CSS Criton 3TD-X Kit Speaker Review

[musicisair]

No CAD at the moment as we haven't started CNC'ing anything. We prototype on the table saw first. What we are looking at first is a 30 degree angle with the cut being 2.5" in from the top corner and 11.5" down the side for the side edges and another 30 degree cut the width of the top starting at 2-2.5" down from the top edge. The center will be very hard to do facets on because the tweeter is very close to the top edge. You can hit all the corners of the cabinet with a chamfer around the woofers to make a better visual match though. I'd love to see your results when you finish!

I think that will look pretty great. I'll have to mock it up in some software and play with some ideas to try to make them match visually; the chamfer around the woofers on the center is probably the way to go.

I happen to be working on a VCAD simulation of the 3TD-XC (and got loads of help from Dan over email; thanks again, Dan!) and got a simulation of it that is remarkably close to @amirm's Anechoic FR of the 3TD-X. It even has the peaks at 850 and 10k+, and the dip at 2200k:

If anyone wants a copy of the VCAD project to toy with let me know!

 

[MarcT]

correct. The CSS Typhon is on sale for $3K -- $2.5K cheaper than the built 3TD-X. I wonder how they compare...

That Typhon is interesting at that price, but I'm not sure I want a multi-box speaker setup. And I wonder if it sounds as good as the 3TD-X.

 

[Kerry Armes]

I think that will look pretty great. I'll have to mock it up in some software and play with some ideas to try to make them match visually; the chamfer around the woofers on the center is probably the way to go.

I happen to be working on a VCAD simulation of the 3TD-XC (and got loads of help from Dan over email; thanks again, Dan!) and got a simulation of it that is remarkably close to @amirm's Anechoic FR of the 3TD-X. It even has the peaks at 850 and 10k+, and the dip at 2200k:

View attachment 408496

If anyone wants a copy of the VCAD project to toy with let me know!

This is really just luck that it ended up kind of close to the overall response. The individual driver responses are not actually close to the actual roll-off we have with the crossover. You can see the tweeter is operating with basically full output over essentially its whole bandwidth.

 

[musicisair]

This is really just luck that it ended up kind of close to the overall response. The individual driver responses are not actually close to the actual roll-off we have with the crossover. You can see the tweeter is operating with basically full output over essentially its whole bandwidth.

I've only been doing loudspeaker simulation for a whole 2 days now, so I've most certainly got many (most?) things very wrong; so I'm not going to argue with you about just getting lucky

In the simulation, if I change the 3TD-XC's crossover to use the 3TD-X's crossover instead it flattens out quite a bit.

FR simulation of 3TD-XC's baffle diffraction with 3TD-XC's crossover network:

FR simulation of 3TD-XC's baffle diffraction with 3TD-X's crossover network:

 

[Rick Sykora]

They told me they don't make a measurable difference, either at component level or for the speaker as a whole.

As more than a few posts have tried to clarify, suggest the OP state more clearly the price for a kit vs finished speaker.

Also, the finished one sells with the crossover upgrade. For the sake of full and accurate disclosure, @Kerry Armes did Amir test the base crossover or an upgraded one?

 

[Kerry Armes]

As more than a few posts have tried to clarify, suggest the OP state more clearly the price for a kit vs finished speaker.

Also, the finished one sells with the crossover upgrade. For the sake of full and accurate disclosure, @Kerry Armes did Amir test the base crossover or an upgraded one?

It had the upgraded crossover in it. The finished version only ships with the upgraded crossover now. We used to offer the option for the standard crossover on finished speakers speakers, but only had a single customer over four years buy any with a standard crossover so we removed the option. We could still technically do it as job if someone asked.

To reiterate, the part values are exactly the same so there is no measurable difference in the frequency response.

 

[Kerry Armes]

I've only been doing loudspeaker simulation for a whole 2 days now, so I've most certainly got many (most?) things very wrong; so I'm not going to argue with you about just getting lucky

In the simulation, if I change the 3TD-XC's crossover to use the 3TD-X's crossover instead it flattens out quite a bit.

FR simulation of 3TD-XC's baffle diffraction with 3TD-XC's crossover network:
View attachment 408504


FR simulation of 3TD-XC's baffle diffraction with 3TD-X's crossover network:
View attachment 408503

No worries and good for you for diving in headfirst. It would be helpful to see the raw files you’re using as well as the impedance and crossover network. There is definitely something off between what you have and the measurements for the center.

 

[Rick Sykora]

I've only been doing loudspeaker simulation for a whole 2 days now, so I've most certainly got many (most?) things very wrong; so I'm not going to argue with you about just getting lucky

In the simulation, if I change the 3TD-XC's crossover to use the 3TD-X's crossover instead it flattens out quite a bit.

FR simulation of 3TD-XC's baffle diffraction with 3TD-XC's crossover network:
View attachment 408504


FR simulation of 3TD-XC's baffle diffraction with 3TD-X's crossover network:
View attachment 408503

Welcome to ASR!

If trying to learn VituixCAD, an example project is often helpful. If @kimmosto does not have a good one, this is one from a solid passive design: https://www.audiosciencereview.com/...ers-h-v-off-axis-measurements-included.41757/

 

[kimmosto]

Welcome to ASR!

If trying to learn VituixCAD, an example project is often helpful. If @kimmosto does not have a good one, this is one from a solid passive design: https://www.audiosciencereview.com/...ers-h-v-off-axis-measurements-included.41757/

I would start studying and playing with a project having simpler and more intuitive/natural crossover topology. For example acoustically decent 3-way design usually requires ca. 12 XO components, and 2-way just 7-8 pcs. Original Mechano23 had 13 components which may give impression that it's required or worth while it's probably not. In addition, it has shunt notches which are worse that series for attenuating THD and IMD boosted by cone break-up.
VCAD user manual links to few very old projects (Kontiainen and Epe-3W) with vxp and measurement data. Those are just examples to play with. Not any reference designs for sure

 

[musicisair]

No worries and good for you for diving in headfirst. It would be helpful to see the raw files you’re using as well as the impedance and crossover network. There is definitely something off between what you have and the measurements for the center.

Thanks Kerry!

I've zipped up the VCAD project if you want to take a look. It contains the `.zma` files and `.frd` files used in the simulation. I also added the documents/images I used to trace the curves, as well as the 3TD-XC's Plan PDF from the CSS website (all in the Datasheet folder). Here is the complete project archive: https://drive.google.com/file/d/1O7usgXBGelYYmCKdW53RFx6HtL90N9lS/view?usp=sharing

For the curves:
* I first traced the frequency responses for each driver using the SPL trace tool[1], then assigned them to the drivers and ticked the "Minimum phase" for each.
* then used the Diffraction tool to for each driver: using the `Sd` value from the datasheet, placement and baffle sizes from the 3TDX-C's plan PDF, and applied the driver's FRD (using "full space" response), then exported these as `<driver>diffraction.frd` (you can find both .frd files for each driver in the .zip file).
* I then replaced my traced .frds with the `<driver>diffraction.frd`s for each driver, and unticked the `Minimum phase` option (only because keeping it ticked made things worse).

For the crossover:
* I implemented the 3TD-XC crossover network that includes the extra 0.47μF capacitor in the LDW6's network (in Variant R2, Variant R1 is the 3TD-X's network)
* I positioned all drivers' X and Y, relative to the tweeter's center, per the enclosure dimensions.
* I used ChatGPT to give me some guessing points for the Z-axis for each driver

(Hopefully) Inconsequential tweaks to the crossover:

* I replaced the 75μF and 125uF capacitors with a bunch of parallel 25μF capacitors
* The inductor values/DCR are a tiny bit different than the CSS schematic, as I used the specs from some Solen-brand air-core inductors

A note:

I noticed today that Audio Express's testing of the LD25x has a slightly different impedance curve than the datasheet on the CSS website. Their testing looks to have it peak at around 890Hz/20Ω, and the CSS datasheet shows about 855Hz/21.4Ω. These seem minor to me, but I thought I'd mention it because I can't intuit what variables really matter.

[1] except for the LD25X's impedance curve; it is very low resolution, so I used a different tool that let me zoom in much closer.

---

I thoroughly I enjoyed learning about just how much I have yet to learn. haha

Thanks @Kerry Armes , @Rick Sykora, and @kimmosto for the tips. I'll try to look at some simpler projects next week!

 

[Ktacos]

3TDX-C

Ok so you're trying to sim the center, and comparing it with the tower? They're two completely different speakers and are not going to measure the same at all.

 

[Paweł L]

Nice kit. Good drivers, simple crossover. For someone who wants to put together decent set of towers hard not to recommend them. The only thing I would change in crossover, based on measurement, is the tweeter level which stands out a little. Unless this tweeter just sounds dead... , kind of like my SB29RDNC, way too damped.

 

[musicisair]

Ok so you're trying to sim the center, and comparing it with the tower? They're two completely different speakers and are not going to measure the same at all.

Yeah, I understand. This came about because I was already working on simulating the Center when CSS Audio shared this review thread on Instagram. I noticed that the curves looked, to my untrained eyes, similar, so I posted my comparison here and things rabbit-holed from there. I should probably just switch over to simulating the 3TD-X tower now that I know there is great anechoic data to use as a reference.

Sorry for hijacking this review thread

 

[Paweł L]

Thanks Kerry!

I've zipped up the VCAD project if you want to take a look. It contains the `.zma` files and `.frd` files used in the simulation. I also added the documents/images I used to trace the curves, as well as the 3TD-XC's Plan PDF from the CSS website (all in the Datasheet folder). Here is the complete project archive: https://drive.google.com/file/d/1O7usgXBGelYYmCKdW53RFx6HtL90N9lS/view?usp=sharing

For the curves:
* I first traced the frequency responses for each driver using the SPL trace tool[1], then assigned them to the drivers and ticked the "Minimum phase" for each.
* then used the Diffraction tool to for each driver: using the `Sd` value from the datasheet, placement and baffle sizes from the 3TDX-C's plan PDF, and applied the driver's FRD (using "full space" response), then exported these as `<driver>diffraction.frd` (you can find both .frd files for each driver in the .zip file).
* I then replaced my traced .frds with the `<driver>diffraction.frd`s for each driver, and unticked the `Minimum phase` option (only because keeping it ticked made things worse).

For the crossover:
* I implemented the 3TD-XC crossover network that includes the extra 0.47μF capacitor in the LDW6's network (in Variant R2, Variant R1 is the 3TD-X's network)
* I positioned all drivers' X and Y, relative to the tweeter's center, per the enclosure dimensions.
* I used ChatGPT to give me some guessing points for the Z-axis for each driver

(Hopefully) Inconsequential tweaks to the crossover:

* I replaced the 75μF and 125uF capacitors with a bunch of parallel 25μF capacitors
* The inductor values/DCR are a tiny bit different than the CSS schematic, as I used the specs from some Solen-brand air-core inductors

A note:

I noticed today that Audio Express's testing of the LD25x has a slightly different impedance curve than the datasheet on the CSS website. Their testing looks to have it peak at around 890Hz/20Ω, and the CSS datasheet shows about 855Hz/21.4Ω. These seem minor to me, but I thought I'd mention it because I can't intuit what variables really matter.

[1] except for the LD25X's impedance curve; it is very low resolution, so I used a different tool that let me zoom in much closer.

---

I thoroughly I enjoyed learning about just how much I have yet to learn. haha

Thanks @Kerry Armes , @Rick Sykora, and @kimmosto for the tips. I'll try to look at some simpler projects next week!

I will take a look. Thanks, as I don't have the frd/zma files and X-over data to see what CSS came out with.
As for the difference in tweeters measurements, don't worry, +/-10% in fs is factory tolerance and quite common. If you get a pair of any drivers with fs within 5% you will be very lucky. And a few % in fs difference doesn't have that much of a influence on freq response in usable range in tweeters

 

[Paweł L]

So I redone your CSS project. I hope your simulated frd with baffle step is correct, cause a lot of values didn't make any sense to me, as a well as driver centers - most mid woofers have 'z' 15-20mm behind the tweeter. The tweeter has fs around 800-900Hz so min X-O point should be around 2.5kHz to keep the voice coil from baking itself. Also looking at Vance's review of the tweeter the distortions increase steadily past 3kHz. This the simplest sort of a proper X-O I could come up with. However buying the kit and implementing original X-0 with some value changes should also work in Center channel and it would more sense than building one from scratch.

 

[Paweł L]

Wow I didn't know CSS actually gives plans for the data including X-over. Well, musicisiair definitely has the wrong data for individual drivers freq response, and that's why Vituix was showing me hills and valleys

 

[musicisair]

Wow I didn't know CSS actually gives plans for the data including X-over. Well, musicisiair definitely has the wrong data for individual drivers freq response, and that's why Vituix was showing me hills and valleys

I just traced the curves using the SPL trace tool from their driver datasheets. I did it multiple times, trying to be as precise as physically possible (using the logarithmic seeing when appropriate). Maybe I still did something wrong.

I'll take a look at your project, and another close look at the FRD files I created tomorrow.

 

[Paweł L]

Measured vs traced and baffle simulated is most likely the difference. Also ticking min phase shouldn't make much difference since when you apply baffle response correction to driver's frd, the sum will be calculated with min phase. Calculated phase actually might change when the frd's upper and lower limits are extended - it's a very delicate parameter, but it will ruin the simulations if not done correctly. Also I don't use smoothing in my frds as they already come 'averaged' in term of resolution.
When you look at the Vance's review of the tweeter, his measurements clearly show the influence of the baffle - there are hills and valleys below 5kHz with large dip somewhere in 2-3kHz. Woofers and mids should have clear '6dB baffle step loss' between 100Hz and 1kHz. You can see that in Vance's midwoofer reviews, where he even states the size of the baffle.

 

[musicisair]

So I redone your CSS project. I hope your simulated frd with baffle step is correct, cause a lot of values didn't make any sense to me, as a well as driver centers - most mid woofers have 'z' 15-20mm behind the tweeter. The tweeter has fs around 800-900Hz so min X-O point should be around 2.5kHz to keep the voice coil from baking itself. Also looking at Vance's review of the tweeter the distortions increase steadily past 3kHz. This the simplest sort of a proper X-O I could come up with. However buying the kit and implementing original X-0 with some value changes should also work in Center channel and it would more sense than building one from scratch.

I've rechecked and triple checked, regenerated the traced frd and zma files, and then the diffraction responses, rechecking everything again, and as far as my noob self can tell, my frd and zma files are as accurate as they can be without measuring, but the response still looks like an 8.6 earthquake on the Richter scale. The crossover network is accurate, per the schematic in the published plans. So I'm thinking it's mostly likely down to: shoddy traced infinite-baffle anechoic FR/zma, misaligned driver z-axes, omitting the simulated enclosure FR step, or just plain-ol user error (in that order, but maybe my confidence is unwarranted).

I wish driver manufacturers published their raw data for use in pure simulations, or does simulation fall far too short of reality for on-axis only raw anechoic data in a 3-way crossover to be all that useful?

 

[a4eaudio]

I wish driver manufacturers published their raw data for use in pure simulations, or does simulation fall far too short of reality for on-axis only raw anechoic data in a 3-way crossover to be all that useful?

What baffle would they measure on? If they used 8" wide and you were using 9-1/2" wide, it wouldn't be good enough. Just like going from their infinite baffle to finite using diffraction simulations you'd still have to do the same. And while the on axis sim might be close, no telling about the off axis.