By Pete Basel    9-12-2008
Classic SPICA TC-50 Loudspeaker Restoration
TC-50 Pair in Black - Serial #13593 Service Code H3.7-8
I have been interested in the SPICA TC-50 audiophile speaker since about 1984 when I became aware of their interesting and different design. I simulated the TC-50 system in 2004 using CALSOD, found a few areas where it could be improved (without significantly altering the transfer function), and purchased a pair in 2007. I have an interest in some of the legendary classic speakers and have restored several of my favorites. I've been replacing the non-polar electrolytic capacitors that are often used, with film types and a series resistor to duplicate the equivalent series resistance (ESR) of the original capacitor. These pages cover the restoration of a pair of the later SPICA TC-50 systems. The owner (MR) has asked me to restore this pair using this method of matching the original capacitor ESR. He is also interested in the following modifications:
- Tweeter Protection Crossover Mod
- Woofer Crossover Resonance Mod
- Additional Cabinet Brace
- Enclosure Damping
- Air Core Crossover Inductors and Star Wiring (under consideration)
- Bi-Amp Input Terminal Cup
- Completely New Outboarded Crossover with Air Core Inductors
I was concerned about shipping these speakers but the owner was not; I was surprised to find them triple boxed when they arrived. The inner two boxes were the originals from SPICA. These are a very clean black pair and appear to have never been worked on.
Close inspection showed few flaws. One felt pad is somewhat low on the baffle by about 1/8". The woofer in system A has a small blemish in the cone treatment, nothing serious, might have been from the time of manufacture. The surround for this woofer is a bit out of round which I often see in other TC-50s. I believe that this was simply due to a production tolerance as to where the edge was glued on the frame and do not expect it to be an issue as long as the voice coil does not rub.
Initial Listening Test
MR's TC-50s and my own pair were set up with an A/B switch to allow them to be compared. One goal was to see how similar they are given that both are later pairs with high serial numbers, and another - simply to get a subjective baseline of MR's pair. It should be noted that the tweeters in mine have been repaired with the newer, and slightly different Audax/ACC dome replacements. My pair are clearly brighter sounding, estimated to be about 2 dB. Subjectively, the initial impression is that MR's just sound more believable - like real music. MR's pair sound more focused, with a more believable center image, warmer in a way that makes them more natural sounding. The less mid bass or warmth with my pair imparted a slight hollow sound to the music, particularly vocals. All of the TC-50s were similar and shared a family sound.
It was fairly obvious that one of MR's speakers was louder to an extent that I felt that the balance control needed to be nudged a bit. TC-50s are sold in matched pairs sharing the same serial number with them marked A and B to distinguish the two. Most of the tonal observations and comparisons were done in mono only using MR's more efficient A speaker.
Pre-Restoration Baseline Measurements
The TC-50s as received were placed in the center of the room, back to the floor facing up, and a Neutrik 3382 measurement mic placed 1m away in line with the top of the woofer frame. A MLS type frequency response measurement was taken using the Liberty Audio Suite (LAUD) measurement system. The time markers were adjusted on the impulse response to remove the time of flight delay. This is required to obtain a true representation of the system phase response. The first frequency response (FR) plot here is of system A, with 1/3 octave smoothing and a 10 dB per division scale. This is typical of marketing literature since it shows a much smoother response than when no smoothing is used along with an expanded scale. Often even more smoothing is used, 1/2 octave or even full octave.
Fig. 1    System A - MLS Frequency Response With 1/3 Octave Smoothing, Grille Off
The rest of the plots do not use smoothing and are shown with a 5 dB per division scale. The measured amplitude and phase response are shown below for system A, grille off, 1m on the optimum axis in line with the top of the woofer frame. Edit: Rereading the SPICA TC-50 manual reveals that the official optimum axis is 7.5" up from the base of the speaker. The top of the woofer frame is about 9" up which is closer to the tweeter axis. Refer below from Figure 8 on to see these responses retaken at the true optimum axis. Note that these measurements are only valid above 400 Hz or so due to the necessary windowing of the time domain data which is standard with MLS testing. Agreement with both simulation and published specifications is good:
Fig. 2    System A - MLS Frequency and Phase Response, Grille Off
Same plot as above but for system B. There are significant differences as was noted in the listening test.
Fig. 3    System B - MLS Frequency and Phase Response, Grille Off
The differences are easier to see with the responses overlaid as shown below. There is about a 1.5 dB lower output from B (red) around 600 Hz, and a 3.5 dB difference around 1.5 kHz. The tweeters match quite well in the passband but one rolls of a bit sooner than the other. The woofer difference around 600 Hz might be a cone mass, or motor strength difference, however the higher frequency output above 1 kHz is strongly influenced by how the cone breaks up, and the effect of the coating treatment. One has to wonder if this is a result of aging, a mismatch when fabricated, or a result of in use damage. Tests of the woofer Thiele and Small (T&S) parameters will provide additional insight:
Fig. 4    Systems A (white) and B (red) - MLS Frequency Response, Grille Off
A FR plot of system A with the grille on (red) and off (white) is shown below. The grille actually improves the response. It is likely that the beveled edge reduces diffraction although one might think that the felt should absorb most of the acoustic energy before it reaches the edge:
Fig. 5    System A MLS Frequency Response Grille On (red) and Off (white)
The step responses are shown for systems A and B below as derived from the impulse response. The rise time or speed of the leading edge is an indicator of the high frequency extension of the system; infinite bandwidth is required for infinite rise time on the leading edge. Finite bandwidth causes the rise time to be slower. There is ringing after the leading edge that is related to the nature of the high frequency (HF) rolloff of the system; more ringing is seen with a steeper rolloff. The length of time during which the flat top is maintained is a function of the low frequency (LF) extension. Response to DC would be required to maintain a flat top for a perfect step response. The shape of the decay back to zero, and the ringing or lack there of about zero is a function of the low frequency rolloff of the system. Most traditional multiway loudspeaker systems cannot produce anything resembling a step response as a result of their non-linear mid-band phase response. The TC-50 does quite well considering that most speakers do not produce anything close to a step waveform. An example of a 3-way speaker with a traditional crossover is shown here in section F step response plot e labeled 3-way. The measured step response for MR's TC-50 systems A and B are shown below:
Fig. 6    System A Step Response
Fig. 7    System B Step Response
The vertical polar response is shown below as a family of curves for microphone positions 15 and 30 degrees above and below the optimum listening axis. I read the TC-50 owner's manual again before taking these measurements and noticed that the optimum listening axis was half way up the enclosure, or 7.5" according to the manual. I checked this against the guideline to use the top of the woofer frame as the reference axis and found this point to be about 9" up rather than 7.5. Retesting on the 7.5" high axis showed much improved responses which are included in the 0 degree reference below, and in all the curves retaken that follow. I was going to replace the first set of curves above, however they show how sensitive the system response is to listening height and have been left for this reason.
Note that the on axis response shown in red is much flatter than the measurement taken in Figure 2. The dip at the crossover frequency is not as deep, and the high end is no longer elevated. This curve is within +/- 2.5 dB from 400 Hz to about 15 kHz with much of this error taken up by the dips at 3 and 5 kHz. The crossover behavior is very similar to a first order system where the off axis response is asymmetrical with the drivers moving more in-phase for the down direction and out of phase going up. At 15 degrees up, shown in white, the drivers are nearly ideally out of phase at the 3 kHz crossover as can be seen by the deep notch. Moving 30 degrees up brings the drivers back in phase and is nearly on axis to the tweeter showing about 5 dB more output above 8 kHz and better extension to 20 kHz. Moving 15 degrees down, shown in maroon, the tweeter output is attenuated because it is far off axis due to the sloped back baffle, but the dips around the crossover region are nearly gone. Thirty degrees down in green shows more attenuation of the tweeter output.
Fig. 8    System A Vertical Polar Response +/-15 and +/-30 Degrees
Frequency and Phase response measured on the factory optimum (7.5" up) axis for system A. Note that both the amplitude and phase response are significantly better than in Figure 2 above. Agreement with the factory measurements shown below in Figures 19, and 20 is also much better.
Fig. 9    System A MLS Frequency and Phase Response Grille On
Frequency and Phase response measured on the factory optimum axis for system B. The phase response is remarkably flat, better than the factory measurements and I believe that this is a result of the out of spec woofer. The amplitude response is also very good, however it shows droop from 500 Hz to 2 kHz as if there is too much baffle step compensation.
Fig. 10    System B MLS Frequency and Phase Response Grille On
Overlaid frequency response curves for systems A (red) and B (white) measured on the factory optimum axis are shown below. The tweeters showed very good agreement in the previous comparison measured about 1.5" higher, and it should be noted that one felt pad is significantly lower than the other, by about 1/8". I believe that this offset has more impact on the response at the lower measurement point. The B woofer is consistently lower in output as is seen in the white curve:
Fig. 11    Systems A (red) vs. B (white) MLS Frequency Response 7.5" Axis, Grille On
Overlaid frequency response curves for systems B with the grille on and off measured on the factory optimum axis are shown below. The results are similar to the previous measurement showing a better response with the grille on:
Fig. 12    System B MLS Frequency Response 7.5" Axis, Grille On (white) vs. Off (red)
Nearfield Woofer Measurements
The woofer low frequency response was measured using the nearfield method where the microphone is placed .25" away from the center of the dust cap. A swept sinewave test is used rather than MLS because it has been shown to be more accurate for low frequency measurements.
The response below shows excellent agreement with Thiele and Small theory taking into consideration that the system crossover network provides baffle step compensation. Marker two shows the woofer output to be 5.6 dB down at 1 kHz as would be expected for a system with full baffle step compensation (BSC). Also note that the output is -10 dB at 40 Hz and -22 dB at 20 Hz showing perfect agreement with the expected 12 dB/oct rolloff of a 2nd order system. The nearfield measurement confirms that the driver and crossover are operating properly, however being near field the effect of the baffle is removed as compared to a free space anechoic measurement. This makes it difficult to combine with the MLS measurement.
Fig. 13    System A Woofer Nearfield Frequency Response
Measurements for system B are similar, however the reduced output is seen again, with more response droop at 1 kHz. Marker two shows the woofer output to be 6.9 dB down at 1 kHz, or 1.3 dB less than system A. It is interesting that the system response shows a much larger difference. Perhaps part of the problem involves the low end response of the tweeter due to a problem with the tweeter and/or crossover network. Marker one shows the system to be about 3 dB down at the system resonant frequency of 65 Hz providing good agreement with T&S theory:
Fig. 14    System B Woofer Nearfield Frequency Response
Fig. 15    System A (white) vs. B (red) Woofer Nearfield Frequency Response
The system input impedance was also measured using the swept sinewave method and are plotted below. The system Thiele and Small parameters were also measured with the following results:
- System A T&S parameters: Fc = 69.5 Hz, Qtc = .79, Qe = 1.17, Qm = 2.45, Rdc = 3.67 ohms
- System B T&S parameters: Fc = 65.1 Hz, Qtc = .86, Qe = 1.24, Qm = 2.85, Rdc = 3.53 ohms
Fig. 16    System A System Input Impedance
Fig. 17    System B System Input Impedance
Fig. 18    System A (white) vs. B (red) System Input Impedance
Measured Responses vs. SPICA Measurements
It is interesting to compare these acoustic measurements with those provided by SPICA in the TC-50 literature. The FR and phase response plots are shown below. Please note that these are for the earlier and slightly different version of the TC-50. The response below 300 Hz is probably more a function of the windowing used to process the data than the actual TC-50 response. The trends do look similar, however the newer TC-50s show a rising high end. The newer version was also tested by Stereophile and the rising high end was noted in that report as can be seen in Figure 4 of this link. The HF response is highly dependent on the measurement point which was noted by Stereophile, and confirmed here in the above measurements. Also note that the low frequency response does not rolloff at the expected 12 dB/oct for a sealed box system. This is probably due to the time windowed measurement method used where the low frequency response shape is more a function of the time window rather than the system under test:
Fig. 19    TC-50 Frequency Response as Published by SPICA
Fig. 20    TC-50 Frequency Response as Published by SPICA
Fig. 21    TC-50 Input Impedance as Published by SPICA
Front Baffle Felt Pad
The felt pad has to come off, obviously to get inside the boxes. I have been curious about the felt as the cutout seems to be too small in the newer version of the TC-50s. The following curves show the system A frequency response with grille on (red), grille off (white), grille and felt off (yellow), and finally with the tweeter felt blocked (maroon) with business card stock. No acoustic absorber, such as felt is perfect, and I wanted to see the effect of reflections the same distance away from the dome just to see where the response differences show up. Grille on and off has been covered previously, however it is interesting that the 5.5 kHz dip moves up to about 6.5 k and is of lower amplitude with the felt removed - yellow curve. Further, the 3.5 kHz dip moves down to about 2.5 kHz with the felt removed, this is probably simply due to the increase in output amplitude and is not a reflection off the felt.
The reflective card material causes a dip at 10 kHz (maroon curve) not seen in the other responses, this suggests that the felt has good attenuation around that frequency. The dips from 5 to 5.5 kHz are nearly the same as with the felt and are not seen for the no felt case suggesting that these might actually be caused by reflections from the felt. The large dip at 3.2 kHz is simply due to the phase relation at the crossover point and not a result of the felt alone.
The earlier TC-50 had a larger rectangular opening around the tweeter and that makes more sense in my view so as to not block the off axis high end response. It would be interesting to know the reasoning behind the new smaller opening. I understand that the new revision of the tweeter was more efficient and more attenuation was required however this could have been accomplished with a pad in the crossover.
Fig. 22    TC-50 Frequency Response Grille On (red), Off (white), Felt Off (yellow), Felt Blocked (maroon)
The following picture shows how the business cards were placed to block the felt absorption and cause reflections.
Fig. 23    TC-50 With Business Cards Around Tweeter
More to come as time permits.