Just a reminder – having accepted the Terms/Conditions on the pop-up, you have agreed that:
- The page is presented entirely as an opinion piece;
- All opinions on the page are false unless proven otherwise;
- You, your assigns, heirs, agents and any third party viewing the page as a result of your access to it (and for whom legal rights against the site owner are vicariously yours In Toto) forfeit all legal rights against the site owner in respect of anything published on the page;
- © You will not copy the page or any of its content; and
- If you do copy anything in breach of Condition No. 4 and publish it elsewhere, then liability is yours solely and to the outright exclusion of the site owner.
Tweeter failure in a Chario brand “Reference 1000” bookshelf speaker led to a major overhaul of the pair after a plethora of appalling defects became evident during a strip-down. The speakers had the potential to be fine, but were let down very badly by cost-cutting, stupid crossover features and apathetic assembly, most of which I’ve addressed some 20 years later – short of ripping out the crossovers and converting them for active bi-amplification that is.
The speakers are from around 2002 and are a great “reference” for the kind of audiofool garbage I purchased in a past life and would never buy again. And there had been around 1000 examples of that! They were expensive and replacement with anything half decent would be too, but since I wouldn’t be seen dead in a hi-fi store, a repair was in order after one of the original tweeters went inevitably open-circuit.
The tweeters were very cheaply made Scan-Speak model D2905/9000. These were replaced by more modern R2604-833000 units which hopefully aren’t as poorly made.
Shoddy assembly of the speakers in general also became evident once the drivers were removed, so the repair became an major overhaul which in turn revealed among other things, a ridiculous flaw in the crossover design that was likely catering to tube amp audiofools, so corrections were necessary left, right and centre. A bit like owning an Alfa Romeo or Fiat I guess. They perform well for a while, then you get what’s coming.
They look identical to speakers called “Hiper 1000” (yes – spelt that way), and others called “Premium 1000” look similar. The word “Reference” appears only on a barcode sticker at the back.
I purchased these new while working in Hong Kong when I had no free time to build things myself. Apparently they were manufactured in some kind of sheltered workshop specialising in spaghetti electrics. Unreliable, non-soldered crimp connectors barely hanging from the driver terminals, wires flapping about against the port tubes to explain the lovely added vibrato, and binding posts expected to maintain electrical continuity with the crossover PCB via plain nuts tightened down onto un-plated brass pads without so much as toothed washers! And it got a whole lot worse when I had to dig deeper. The crossover PCB was mounted on a soft pad that covered just part of it. A wood screw tightens a floating end of the PCB down to the rear baffle flexing the board against the pad. This causes a resistor wire to detach from its solder joint! Also, the binding post contact pads on the PCB distorted badly as the nuts were tightened. The whole PCB distorted around the posts as there were no spacers to prevent their solder joints from pushing hard against the rear baffle. And there was an inexcusable flaw in the woofer circuit that required deletion! A typical example of audiophile-targeted incompetence IMO! “Reference” indeed. Bah!
I found a data sheet for the tweeters dated 1989 so they were quite ancient even by 2002. Perhaps they were snapped up as NOS. I have no idea. Anyway, they were designed such that movement of a solder pads formed integrally with the terminals such as by pushing a crimp lug over them as intended breaks a glue joint. Resultant free movement of the pad guarantees breakage of the hair-thin multi-strand wires to the voice coil! Also where these wires connect to the enamelled voice coil there is no solder, so glue covering the joint corrodes the connection leaving an open or high impedance circuit. They measured pretty badly when they worked, but complete failure was just a matter of time.
A soft black foam gasket ring sat on top of the white plastic ring to sandwich the solder pads, but being soft, that’s to me either grossly incompetent engineering, or deliberate planned obsolescence. “Conveniently”, they sold replacement domes/voice coils, but if you could still get them (which you can’t) they’d be more expensive than a new tweeter. The tweeters have been thrown on the junk pile.
The woofers appear to be customised and hand-doped Audax 5.25 inch units that I can’t identify positively, so thankfully they’re intact.
One might assume that the enclosures, ports and crossovers were tuned specifically to the woofers, and the baffle recesses are specific to their peculiar shape¹, so it’s a good thing they still worked. The port opens through the floor and reaches up almost to the top:
- The peculiar driver shape actually serves a purpose in allowing the tweeter to be mounted more closely for better lobe steering, but in the case of these speakers, I doubt that the crossover was designed to any kind of precise driver alignment.
It was too difficult to photograph the crossovers in situ (but I was forced to remove them later – see below), but they seem to comprise at least a second order filter on the tweeter going by the coil and capacitor count nearby the tweeter wires.
The binding posts on the back pass through 6mm holes in the MDF and through the crossover PCB. Electrical connection was solely via standard nuts threaded onto each post. Those bore down on square brass pads on the PCB – just visible in a photo below.
Yes they were from the “bi-wiring” era, or is that still a thing? Complete nonsense of course with passive crossovers, so the shorting bars have always remained in place.
I added stainless spring washers, toothed washers and replaced the nuts with M6 Stainless Nylocs for reliability. Also cable-ties now secure the speaker wires to the ports to stop the unwanted “vibrato”.
Amplitude response of speaker with the original tweeter
Fortunately I measured the acoustic output of the speakers a year or so before the tweeter failure:
The quasi-anechoic gated near-field measurement is as raw as I can make it and shows a pretty ragged tweeter response (±4.5dB). The woofer seems quite reasonable at least as far as the measurement method allows. That inexplicable drop suggests a crossover frequency of around 2 to 2.5kHz. I found references suggesting anywhere from 1450Hz to 2.5kHz on the ‘net for other Chario models with the same drivers, but the precise number is of no consequence. The measurement suggests that the tweeter output was 2 or 3 dB down and that there were obvious severe problems. These had previously been addressed to some extent with a multitude of JRMC PEQ filters generated in REW, but the overall drop-off relative to the woofer SPL meant that was pretty much a lost cause because high Q “boost” filters are banned in my book.
Choice of new tweeters
I didn’t care about a perfect match for the originals as they were clearly nothing to write home about. There is something that looks identical called a D2905/930000 (and other variants of the same driver) that would drop straight in, but nothing I could find had comparable T/S parameters and they all probably have the same terrible internal construction as my originals given that they look identical. And they’re unjustifiably expensive.
The nominal impedance needed to match, and the quoted efficiency needed to be 2 or 3dB higher. Also since I had no schematic for the crossover network to confirm a crossover point (nor the motivation to measure it electrically), I needed a tweeter with a low resonant frequency and good LF extension for an adequate safety margin given I had no intention of changing the crossovers. At first impression, getting them out looked like risking cracking the PCBs anyway, since they seemed to have some kind of double-sided adhesive tape or other adhesive behind them. This turned out to be a false assumption on my part (see below). And of course the faceplate had to be the same diameter to fit the recess. Anyway it all lead to the Danish made R2604-833000 ring radiator tweeter. It’s 2dB more efficient (92 vs. 90) which is ideal and if its output is too high a shelving filter could be keyed into JRMC easily enough. They’re probably built just as crap-like as the originals, but these are hardly high fidelity speakers warranting an OCD spending frenzy. And indeed upon arrival there was immediate confirmation in a faceplates made of plastic. Euch!
- Aside: There are Vifa and Peerless versions of this tweeter with different model numbers. They look identical from the front, but are made in China and don’t have the double magnet. I’ve placed a personal ban on Chinese products where alternatives are available – even if more expensive. China is an all too powerful international bully run by low-witted descendants of Mao’s generals. They need to be pulled inline with international values and smarten themselves up before I buy anything Chinese with a clear conscience.
Installing the new tweeters
Of course nothing is meant to be easy, so a “simple” tweeter replacement soon becomes a major overhaul!
5 screw holes instead of 4, and connection terminals that protrude beyond the baffle cut-out aperture requiring filing of recesses in the MDF.
Another pet peeve of mine is the tight arse or bone lazy use of wood screws to mount speaker drivers into MDF. It’s pathetic! Wood screws are for wood! Tee-nuts ← should always be installed in the back of an MDF baffle before assembling the enclosure and machine screws must be used. The way I usually do it is shown on the right. →
Woofer screw holes drilled out to 6mm ∅ for M5 bolts and tee-nuts. New 5mm ∅ holes for the tweeter’s M4 bolts and tee-nuts. The old holes were filled with PVA and cosmic dust then sanded smooth:
The crappy white polyfill was removed in favour of egg crate foam cut to the same shape plus a small amount of wool.
Speaker wires got new crimp connectors both crimped and soldered, then covered in heat shrink tubing:
A clamp presses in the tee-nuts:
Plywood block with cardboard padding to protect the veneer.
Beeswax applied to veneer.
Almost forgot the gasket, but tweeter installed:
Faceplate about 1mm thicker than the old one, but that’s just too bad. Why do these driver manufacturers insist on inconsistency? 4 holes, 6 holes and even 5 holes and plates of different thicknesses! What a PITA! At least they kept the diameter the same. There’s the only praise they get.
All together again:
With MACHINE SCREWS and TEE-NUTS like an engineer would do it!
Back where they belong
Attached to a Pioneer A-30 integrated amp (not Class-D!) set in Direct mode (as a power amp) with an old Sony Vaio laptop and a Focusrite Solo USB audio interface (that equals ANY highfalutin audiophile DAC‘s sound and cost under AUD150 on special):
All the old tweeter resonance-correction filters in JRiver’s DSP Studio were deleted – just leaving a 17Hz high pass and a 50kHz low pass filter in place (the latter because years ago I foolishly converted a bunch of SACDs to 176.4kHz FLAC not realising they contained a mess of low level noise and nothing more over 50kHz).
Do they work?
Of course they do! As a super-precaution I did check the amplifier just in case it did something to the original tweeter which it didn’t. DC offset a mere 5mV and a clean (non-oscillating) signal into a dummy load seen on my oscilloscope. And they sound fine.
This is on-axis and gated at about 600mm. Can’t be bothered doing polar measurements.
- A note about the slight rise across 20kHz is under Crossover Topology below.
- A dip there suggests that 3kHz might be the crossover point, but I don’t think that’s the case. Under the heading Crossover Topology below, the -3dB point for the high pass section of the crossover is found to be 2.2kHz.
SPL numbers are uncalibrated (there is no way I’d ever sweep that loud!), but the scale is the same as for the old measurement. Looks better at ±2.5dB and the additional 2dB efficiency certainly came in handy, but to be fair, the original measurement (scroll up) had the grille cloth in place and this one didn’t. Same Earthworks calibrated mic used both times. There’s a broad droop of around 1.5dB between 2 and 4kHz that makes no sense.
Simple PEQ applied in JRMC flattens it to:
The goodness was short-lived, so the story isn’t over yet!
The new tweeter in one speaker started to make intermittent crackling sounds while music was playing. WTF! Swapping the speakers left to right relinquished the power amp, PC and DAC from blame. The crackle followed the speaker. Swapping the tweeters from the left to right enclosures showed it wasn’t a tweeter either. The crackle stayed with the right enclosure. Must be something between the binding posts and the driver. The crossover had to come out after all.
And there it is in all it’s “glory”:
It wasn’t double sided tape after all. Just a crummy round foam pad that’d been squished in there for 20 years:
And the further I looked, the worse it got! Who does this kind of thing?
The nameplate need not be metallic! I guess it was done purely for a “luxury” look. But why place a potential shorting bar right across the customer’s amplifier? The mm clearance might be OK elsewhere, but they’re only anchored into soft MDF! I think the answer is to add a little PVA when screwing the posts back in – just to firm up the MDF a bit.
But what struck me mostly was the terribly tarnished and bent nature of the brass pads where the binding post nuts bear down to make electrical contact. An inspection of the back of the PCB and inside the speaker enclosure explained the problem. There are two solder joints securing each brass pad to the PCB. When the nuts are tightened down, the solder joints press hard into the rear baffle, bending the pads.
No wonder the speakers had a history of intermittent low volume depending on the day. Sometime left was low, sometimes right, sometimes both! High impedance connections!
Here’s a photo taken later. A pair of dimples is seen clearly in the veneer alongside each post:
The photo shows added industrial nylon spacer washers with a spring washer underneath each. Under each spring washer is a flat stainless washer to prevent the sharp edge of the spring washer from sinking into the veneer and becoming ineffective. → The original nuts might have been repurposed under the PCB, but there was insufficient thread. Thackery washers were considered, but I didn’t have any and they’re too expensive. A 3mm thick neoprene tape was added at the far end of the PCB. These features are not original. The new hardware maintains a space between the PCB and the rear baffle when the Nyloc nuts are tightened down onto the toothed washers which bite into the new solder coating of each brass pad (see below). More importantly, the spring washers guarantee that the toothed washers always press hard into the solder, because the Nylocs aren’t going anywhere. This is called engineering! To have sold the speakers in their original state and also used the word “Reference” in their marketing, I’d be hanging my head in shame! 😳
- Aside: This reminds me of some “typical” audiophile bull dust marketing by a speaker manufactures called “Magico” or something. It went on about the “sonic glory” (or whatever) of their enclosures because they went to the extra trouble of veneering both sides of each panel. Well guess what? That’s how it comes from the suppliers! If not veneered on both sides the MDF stock would all warp!
OK settle down Ian and move on
Brass pads given the Scotchbrite treatment and a coating of solder – each filed flat after the photo was taken:
Many solder joints re-flowed and some hot melt glue added to stop things rattling:
A wood screw goes through that hole in front of the wirewound resistor to secure that end of the board to the rear speaker panel. As the screw is tightened, the board bends because the original foam pad under the board (which looked like it was grabbed from under someone’s table leg) falls short and the solder joint to the resistor lead breaks! The resistor was hanging loose in a dry joint. More ignorance or just plain indolence on the part of the manufacturer? Probably. 😆
The answer was a full sized neoprene pad and the screw now goes through it and the neoprene tape seen in a previous photo.
- Aside: I do not like passively crossed speakers! They rely on inefficient inductors, low tolerance/short lived electrolytic capacitors, high wattage/power-wasting resistors and usually have low order filter slopes due to the added expense and size of going any higher. Low order filters equals additional intermodulation distortion that is very audible. Due to these limitations, I’ve never studied them nor bothered designing one, so this rebuild is a learning exercise for me that will never be put to good use! Real hi-fi requires active crossovers and multiple amplifier channels. These speakers do not warrant the trouble in conversion however.
I traced out the crossover network and the tweeter section is a simple second order high pass with one exception. There’s a leading series resistor to attenuate output, but it has a 1μF MKT capacitor across it. This looked odd, so I modelled it in 5Spice.
The inductor values were actually measured and series resistance added per the measurements, but I modelled the driver as a simple 4Ω resistor just to get an idea. And it simulates like this:
Amplitude red / Phase blue
The 1μF capacitor causes the response to rise gently above around 10kHz. It’s up 1.5dB by 20kHz (and the full 8dB by 300kHz). Deleting the capacitor leaves the response flat across 20kHz at roughly -8dB. The capacitor would have been to counteract the falling response of the original tweeter per the Scan-Speak data sheet:
My own measurement actually confirms this with a slight rise for the new tweeter which has a flatter natural response, so I lifted that capacitor. Nobody can hear that high, so it was probably associated with a marketing gimmick.
Output of circuit with the capacitor deleted shows a -3dB point of around 2.2kHz:
The bass section is a 2nd order low pass with two peculiar sections.
Again I’ve used a simple resistor (R3) for the woofer driver just to get an idea.
Firstly, a parallel RLC network (at the top) forms a band-stop filter to manipulate the bass driver response no doubt to smoothen the natural behaviour of the driver in its ported enclosure.
Here is the electrical output assuming a low output impedance transistor amp (0.1Ω shown):
My guess is that the suck-down (dampened band-stop) at around 880Hz is massaging in some baffle step compensation and addressing a front baffle edge ripple in the frequency response. Since my microphone measurement shows a well behaved output in that region, I can only assume that this is working as intended despite not being applied to the tweeter signal. Any filter like that should be applied prior to splitting the signal! It’s idiotic on face value, but probably works to the benefit of decent acoustic summing with the tweeter on account of the phase shift that it applies.
But change the input impedance to 3Ω (as if using a valve amplifier) and the bass is pulled down by 3 or 4dB:
This of itself isn’t an issue because the tweeter level comes down by the same amount in the simulator (and no doubt any valve amp has its overall gain set to include output transformer behaviour), but there is an additional suck-down seen there at around 100Hz caused by a second very stupid set of components placed across the input terminals (at bottom left in the schematic) forming another band-stop filter – namely C3 and L4 with a resonant frequency of 94Hz. R6 represents the measured resistance of the coil L4.
The filter is useless and stupid!
With the low output impedance of a solid state amp, if you remove C3, L4 and R6 from the simulator, the output is identical to the first graph, but with the benefit of much lower current demands being placed on the power amplifier.
A woofer in a ported enclosure exhibits two impedance peaks. I’m guessing they’re at around 70 and 100Hz for these speakers. Maybe Chario’s “engineers” were attempting to counteract those in one broad brush so that the speakers didn’t sound boomy when used with valve amplifiers. But it’s an Elasoplast on a hernia. The series resistance of the coil (9.2Ω) over-dampens the Q and instead of targeting anything specific, it merely provides a broad bass reduction centred at 94Hz as shown. And although it might disguise (but not target) the two resonances when a valve amp is used, this simply shows one of the many failings of any valve amplifier when connected to woofers, it is obviously a completely ridiculous circuit for a solid state amp because it presents a low impedance load for nothing, so I disabled this section by lifting the capacitors making up C3.
So the speakers might now sound boomy when driven by valve amplifiers, but that obviously doesn’t matter half a toss because just like turntables in 2021, they’re pointless souvenirs of a bygone era (that people remain passionate and even emotional about for nothing) and I don’t waste my time with them. 😀
Upon looking at the electrical simulations, it appears that there ought to be a giant suck-out between the woofer and tweeter acoustic outputs. Also with second order crossovers like this, the wires to one of the drivers would normally be reversed to account for the 180° phase misalignment between high pass and low pass sections, but this is not the case for these speakers. At 2.2kHz (which might not even be the acoustic crossover point) the simulated phase misalignment for this crossover is around |-160°-70°| = 230°, yet no wires are reversed, and there is no serious anomaly across the crossover region in my microphone measurement to account for that and the 50° remainder. I doubt that they would have even considered acoustic driver offsets and suspect that the misaligned filters were the result of empirical development rather than actual engineering. Maybe later I’ll measure the separate output of each driver with the microphone using a loop-back timing reference to see what’s really going on. I’m suspecting they cross at around 1.5kHz where the high pass section is around 6dB down.
I’m not one of those “re-capper” obsessives. If they measure the same then they work the same and have no contribution to a “sound”, but the bipolar electrolytics which remained were very cheap (≈ 30¢) 85°C/63V Jamicon parts and were around 20 years old, so I decided to lift those and install new Nichicon 105°C/100V parts. And it’s a good thing I did because the Jamicons were all out of spec – some very high (a 47μF cap read 63μF). Where low tolerance capacitors are used in audio filters, they must be measured and matched, or else the circuit will not perform to design! Where tolerances are typically around ±20% for electrolytics, multiple caps must be purchased and only selected ones used! New simulations of the low pass section using the actual read capacitance values of the old Jamicons (rather than the labelled values) explains the broad 1.5dB droop in the measured frequency response between 2 and 4kHz mentioned earlier. The new caps might remedy that with any luck.
And the 10μF caps in the tweeter circuits were crummy 10% tolerance MKT parts and they didn’t even use the same ones for left and right speakers! Hey let’s just grab whatever there is in the junk box. No one will know right? WRONG!
So obviously they’ve been lifted. New inexpensive 220V 5% MKP (Kemet) parts were installed. Fortunately the PCB had a multitude of through-holes to accommodate caps of various footprint. No lamebrain “audiophile” parts for me! Jantzen, Mundorf and the like have a valid place only in strange fantasy worlds where money grows on trees and the salesman is your “friend”.
(Three capacitors left off each board as noted)
Last time! All back together.
And they sound as good as new. These little bookshelves now have a bit of extra kick to them. Moreover the Pioneer A-30 amp is doing its thing without being strangulated by a STUPID crossover flaw and high impedance connections!
More to come if I can be bothered measuring the acoustic response again. I’m kind of half interested in removing the shorting bars from the binding posts and using a couple of jumper cables to reverse the tweeter polarity. This would hopefully show a sharp null at the actual acoustic crossover point. Anyway I’m keen to see if that broad trough across the crossover region has filled in.
New microphone measurements show that the tweeter driven by itself has a dip at 3kHz. This in not shown in the Scan-Speak data sheet → which I have no reason to doubt. It’s probably related to the enclosure or baffle shape. Indeed I just found a “review” in one of those flowery language Stereofool magazines for a “Premium 1000” which is almost the same speaker and it has a 3kHz dip as well. The storyteller refers to it as a trough in “the presence region” whatever that is.
The acoustic crossover point is around 1.57kHz.
The inverted tweeter trace shows no sharp dip – only a slightly deeper broad trough indicating poor phase alignment between the drivers. It’s actually out by 72 degrees when not inverted, with the woofer leading. Also, inverting the tweeter causes an undesirable hump at 1.2kHz.
The normal summed amplitude trace has generally a ≅1.75dB improvement over the crossover region thanks to the new on-spec capacitors. Also the rise across 20kHz is flattened.
↑ Same scale as before. ±2dB.
END (or maybe I’ll measure horizontal off-axis, but I’m pretty sure it’ll be a dog’s breakfast).
… and yep:
These were done at a larger microphone radius of around 900mm. Previous traces were done at around 600mm.
The top trace is about 5° off-axis (line of sight to where I listen), but even by 15° it’s seriously compromised and a camel hump develops at 2.8kHz very quickly. Also, there is way too much off-axis roll-off above a mere 5kHz. Any EQ applied to the on-axis response will exacerbate the poor off-axis anomalies and therefore should not be applied (apart from a small cut at 5.8kHz to negate a resonance that’s consistent through the upper traces). These traces confirm my suspicion that the speakers were developed empirically to an on-axis response to suit a marketing strategy. This in apparent total disregard for off-axis performance.