Hivi 3.1A DIY Speaker With Sehlin Mod Review

[Milan]

I thought I could do impedance sweep with my dx3pro, didn't realize coax inputs on it are digital and not analog. I ordered audio interface so I can do the test, I'll post when I get the results.

I am interested in looking a bit more into the difference and posting results to contribute a bit to this thread, I appreciate the advice on what to do, but also on what not to do, and I want to help other people make right choices (and avoid bad ones). I am definitely a beginner when it comes to speaker building and electronics in general.

I thought it would be interesting to see and compare how flush mounting vs surface mounting changes fr for this specific build, I could 3d print a ring to bring the mid back out to the surface and measure. I got suggestion to do impedance sweep of individual drivers and the entire speaker, I plan to do that as well. It's been years since measurements have been posted for these drivers, it's not impossible that something changed in the manufacturing process or quality control since there seem to be some significant differences between the units, at least to the limited sample I had a chance to measure.

Speakers will be getting grills at some point so I plan on chamfering the inside of the frame and do some wool felt treatment to the baffle to reduce inevitable diffraction grills will cause, I will also share my findings when that happens. I might even try to 3d print something that's a bit thinner than what came with the kit, I'll see how much time I'll have.

I am not interested in changing the crossover I made in any significant way (or even try to pry it out from those plastic pins). My goal is to try to eliminate as many other variables as possible before I definitely put blame on filtering for that 6.8khz peak. Horizontal directivity didn't seem to deviate from measurements on spinorama website. I have measured horizontal fr from 0 to 60 in 5 degree increments and just visually compared graphs to the ones on spinorama website, no major anomalies that I could notice, fr changes in similar manner. I didn't measure vertical. Now I know this isn't exactly scientific but I cannot replicate klippel accuracy anyways so for my purposes this is enough. Distortion graphs match Amir's. A bit cleaner down low since I plugged the port. Plugging the port doesn't cause any fr changes above bass area. Waterfall doesn't show any ringing at 6.8khz, I think around 4ms at 8khz is the longest trail I saw in that area. Those measurements have high noise floor, i will repeat when I look more into how to dial in the settings right.

I am happy with the sound, increased brightness and some eq works for my room, left wall is 20' away, right one is closer but treated, 18' tall ceiling, thick carpet. Living room is stupid big, way over 10000cuft of open space. I have never heard speakers that play this clean in my life, I always had to crank volume up to fill the space. Both me and my wife noticed that we tend to crank up the volume higher than before, but any reference to sound being played too loud coming from distortion being introduced is lost with these speakers. I never realized how much of my listening experience was degraded by distortion, fr was never an issue as that can always be tweaked with to a certain degree.

If anyone else in interested or has any suggestions on how to narrow down the source of 6.8khZ peak, please let me know, I will be interested in looking into it. Also, if you there's any other measurements I should do, or anyone wants to see.

Big thanks to everyone in this thread, you have been very helpful and I have learned a lot.

 

[Tinker]

Glad to hear you're getting it sorted out! It appears they are functioning a little better now now? I hope so.

I like the recess flush fit on the midrange. Wouldn't change that.

They are really clean sounding and benefit greatly when paired with a sub. Maybe two subs in your large room. Not sure why you plugged the ports?

The crossover has those little plastic snap-in clips. You would have to remove or break the little platic tang that holds it in. Did you glue them into the MDF? Maybe you could pry it out?

Keep at it, I'm sure you'll figureit out. It's worth it.

Good luck!

 

[Milan]

Glad to hear you're getting it sorted out! It appears they are functioning a little better now now? I hope so.

I like the recess flush fit on the midrange. Wouldn't change that.

They are really clean sounding and benefit greatly when paired with a sub. Maybe two subs in your large room. Not sure why you plugged the ports?

The crossover has those little plastic snap-in clips. You would have to remove or break the little platic tang that holds it in. Did you glue them into the MDF? Maybe you could pry it out?

Keep at it, I'm sure you'll figureit out. It's worth it.

Good luck!

I plugged the ports because I wanted to put them close to the wall and wanted to minimize boundary issues as much as possible. I used no 13 rubber stopper, fits perfectly in the port, completely sealed. I built 4in deep riser for them, hollow frame with 3in rockwool fill. Same behind them, 4in deep frame with 3in rockwool. I spent a lot of time finding the right spot for them where I could minimize sbir and control it without having to do to much treatment. I think in this position a lot of sbir cancel each other out, but there was a lot of trial and error. They sit on the hearth flanking the fireplace. I have 1st order sbir at 140 from the front wall, but very well controlled and low in amplitude, and smooth crossover at 90. Gonna wrap those frames at some point when I decide to make grills for them.

They replaced lxminis, and that meant one hypex 252 and minidsp not in use anymore. So I made sealed sub with Dayton 15in driver, bridged hypex to power it and added that to 2 hsu 15s I had. I ran MSO, and holy crap, that is one of the best pieces of software I have ever used in my life. I am at +- 1db at 3 listening positions up to the first sbir, which is around 3db at 140 and not really an issue, secondary ones I can't even tell on the graph. Treating the fireplace so I can put center in it was an absolute nightmare since all the gear is in there but I made it work with extended baffle and A LOT of rockwool.

I don't know if this can help someone add more subs to their living room, but what worked for me was to cover them with ottoman slipcovers and use them as stands for stuff. Not perfectly placed, but quantity over quality + MSO = WIN. Wife's happy

 

[Milan]

I don't have a lot of free time, so I am testing one thing at a time, in regards to where I deviated from original design and comparing with the original. I made no mods to the cabinet besides top and bottom roundovers which shouldn't influence the sound in any way, nothing on the baffle. Mounting mid flush is the only mod I did. These are the measurements with the mid flush with the baffle (black), surface mounted (red) and protruding 2mm from the surface (blue).

What i can gather from this is that flush mounting seems the best, because it doesn't cause a dip between 8 and 10khz, and keeps the response smoother and on target 3 to 5khz. 2.1khz also seems not to benefit from anything than other than surface mounting. 6.8khz peak unaffected.

Another thing i did was pop the grille frame on and measure that. There's no fabric on it, I just wanted to see what the frame does to the response, my assumption is (from the research i did) the frame makes much more of an impact than the actual cloth, provided it is reasonably transparent. Grille of (black) vs grille on (red)

Uneven tweeter response and diffraction ripples, as to be expected, some affect on 2-3khz.

I think it's safe to assume that flush mounting didn't cause any of the obvious issues in frequency response.

Another thing I notices comparing my measurements to spinorama is the drop i have 2.5-4.5khz, where there's definitely not a dip but a raise there on klippel data, around 3db. Off axis behavior is okay but the response definitely dips in my measurements. Mid recessed or flush, spinorama shows a peak at 4.5, not a dip. That's definitely midrange domain, approaching crossing region. There's obviously an issue in the crossover region, possibly midrange driver(s), less likely tweeter. I will do impedance sweeps hopefully this weekend, but as far as crossover goes, that one is really out of my league and could appreciate some help figuring that out.

Crossover is basically mod 6 with slight deviations. L1 unwound 16 turns as per @Scoobis instructions to get to 1.2mh. R1 is 7ohm instead of 6.8, R3 is 2.2 ohm (i found scotts post where he says anything between 2 and 2.5 is okay there, so I split the difference), and L1 is 0.15mh but 20awg instead of 18awg. Both crossovers are identical and both speakers measure almost identical besides slight difference (0.5db) in upper midrange/lower treble, which I believe is not crossover related because it follows the midrange driver.

Okay, so now I am committed to figuring this out for sure, I am even considering building the crossover from scratch. Before doing that tho, I was also thinking of measuring the midrange driver on its own and comparing it with the factory specs, i would appreciate if someone could point me to the relevant procedure and measurements for that. I will, hopefully, have the capability of measuring impedance and frequency response after this weekend, I do not own any analyzers, oscilloscope or anything more advanced than mili and megaohmmeter and regular multi. My gut tells me it's the midrange, and that the handover between mid and tweeter has issues, I just don't know how to confirm that for sure and eliminate any other variables.

 

[Milan]

Alright, so, I pulled the crossover out, checked all the resistors, everything is to spec. Traced out and did continuity checks on the board, everything that is supposed to have continuity has it, what isn't supposed to have it, doesn't have it. I pulled one lead of each capacitor and measured, all to spec. I went to check if I maybe swapped L2 and L3 somehow. I don't have LCR but with miliohm I was able to see which one is has less resistance = less inductance, and they are positioned correctly. Everything checks out, there's doesn't seem to be issues with crossover wiring.

I also checked tweeter and midrange response in the cabinet, unhooked one at the time, but I did leave woofer connected. Probably should have done it without the woofer but even with woofer hooked up you can kind of tell what's going on. W+T (blue) W+M (red) sum response (black) and Amir's cea2034 (cyan). I also compared everything to Amirs individual driver measurements but didn't plot that because it wasn't in his zip file, I just used his graph.

Midrange driver in amirs measurements drops 20db from 5 to 8khz, in my measurements it's around 10db. Rolloff starts way earlier in mine, around 3khz which explains dip between 3 and 5 and the slope is shallower which explains extra energy summed with tweeter response around that 6.8khz peak.

Tweeter response shows that 6.8 might not be a peak, but that there's a loss of couple of db between 8 and 12khz. This could be explained by extra padding on the tweeter, 7ohm resistor and higher resistance of L1 thinner wire.. And whatever mod 5 does to attenuate the highs additionally. And 6.8khz seems like extra energy from midrange caused by slower rolloff, somehow keeping response the same as Amir's.

So, again, I don't think issue is with tweeter here since my measurements show 14db difference on tweeter between 4 and 7 khz and Aamir's show around 12db in the same region which is really close considering that mod 6 changes tweeter filtering.

Woofer shows around 5db drop between 500 and 1000 in both mine and amirs measurements which means that the difference between cyan and black in the same region is caused by midrange and not woofer.

My conclusion is that, for whatever reason, midrange rolling off earlier in hand off to the woofer, as well as rolling off earlier in the handoff to the tweeter but also with the shallower slope. If anyone can explain to me how would this be crossover related I would appreciate that.

In the next few days I'll try to get midrange impedance sweep and free air measurements to see what the heck is going on with the midrange. I either have 2 drivers way out of spec, or I messed up both crossovers in the same way and can't figure out what I did wrong. Midrange crossover is the only thing where I didn't deviate from mod 6.

And no, it's not polarity issue, I am getting huge dips in bass and tweeter handoff if I swap the leads on the midrange.

And again, both speakers have almost identical FR (black and red) which means drivers and filtering are similar, so I'll be trying to figure this out on a single crossover and single midrange driver so I don't have to waste more time on this than I am already doing.

 

[Milan]

Okay, I might be onto something here. My assumption was that Scott's model should work with my speaker and that there shouldn't be any issues.. But if the drivers are different enough, same filtering wouldn't yield same results. So I tried addressing issues I noticed with crossover mods. Well, more like reverting some of the changes in mod 6 to previous mods.

Biggest issues in my response are 6.8k peak and 1k dip. I went back and replaced:
C1 from 1.5 to original 3.3uF,
L1 from 0.15 to 0.1mH
I would be lying if I said I really understand how these two influence tweeter hpf - Fc going lower in fr, but Q shifting higher, respectively. That's the best I could understand this.
I also lowered R3 from 2.2 to 1ohm (because I had spare 1ohm resistors) just to see what's going on in mid to woofer handoff. Mod 6 (green) and new experimental crossover (red)

What immediately stood out to me is that higher output of the woofer is making the cancelation worse at 1khz. Going from 1.5mH to 1.2mH on L1 effectively made cancelation worse by introducing more energy in 1khz region. My assumption is that reverting back to 1.5mH should make the cancelation smaller.

Reverting C1 and L1 to original values gave a bit more energy around 12k, but lowered that 6.8khz peak around 1.5db without significantly affecting upper midrange. This is a big win in my book, treble regions seems smoother now.

I wanted to see if I could anything else with spare parts I have, and I started messing with c3 and c4 by adding 6.8 3.3 and 1.5 caps in parallel. Most of those just caused cancelations in midrange/treble area, but one mod seemed like it might had positive result (tho pretty small). I put 2.2 ohm back on the woofer and paralleled 1.5uF to C4 (green) and compared to previous graph (red)

C4 11.5uF seems to make upper midrange a bit more linear without having negative impact on treble, so I might keep that or try 12uF cap there, they are cheap enough. Again, padding the woofer back definitely didn't make the dip worse, so I am pretty confident I might get good result with original 1.5mH L4.

My ultimate goal with this is to get on axis fr as linear as possible without major dips. I am okay with peaks because this is going to get 5-6 peq filters to flatten everything out. Like I said before, side and ceiling reflections are not an issue in my room so any non linearity has less effect in my case than theory suggest. I did 0-60 with 6 peq bands and the grille on, and I am happy with results. Will be even happier if I get that 1khz dip under control so I can get away without boosting that area too much. REW applied boost at 2.3khz which I'll probably skip, else looks okay to me.

 

[Vze4dckm]

Hi all,
This is my first attempt at DIY speaker building. Very satisfying thus far!
Question about the terminal plastic plate: I like the look of "tube connectors" (have no idea what it means), so I cut a piece of hardwood to match the opening before covering the back side with veneer. Then I will drill through to mount a pair of banana plugs. Can I just unscrew the terminals that are on the plastic plate and use these directly?
Planning on mod 5 for xover (parts are en route)

 

[Reydepancakes]

Hi all,
This is my first attempt at DIY speaker building. Very satisfying thus far!
Question about the terminal plastic plate: I like the look of "tube connectors" (have no idea what it means), so I cut a piece of hardwood to match the opening before covering the back side with veneer. Then I will drill through to mount a pair of banana plugs. Can I just unscrew the terminals that are on the plastic plate and use these directly?
Planning on mod 5 for xover (parts are en route)
View attachment 536186View attachment 536187

I'd say your time is better spent just using the original black plastic insert or getting new plugs.

 

[Vze4dckm]

I see that the C5 upgrade from 47 to 68uF can result in a very crowded board
Does it make sense to keep this big boy and expand out (like what @Milan did)?
Or, should I just get a 100V that fit nicely on the stock board?
I could also get a wider board to fit all parts comfortably, and ditch the stock board.
I know nothing about these components, but based on cost, one would assume the 400V should offer a better result?

 

[a4eaudio]

Or, should I just get a 100V that fit nicely on the stock board?

A 68uF 100V NPE in that location of the xo should be fine.

Depending on where you are located, a 160V option LIKE THIS ONE would be a little smaller than the 400V if you wanted to go in that direction.

 

[Astray]

I wonder how this compares to the HiVi-Swans D3.1 MKII. Do you think they would benefit from the Sehlin mod as well?

 

[GGroch]

I wonder how this compares to the HiVi-Swans D3.1 MKII. Do you think they would benefit from the Sehlin mod as well?

The D3.1 MKII is much more closely comparable to the DIY 2.2a using nearly identical drivers and dimensions. The 2.2a seemed to have less to criticize in FR with its original crossover. See my post in the DIY3.1/2.2a thread with a photo of the D3.1 MKII crossover. It is significantly more sophisticated (ie, has more components) than the 2.2a.

This photo was from a GR Research YouTube video. They tried to tear one down for crossover mod evaluation. He could not figure out how to remove the woofer (it appears the black front facing is not removable). So, removing and replacing the crossover may be very difficult.

 

[Astray]

The D3.1 MKII is much more closely comparable to the DIY 2.2a using nearly identical drivers and dimensions. The 2.2a seemed to have less to criticize in FR with its original crossover. See my post in the DIY3.1/2.2a thread with a photo of the D3.1 MKII crossover. It is significantly more sophisticated (ie, has more components) than the 2.2a.

This photo was from a GR Research YouTube video. They tried to tear one down for crossover mod evaluation. He could not figure out how to remove the woofer (it appears the black front facing is not removable). So, removing and replacing the crossover may be very difficult.

So I should treat this as a DIY 2.2a and do the Sehlin mod?

Are you talking about this review? A lot of the comments suggest that it's actually not that difficult to disassemble if you know what to do.

 

[GGroch]

...A lot of the comments suggest that it's actually not that difficult to disassemble if you know what to do.

I missed that, still I have not seen any photos of the innards besides GRs. As a 2.2a owner (I think they sound great) I'd be interested in seeing how different they are and the exact components.

Like Amir, I am skeptical of Danny's expertise, but I figured removing drivers to access crossovers was his superpower.