2018/12/29 created
2024/03/19 updated

Current revision of Gaudi II: Rev.2.2 (May 2023)
The digital crossover network, dbx DriveRack VENU360, was employed in the previous revision (Rev.2.12). I flattened the frequency response by using its equalizer feature, but another problem surfaced. The transient response of the midrange wasn't good. For the solution, I improved the loudspeaker SS-309A to Rev.B. The improvement includes replacement of the squawker and other minor improvements. I carried out system-level tuning with SS-309B.
For the past revisions of Gaudi II, click here.

Configuration

The system configuration is the same as the previous revision except the improved loudspeaker, SS-309B.
Since the specification of the loudspeaker has been changed, I redid system-level tuning. As a result, the settings of the network (VENU360) has been largely changed.

The amplifiers that were used in the old Gaudi system (Gaudi Ver.1) are still used in this system. I postponed building new amps because the amps don't affect sound quality so much, compared to loudspeakers and room acoustics.

I drew a new level diagram, because the gains of the preamp and the crossover network are different from those in the system design.

---

Layout

The component layout is as shown on the right (click to zoom up).

The loudspealer has been changed from SS-309A to SS-309B.
Other than that, the layout is the same as the previous revision (Rev.2.12).

The system layout is as shown in the figires below. It's the same as Rev.2.12 too.

---

Cabling

The cables used in Gaudi R2.2 are as shown in the table below.
I did away with the questionable cables made by 47 at last.

From	To	RefDes (model#)	Wire	Length	Sender connector	Receiver connector	Note
ADP (SL-1200GR)	MC head amp (HA-213)		Low capacitance unbalanced wire (1-core shielded)	1.5m	RCA plug	RCA plug	SL-1200GR's accessary Tied with earth wire Stereo pair
MC head amp (HA-213)	Preamp (PA-210)		OFC balanced wire (2-core shielded) BELDEN 8412	1.5m	Teflon insulated RCA plug COLD-GND	Teflon insulated RCA plug	Unbalanced Stereo pair
DAC (UD-301)	Preamp (PA-210)		4-core shielded wire Canare L-4E6S	2m	RCA plug (Canare F-10)	RCA plug (Canare F-10) COLD-GND	Unbalanced Stereo pair
Preamp (PA-210)	DAR (MR-2000S)	CB2	4-core shielded wire (Canare L-4E6S)	2m	RCA plug (Canare F-10)	RCA plug (Canare F-10)	Unbalanced Stereo pair
DAR (MR-2000S)	Preamp (PA-210)	CB3 (Canare RC018)	1-core shielded (Canare GS-6)	1.8m	RCA plug (Canare F-10)	RCA plug (Canare F-10)	Unbalanced Stereo pair
Preamp (PA-210)	Network (VENU360)	CB5 (Canare RC03-B2)	4-core shielded (Canare L-4E6S)	3m	RCA plug (Canare F-10)	XLR pulg COLD-GND	Unbalanced Stereo pair
Network (VENU360)	Bal/Unbal (CleanBox)	CB6 (Canare EC01-B)	4-core shielded (Canare L-4E6S)	1m	XLR jack	XLR plug	Balanced Stereo pair
Bal/Unbal (CleanBox)	Tweeter amp (MA-208)	CB4 (Canare RC03)	1-core shielded	3m	RCA plug	RCA plug	Unbalanced Stereo pair
Network (VENU360)	Bal/Unbal (CleanBox)	CB7 (Canare EC01-B)	4-core shielded (Canare L-4E6S)	1m	XLR jack	XLR plug	Balanced Stereo pair
Bal/Unbal (CleanBox)	Squawker amp (MA-215)	CB1	4-core shielded (Canare L-4E6S)	1.5m	RCA plug (Canare F-10)	RCA plug (Canare F-10)	Unbalanced Stereo pair
Network (VENU360)	Woofer Amp (DAD-M100pro)	CB14,15 (Classic Pro CXX050)	2-core shielded	1.5m	XLR jack	RCA plug (Canare F-09)	Unbalanced Mono x2
Tweeter amp (MA-208)	Tweeter (T925A)		OFC AWG20 x4-core (Canare 4S6G)	2.7m	Spade terminal Phi=3.5mm (Nichifu TMEX 1.25Y-3.5)	Spade terminal Phi=8mm (RS Pro 613-9485)
Squawker amp (MA-215)	Squawker (D1405+H400)		OFC AWG20 x 4-core (Canare 4S6G)	2.7m	Spade terminal Phi=8mm (RS Pro 613-9485)	Soldered
Woofer Amp (DAD-M100pro)	Woofer (FW305)		OFC AWG20 x4-core (Canare 4S6G)	2.4m	w/o plug	Spade terminal Phi=8mm (RS Pro 613-9485)	The cable length is from amp to woofer terminal

*1: The red texts indicate the cables that are changed in Rev.2.2.
For details of the line cables, see the PDF below:
[Design document on line cables (LineCable_Design.pdf)]

---

AC Power Supply

The figure on the right shows the AC power connection (click to zoom up).

It is the same as the previous revision (the figure is Rev.2.12).

---

System-level Tuning

The system-level tuning was redone, since the loudspeaker (SS-309B) was modified (the procedure is the same as the previous revision).

• 1. Measurement of frequency response in quasi-anechoic method, and adjust of the gains of each band, plus compensation with PEQ
• 2. Measurement of the delay times of each band with one-cycle sine wave, and adjust of time alignment
• 3. Measurement of frequency response and waveforms at the listening position
• 4. Confirmation of sound quality by trial listening

I repeated this sequence several times to optimize the system.

The finall settings of VENU360 is as follows:

• Configuration: stereo 3-way
• Crossover frequencies: fc1=900Hz, fc2=6.3kHz
• Filter type: fc1:LR24 (4th-order Linkwitz-Riley, 24dB/oct), fc2: BW18 (3rd-order Butterworth, 18dB/oct)
• Input gain: L-ch: -3.6dB, R-ch: 0dB
• Output gain: HIGH: +5.3dB, MID: -0.5dB, LOW: +17.0dB
• Limiter: off
• Polarity: HIGH: normal, MID: normal, LOW: normal
• Phase: 0deg
• Time delay: LOW-L: 0.06msec, LOW-R: 0.06msec (when tweeter is positioned at P0, HIGH-L: 0.52msec, HIGH-R: 0.52msec)
• Equalizer: GEQ: off, PEQ: off
• Infrasonic filter: 18Hz (BW12)
• Input clip level: +28dBu (=19.5V)
• Output clip level: +8dBu (1.95V)
• Other settings: default

From here, the outputs of each band of VENU360 are denoted by LOW, MID, HIGH.
The suffixes '-L' and '-R' denote the left and right channels, respectively (e.g. LOW-L, MID-R).

As in the previous revision of Gaudi II, woofer amp is directly connected to LOW of VENU360. In this manner, woofer amp generates a little noise. To reduce this, the attenuator of woofer amp is turned to by -12dB.

Please see the report on SS-309 for the details of the tuning.

As described in the report, I tried out the three positions of tweeter.

P0 (Position #0) Seemingly most adequate position	P1 (Position #1) Tweeter is lifted by spacers. It's aimed to reduce distortion caused by diffraction.	P2 (Position #2) Tweeter is aligned with the squawker driver. Time alignment is achieved w/o using the delay feature of the crossover network.

I selected P2 in the evaluation of SS-309B, but I liked P0 after listening to various kinds of music. I placed tweeter at P0 for a month or two.
Finally, I selected P2 again.

Measurements

This measurement was carried out according to the method specified in the section "Loudspeaker Mesurement" of the supplemental information "Method 2: Measurement Methods".
In the quasi-anechoic measurement, the distance between the woofer and microphone in the near-field measurement is d=13mm, and the merge frequency, fm=460Hz. The window width of impulse response is -2.0 ~ +2.5msec.

I was so interested in difference in frequency response among the different tweeter positions. I made measurement for each tweeter position. The data shown below are quasi-anechoic frequency response for P0 and P2 (L-ch only). All the measurements were carried out after time alignment was made.

Tweeter position #0 HIGH: Inverted	Tweeter position #2 HIGH: Inverted
Quasi-anechoic Frequency Response P0, Inverted Nearly flat	Quasi-anechoic Frequency Response P2, Inverted Fluctuation between 1kHz~3kHz is larger
Sine Wave Response at 6.3kHz (=fc2) P0, Inverted The waveform is distorted. Tweeter position #0 HIGH: Normal	Sine Wave Response at 6.3kHz (=fc2) P2, Inverted The waveform is distorted a little. Tweeter position #2 HIGH: Normal
Quasi-anechoic Frequency Response P0, Normal Dip appears at fc2	Quasi-anechoic Frequency Response P2, Normal Deep dip appears at fc2
Sine Wave Response at 6.3kHz (=fc2) P0, Normal Fine waveform	Sine Wave Response at 6.3kHz (=fc2) P0, Normal Distorted waveform

Looking at the frequency responses, the polarity of HIGH should be inverted regardless of the tweeter position. When it is normal, the deep dip appears at fc2.

Interestingly, however, as for P2, SQ (sound quality) is much better when the polarity is normal. The inverted polarity makes music sound awful.
On the other hand, as for P0, treble response is poor when the polarity is normal, just as the frequency response implies. The inverted polarity is apparently better than normal, although the waveform at fc2 is distorted.

In summary, in terms of perceived sound quality, with P0, HIGH sounds better with inverted polarity, whereas with P2, it sounds better with normal polarity. Now, as for which is better between P0 with inverted polarity and P2 with normal polarity, it's subtle.
After I carried out many listening trials by using various music sources, it have determined that P2 with normal polarity is better.
For further details, see the "Self-Evaluation" section.

Upon further consideration of the relationship between measurement data and sound quality, I have come to realize the following points. Frequency response measurements only gauge sound pressure and do not indicate whether distortion is occurring. In the case of multi-way speakers, distortion is prone to occur in crossover regions. However, frequency response alone does not reveal whether distortion is present, necessitating the combination with other data.

While observing waveforms seemed to be the most straightforward and effective approach, I have learned not to rely solely on this method. Without careful consideration of measurement conditions, one may obtain inaccurate data. In this instance, I believe it would have been essential to observe not only the waveform at the crossover point but also those of the surrounding frequencies.

Time Alignment Adjustment

For details of time alignment adjustment, see the section "Time Alignment Adjustment" of the supplemental information "Method 2: Measurement Methods".
Now I am accustomed to this method. I didn't spend much time and labor for this adjustment.

The figures below show the waveforms of the speaker in the adjustment (L-ch only).

E1: Input of the crossover network, E2: Output of mic amp

The waveform is much finer compared to the previous revision, because the ringing of the squawker is dramatically reduced.

 

Characteristics at Listening Position

For your information, characteristics measured at the listening position are shown below.
Here is frequency response.

The response drops at around 120Hz and 2kHz, though, the curve is rather flattened compared to the previous revision. The ripples occurs probably because of standing waves in the room. I haven't installed diffusers in the room yet.
The bass response gradually drops below 60Hz as in the previous revision. The response is reduced by about 2dB because the duct of the woofer box is stuffed (closed-box configuration).

The following figure indicates frequency responses of each band (R-ch only). The response in each band was separately measured, and the curves of each band are superposed in the figure. It clearly shows the characteritics of the bands (tweeter, squawker and woofer) since there is no interference each other.

The figure below is the waterfall chart of the right channel.

 

Trial Listening

I did trial listening each time I changed the settings. I used 2xHD "Audiophile Speaker Set-up". The format is DSD5.6M (DSD128). I mainly played performances of jazz combos with and without vocals in the album.

---

Room Acoustics

The room acoustics hasn't been changed since Rev.2.03. I continued to use the acoustic goods I had employed for Gaudi with some exceptions.

---

Self-evaluation

Improvements made to the previous revision have indeed enhanced the sound quality. While I already felt that the previous revision had the best sound quality in Gaudi's history, this time I have achieved even better sound quality.

Because Gaudi II is now able to faithfully reproduce beautiful sounds, I have recently become addicted to easy listening music.

However, if I were to evaluate it at 100% satisfaction, I wouldn't be quite there. I would rate it "B". This is because with the current system design, I feel that achieving the ultimate goal of super Hi-Fi, super Hi-Res seems unlikely.

I will explain in order below.

Sound Quality

It can be said that we have achieved the highest sound quality in Gaudi's history, but I still sense what is known as "horn coloration."
The Classic Pro H4401, which I adopted to replace the Fostex H400, has less ringing and achieves a clearer sound quality. However, there is still a slight sense of discomfort, or rather, it sounds a little different from the natural sound of instruments. This is due to the "horn coloration."

Regarding the position of the tweeter, as mentioned in the "Measurements" section, I was undecided about whether P0-Inverted or P2-Normal would be better.

With P0-Inverted, the sound initially appears to have less distortion and sounds clearer. However, the sound images were slightly out of focus. Additionally, depending on the source, the bass response is not enough and the highs may come across as noisy.

P2-Normal, while conceding a bit in terms of clarity, offers sound images in focus and stable stereo imaging. Furthermore, it offers good bass response.

I concluded that P2-Nomal is better. I cannot provide a theoretical explanation as to why simply changing the tweeter's position and reversing the polarity can affect the bass response. However, I have experienced many times before that the bass response can change based on the amount and quality of harmonics in treble region.

There is one more thing that concerns me about the sound quality. As I pointed out in the previous revision, when playing analog records, the analog feel is not perceived. There is no depth in the soundstage, it feels flat. It's as if I'm listening to a CD. Since Gaudi II is a system for listening to analog records, this is a problem.

I believe the cause lies in the low resolution of the crossover network VENU360. A sampling rate of 96kHz is too low. Those who use digital oscilloscopes would likely agree with me, but to accurately view the waveforms of audio signals, a sampling rate of at least 300kHz is necessary. Ideally, it should be over 1MHz.

Problems about Horn-type Squawker

Through the SS-309B project, it was revealed that horns have a greater impact on sound quality than drivers. Additionally, it was found that the shape of the horn is more important than material differences such as wood or resin (refer to the webpage on SS-309). Presumably, round horns may have less distortion than square ones. Avant Garde claims to have adopted a circular horn to eliminate horn coloration, which may indeed be the correct approach. However, such horns are not commercially available. If one continues to insist on using horns, he/she must design and build the horn for himself/herself. Fortunately, it is now possible to create horns using 3D printers. Personally, however, I have lost my attachment to horns.

In the SS-309B project, dome squawkers (Dayton Audio RS52FN-8) were used for comparison, and these proved to be quite good squawkers. At least according to the measurement data, both frequency response and transient response are superior to ED3402 + H4401, and the sound quality is also quite satisfactory. With proper utilization, it is certain to achieve sound quality exceeding that of ED3402 + H4401.

In my teens, I was convinced that horn speakers were superior to direct radiators (cone or dome types), and this belief persisted until recently. I particularly believed that the advantage of horn-type speakers lay in their superior transient response. Although their frequency response was not outstanding, I believed that horn speakers were better because they excelled in transient response, which has a greater impact on sound quality. Now that it's clear their transient response is inferior to dome speakers, I no longer feel attached to horns.

Horn-type speakers are inherently difficult to use. They cannot cover a wide frequency range, and when you employ horn squawkers, special measures are needed to achieve time alignment. There is also an issue of distortion likely occurring in the crossover region between the tweeter and the squawker. Another problem I recently noticed is the need to lengthen the distance between the speakers and the listening position. You need at least 3 meters to create a good soundstage. Although my house is made of wood, it is quite live. Placing the speakers further away increases the influence of first reflections from the side walls. Currently, I am barely maintaining a distance of 3 meters. Ideally, I would like to follow the original system design and set it at 2.6 meters, but this is impossible with horn speakers.

Appearance

I was worried that the appearance would deteriorate when I replaced the wooden horn (Fostex H400) with the resin horn (Classic Pro H4401), but it's not as bad as I feared. My friends even said it looks quite good. Surprisingly, my wife commented that it looks more upscale than the wooden horn, which was unexpected. It seems that without being told, people don't assume it's made of resin. It's amusing to think that a resin horn costing 1,980 JPY looks more luxurious than the wooden horn I bought for nearly 60,000 JPY.

Tweeter position - P0	Tweeter position - P2

User-friendliness

I improved the loudspeakers and redid the system-level tuning, so the usability, such as operability, remains the same as the previous revision.

By attaching the woofer amplifier to the woofer box, space became available on the TV rack where the woofer amplifier was previously placed.

---

Wrap-up

Although Gaudi Rev.2.2 achieved the highest SQ (sound quality) in Gaudi's history, I was not completely satisfied and rather disappointed. While there is still room for improvement, I don't believe that the sound quality will improve significantly even with further improvements. At this stage, I feel that I have already reached the limits of this system. To achieve the ultra-Hi-Fi, ultra-Hi-Res that I aim for, I feel the need to redesign the system.

For half a century, I believed that horns were superior to direct radiators, but now that belief has crumbled. In particular, a horn-loaded squawker is cumbersome to use, and difficult to suppress distortion occurring in the crossover region with the tweeter. Because the transient response, which is crucial for SQ, of the horn squawker is inferior to direct radiators, it is not suitable for the ultra-Hi-Fi, ultra-Hi-Res that I am aiming for. Another problem is that it doesn't suit the room. Generally, horn speakers require a large space. It may be too late to say this now, but the room where Gaudi II is installed is too small for horn speakers.

On a side note, I recently came across information that the popularity of wooden horns is increasing. Personally, I have decided to abandon them, but I can understand the feelings of those who desire wooden horns. They look cool and, although not Hi-Fi, they produce a unique and charming sound. They have the charm of so-called instrument-like speakers. When the Fostex H400 is fed with a pulse waveform, it generates long ringing, which creates a kind of reverb effect and sometimes makes vocals sound more beautiful. On the other hand, it can have the adverse effect, like making piano sound less like a piano. Overall, it's not modern high-resolution sound but rather a gentle retro sound to the ears. It might sound very good to retro enthusiasts.

This is just my intuition, but if I were to make full use of the D1405+H400, I think the system recommended by Fostex would be the best. It's a system composed of W300A, D1405+H400, T925, and passive networks. I think the passive network is better than a multi-amp system where the flaws of the LS units become apparent. The filters for the woofer and squawker are a 12dB/oct type, and only the filter for the tweeter is a 6dB/oct type. By overlapping the sounds of the tweeter and the squawker, the lack of high frequencies could be compensated.

For the next-generation system (Gaudi Ver.3.0), I'm considering a 2-way speaker system adopting cone-type LS units for both the woofer and tweeter. I want to make a decision after careful consideration.

 

---