filter was needed for his older woofer. It is not used for my newer one. Note that the woofer high pass filter is shorted out, but I use it and may be useful for those wanting to use r1 with subwoofers. The...

www.audiosciencereview.com

All the tuning was done in the active crossover.

 

[abdo123]

The Directiva R2 could technically play lower than 200Hz, but is being designed for use with bass modules, rather than being standalone.

I don't think that's true. From what have been shared so far about the R2, it functions similarly to a ported speaker with the port tuned at 200Hz.

check the thread for details.

 

[alex-z]

I don't think that's true. From what have been shared so far about the R2, it functions similarly to a ported speaker with the port tuned at 200Hz.

check the thread for details.

Unless they changed the design since I last checked, the side firing vents are using to extend directivity control, the Dutch & Dutch 8C works the same. Much different application than a ported speaker, where the port resonance is used for SPL augmentation.

The other way to achieve cardioid behavior in the mid-bass region is with side-firing woofers. But that requires an additional DSP + amplifier channel, raising the total project cost. The Kii Audio Three uses this approach.

I am saying that because the speaker will be using an active crossover, how low they make it play is mostly arbitrary. The bass module crossover point being 200Hz is more a matter of distortion and directivity control than an actual design limitation.

 

[abdo123]

Much different application than a ported speaker

While the application is different, the principal is still the same. you're letting air from inside to outside through a vent.

You can read more about it at number 3 of @ctrl 's thread here https://www.audiosciencereview.com/...adiation-via-lateral-slots-like-d-d-8c.37863/

and if i remember correctly the low frequency extension was even worse than what was simulated. The simulations show -5db/oct loss compared to sealed (check number 3 again) but in reality it was closer to -10db/oct. Making the -10dB point of the R2 module at 100Hz. Check measurements here: https://www.audiosciencereview.com/...e-platform-speaker-project.20407/post-1043389

This is not considering that there was signficant boosting involved to get the response flat to 200Hz as it was measured by Tim. (Baffle step compensation and random cardioid losses). Which seem to be around 4dB total. (Check here: https://www.audiosciencereview.com/...e-platform-speaker-project.20407/post-1044803)

So you're looking at 14dB boost at a frequency range below a port's tuning frequency which needs massive amount of excursion just to get it to measure flat at 100Hz. (Which is why the D&D 8C uses an 8-inch woofer with much more excursion, and even then, there is a huge distortion penality).

I hope I didn't spam this thread too much, point is, there is nothing arbitrary about that 200Hz crossover, and it cannot be done any other way (via lateral slots).

 

[Glorfindel]

I promised more updates, so here we go. I had also purchased a Daytonaudio DATS V3, so I set out to measure the Thiele-Small-Parameters of the Purifi woofer myself. This is completely straightforward for many of the parameters, but others (the Vas) require additional inputs in form of the piston diameter and a known mass that is added to the woofer diaphragm. This proved to be somewhat hazardous, but one thing after the other.

For the straightforward measurements, I obtained the following:

fs = 46.3 Hz
Qts = 0.35
Qes = 0.38
Qms = 4.33
Re = 3.9 Ohm

These values differ not insignificantly from the Purifi datasheet, which has
fs = 38 Hz
Qts = 0.30
Qes = 0.32
Qms = 4.5
Re = 3.8 Ohm

However, I find them in closer agreement with the measurements from Erin and HifiCompass, so I don't think I should disbelieve them (except that my Qms seems awfully low). Erin/HifiCompass have measured
fs = 45.5/45 Hz
Qts = 0.367/0.37
Qes = 0.392/0.39
Qms = 5.725/5.3
Re = 3.71/3.8 Ohm

To simulate enclosures in VituixCAD, I need more, however. But when I tried to measure the Vas by calculating the piston diameter as sqrt(Sd/pi)*2, where I use the datasheet value of 56.7cm2 for Sd, and add a mass of 21g to the diaphragm, I get a Vas of 7.6l. The datasheet has Vas=6.4l, Erin and HifiCompass have Vas=4.26l and Vas=4.6l, respectively . However, if instead Iet the DATS V3 calculate the Vas based on the Mmd of 12.6g that Erin provides, I get Vas=4.35l. This value is much closer in line with Erin's and HifiCompass's values, and I don't trust at all my "added mass" approach.

The trouble is that these two different sets of values (the "measured" Vas and the "copied from Erin" Vas) don't yield remotely the same bass behaviour.
If I simulate this enclosure (which is roughly what I had planned),

then I get the following two bass responses (left: "measured", right: "copied from Erin").



The "copied from Erin" version pretty much matches with a simulation based only on the datasheet values of the Purifi:

So I guess/hope that this is where the truth lies.

Still, I am somewhat confused. My questions would be

• Why do I (and HifiCompass and Erin) measure such different parameters than are provided in the datasheet?
• Why do these curves still look so similar (at least if I assume that my Vas measurement was just trash)? I have read somewhere that some T/S parameters are actually "fundamental" (such as Re) and others (such as fs) are actually only derivatory and seem not to be used by vituixCAD. (I seem to be able to confirm this: If I change the fs of the drivers, the curves don't move an iota.) So are these sets of parameters in some sense "equivalent"?
• Should I really trust the datasheet parameters for the Purifi passive radiator, if the datasheet values for the woofer seems so inaccurate?

Another question concerns the "Qa" and "Ql" parameters of the enclosure. I have read that Qa is "absorption" and Ql is "enclosure losses". I understand that Qa is higher if I don't put any filling in the box and lower if I stuff the box completely, and that Ql will be low if my box is leaky and high if it is impenetrable. But what are sensitive values?
My previous simulations all used Qa=30 and Ql=15, but is this remotely realistic? If I change both of them, then the bass is completely lost, so I hope that is not what I will get.


But can I just not put any filling in the box to get a high Qa, or will this have other detrimental side effects? I understand that I should do a solid job on the woodwork to raise the Ql, but what value could I possibly aspire to?

I think it is obvious that I am a little out of my depth. If you could provide some thoughts and suggestions, it would be greatly appreciated (I fear this may become a recurring theme. I hope that I could compensate a bit by providing a documentation that is interesting to others.)

Cheers,
Gian

 

[Jolly Joker]

While the application is different, the principal is still the same. you're letting air from inside to outside through a vent.

You can read more about it at number 3 of @ctrl 's thread here https://www.audiosciencereview.com/...adiation-via-lateral-slots-like-d-d-8c.37863/

and if i remember correctly the low frequency extension was even worse than what was simulated. The simulations show -5db/oct loss compared to sealed (check number 3 again) but in reality it was closer to -10db/oct. Making the -10dB point of the R2 module at 100Hz. Check measurements here: https://www.audiosciencereview.com/...e-platform-speaker-project.20407/post-1043389

This is not considering that there was signficant boosting involved to get the response flat to 200Hz as it was measured by Tim. (Baffle step compensation and random cardioid losses). Which seem to be around 4dB total. (Check here: https://www.audiosciencereview.com/...e-platform-speaker-project.20407/post-1044803)

So you're looking at 14dB boost at a frequency range below a port's tuning frequency which needs massive amount of excursion just to get it to measure flat at 100Hz. (Which is why the D&D 8C uses an 8-inch woofer with much more excursion, and even then, there is a huge distortion penality).

I hope I didn't spam this thread too much, point is, there is nothing arbitrary about that 200Hz crossover, and it cannot be done any other way (via lateral slots).

Do I understand correctly that the reason for a separate bass module is that the slots would dampen the low frequency response if you didn't have a wall between the two? I realize 100-200 Hz can still be localized so you can't just cross over to subs (spread around the room) at that point, but is there some reason not to care about directivity in the 100-200 region other than needing large and efficient elements to get sufficient SPL with a slotted box?

Sorry for the off topic, the subject kinda raised wider questions.

 

[bigjacko]

Do I understand correctly that the reason for a separate bass module is that the slots would dampen the low frequency response if you didn't have a wall between the two? I realize 100-200 Hz can still be localized so you can't just cross over to subs (spread around the room) at that point, but is there some reason not to care about directivity in the 100-200 region other than needing large and efficient elements to get sufficient SPL with a slotted box?

Sorry for the off topic, the subject kinda raised wider questions.

It is not dampen but cancellation. The magnitude and phase of the two sound waves combines and determine the combined sound wave.

At below transition frequency, direction of the sound wave is not important because room modes and reflection come into play.

 

[D!sco]

This is a really nice and thorough first attempt. I don't really have a lot to add, I'll just say that if I had the money for everything but an anechoic chamber, I'd be doing what you listed.

I'd just be happy with how much better the measurements are vs the datasheet specs. Without a sub you can't really expect much at 30hz, I have a hard time figuring out what purifi was thinking with their 4" sometimes...

 

[hex168]

Something else to try to see if you prefer a waveguide with a 4" woofer or not; it won't cost you much:

https://www.digikey.com/en/products/detail/peerless-by-tymphany/BC25TG15-04/6557400

It was used here with a 4" woofer. A Purifi would be quite a step up:

Affordable Accuracy Monitor Review

This is a detailed measurements of a modified Pioneer SP-BS22-LR speaker by @Dennis Murphy called Affordable Accuracy Monitor (AAM). He was kind enough to send me a sample. I let him explain the mods but basically the tweeter was replace with a higher quality one and a new modified crossover was...

www.audiosciencereview.com

 

[AaronMat]

This is my active desktop speaker. Just finished it but haven't got around to measuring & tuning them. Two PRs & one ICEPower 100as2 in each box. The right side speaker has a minidsp 2x4hd and a baby lil ac/dc converter.