The HiVi DIY 3.1 kit is still available on Amazon including knockdown cabinets. Some of the reviews indicate that the overall sound is a bit bright, but can be corrected by increasing mid and tweeter resistor values. Rich (Turn 2) had done this to a pair that he also brought to the Iowa show, but a wiring error led to some strain on the mids - so we stopped the demo early. Rich indicated that he had modified the resistor values. I also picked up a pair, encouraged by my experience with the HiVi dome midrange in the Indium 7 project and having been impressed with the HiVi planar tweeter in a couple of projects. I found the sound brighter than what I would consider natural, so I developed several modifications of varying complexity to tailor the sound toward a more neutral presentation. I stepped through the process of evaluating and modifying the crossover on the Midwest Audio Club forum HiVi DIY 3.1 Discussion and will summarize the results below.
Stock Crossover and Measurements:
Note: When I measured the woofer, I had the polarity reversed, so ignore the fact that the polarity is reversed in the crossover diagrams - I just needed to do that to get sim to match subsequent measurements.
This actually looks almost identical to what is posted on HiVi's website (but using a more revealing scale). I didn't splice in any low frequency nearfield response, so anything below about 200-300 Hz is not accurate. I guess one could call the response 89 +/- 4 dB without really lying - but most of the 8 dB variation is a steady rise from 85 dB in the 200-300 Hz area to 93 dB around 12 kHz. I would expect this to sound a little bright and, to my ears, it does.
Option 1: A simple fix if the speaker is already built.
The approach here is to leave the stock crossover alone and just add a resistor across the driver terminals for the mid and tweeter - as this would be a really easy fix. A reasonable attempt is a 10 ohm resistor (R5 on the diagram) across the DMN-B and a 5.6 ohm resistor (R4) across the RT-1.3B, which results in a simulated response that would fall into a +/- 2 dB window around an average of about 86.5 dB at 2.83V, 1m.
The parallel resistor fix actually sounds pretty good. Tried with some better quality material and cheesy 80's music. The midrange presentation is slightly forward, but it is over a broad range, so it isn't really peaky or irritating. The 12 kHz peak is down enough to line up with the midrange. The recessed area in between correlates with a BBC dip and is also pretty broad and shallow, so it doesn't really feel like a lot is missing. Bass/midbass is present at the right level, so BSC appears set about right. For a cheap, easy fix, this looks like a win.
Option 2: Resistor replacement.
If one is starting out assembling these, it might be preferable to change the values of the existing series resistors rather than add parallel resistors. In that case, it looks like a similar amount of padding can be acheived using a 6 ohm resistor on the tweeter and a 3 ohm resistor on the mid. This approach gives similar, but possibly slightly less linear response than Option 1.
Option 3: Further smoothing out the midrange response.
With this crossover topology, there is another tool that can be used to help with the midrange peak - R3. Changing R3 from 1 ohm to 2 ohms helps bring the midrange peak down. The value can be increased further. By the time the value gets to 3 ohms, the midrange peak is lower than the 12 kHz peak - which experience tells me won't be desirable. Further padding the 12 kHz peak at that point will shift the overall balance slightly more bass-heavy. I would probably settle on something in the range 2 to 2.5 ohms and call it good.
Option 4: The perfectionist option:
The real limitation of the stock crossover approach winds up being the shape of the high frequency response. If that could be flattened out, then one could make the tweaks needed to really flatten the rest of the response without the 12 kHz area sticking out as an overall peak and sounding "icy" for lack of a better term.
Here is a modified crossover that would allow reuse of the 3 largest inductors, 2 largest caps, and the PC board itself. The woofer filter is the same as in option 3. The 6.8 uF cap in the midrange filter is changed to 5.6 uF, which addresses the area between 1 and 2 kHz. The tweeter crossover has all new components, but they are small.
I sent the HiVi DIY 3.1 with Option 4 "The Perfectionist Mod" to Audio Science Review during the pandemic, where it was awarded the highest rating (the golfing panther). I do my best to follow sound measurement practices, but it was nice to have those validated using a far more costly and sophisticated measurement apparatus (the Klippel Near Field Scanner). I think it helped get the attention of Swans-HiVi also, as they now mention the modification in some of their product literature. My motivation was just to help people get the most out of what is a very nice set of high value drivers and a good, sensible cabinet design. So, I am happy to have largely met that objective.
In the aftermath of the review, it was clear there were still opportunities for at least minor improvement. Amir, the reviewer, had recommended some eq of the speaker to further improve the sound and I came up with a couple of additional crossover variations to address those observations and take another step toward optimal sound from this configuration. The downside of these variations is that they involve higher cost as they change out some of the larger components, which wasn't needed with the original "Perfectionist Mod".
Option 5 - the next step
This tweak was first suggested on page four of the Audio Science Review thread on the modified DIY 3.1's. It involved the following:
1. Increase C5 (in the woofer circuit) from 47 uF to 68 uF.
2. Increase R2 from 2 ohms to 3 ohms in the midrange circuit
3. Increase R1 from 5.6 ohms to 6.8 ohms (7 ohms would be fine depending on which is available) in the tweeter circuit.
The resulting crossover network and simulated response is shown below:
This does bring down the level of the mids and highs a little more and results in a fuller, more relaxed sound relative to Option 4. Below is a comparison of the Perfectionist mod (red line) and Mod 5 (blue line) frequency response. I currently use mod 5 in my pair.
The other observation that was made during the review was that bass could sound somewhat "tubby" in some cases. The cause for this could be a slight dip in the response from 200-400 Hz that was seen on the Klippel measurement. The reduced level below about 450 Hz can also be seen in the red curve above (my measurement of mod 4 "the Perfectionist Mod"). To at least partially address this, I proposed a further tweak of the woofer circuit which became "Option 6". Option 6 requires reducing the value of the woofer inductor L4 from 1.5 mH to 1.2 mH and increasing the capacitor C5 value to 100 uF. The final schematic for option 6 is shown below.
It can be seen below (mod 6 is the green curve) that there is a slight improvement in the 200-400 Hz range, but the difference is not nearly as significant as the difference between mod 4 and mod 5. Also, the woofer inductors are relatively expensive components.
I guess my take is that I would probably use the mod 6 crossover if I were designing this from scratch, but if I was modifying the existing kit, I might be tempted to stop at mod 5. It should be pointed out that if you have the means to measure inductance, it is possible to unwind L4 a bit to reduce the inductance from 1.5 to 1.2 mH.