#46 Tivoli Henry Kloss Model One teardown & repair

A client brought in this Tivoli Model One the other day, with symptoms of (lots of) noise. Once again, i went against the mantra of "don't turn it on; take it apaaaah't", but this time, for good reason - the thing isn't mine.

Interesting / uncommon feature is that it seems to accept both AC mains power, as well as a 12V DC input. So i suppose that lets you use it in your car, if your car's radio is busted, or..? Your guess is as good as mine.

Sure enough, even with the volume at minimum, the thing's noisy as hell. Think "ripping up cardboard boxes" type noises. With the volume above half-way, the in-between-radio-stations noise starts to overcome the abysmal "background noise", so it's reasonably safe to assume it's probably not from the receiver stage.

Slight side-note - just now, for bits & pickles, i googled "Tivoli Model One", just to get a ballpark feeling what these things retail at. First result, a local one, indicating about 200 (yes, TWO HUNDRED) buckaroonies.

[crickets]

Well, let's see what you're getting for that sort of cash. Full disclaimer though - i've had one of these apart a couple years back, ended up replacing the ceramic resonators in the receiver section, aaaaaand it ain't nothin' to write home about. But all in due time...

Bit of an adventure taking this thing apart, actually. Four screws in the back hold the front panel, and two more fasten the rear panel to the housing. As soon as the front comes off, it reveals a whole bunch of cables running between the front and back, and loooots of hot-snot (aka hot-glue or hot-melt-glue) on all the connectors. These latter ones are nice - it's an extra expense on the bill-of-materials, but it definitely eases servicing and no doubt (and arguably more importantly), assembly as well.

Then again, that expense "had to" come from somewhere - all single-sided boards, most likely FR-2. Most definitely not the woven-glass-fiber FR-4. Mostly through-hole construction, apart from the mixed nature of the audio-board. But still, $200-worthy? I don't know... How much of that is the brand-name-fee, i wonder? Odds are, oodles of it, but what do i know?

The front panel reveals the tuner section (connected to a pretty nice vernier dial, with a significant gear-reduction between the knob in the middle, and the tuning-indicator "skirt" around it), and a no-name (likely Shenzhen-special) 3" speaker. Granted, the enclosure is ported (aka "bass-reflex"), so it's gonna sound beefier than a phone, but the laws of physics (and economics, surely) will only let things go so far.

On another "layer", a second board carries the volume potentiometer (50k linear) and the power/mode switch.

Rear panel carries the power transformer (let's call it a 12VA-ish) and the audio-board, that also contains all the ins & outs.

The power amp is an ST Microelectronics TDA7266, of which only half is being used. Guess they must've gotten a REALLY good price on these. Or they just use the same ones for the stereo Model Two, and economies of scale being what they are, it ended up cheaper to just get more of these, than two different chip-amps for the two products.

On the back side are a whole host of SMD passives, and three opamps. A TI NE5532 driving the headphone outputs, and a TL084 and TL072 doing some other chores (mixing, filtering and whatnot).

Capacitor-wise, it's about as horrid as you might expect for mass-market consumer-grade "Made in China" electronics. Main filter cap is a Koshin (which i've seen in a few Harman-Kardon AV-receivers / "home theater" amps), and most of the smaller ones are Yihcon... [shudder]

Back to the troubleshooting - just to make triple-sure, i connected the speaker back onto the power-amp board, and sure enough, plenty of random noise coming through, so odds are pretty good the problem's somewhere around that. On the bright side, the circuitry's not TOO-too complicated, and that in turn reduces the number of potential culprits. So far, barring cold joints, my bet would be either some crapped-out coupling capacitors, or one of the opamps kicked the bucket.

As per the datasheet, the input to the half of the TDA7266 that's being used is on pin 4. Tracing back from that, we see a ceramic cap in series (for DC-blocking / AC-coupling), a resistive divider (2.7K in series with a 10K going to ground, and then another DC-blocking capacitor - this time a 10uF / 100V Koshin. Given its brand, and since it's a through-hole one, out it comes.

A quick power-on test reveals... Nothing but beautiful silence, which means the amp itself should be fine. So either this cap gave up the ghost, or the noise stems from somewhere upstream. Replacing it with a 22uF Rubycon YK seemingly reduced the noise, but there's some still there. Right then, time to dig in a bit deeper.

A bit of deductive reasoning and trace-following resulted in a ballpark-pinout for the 12-pin ribbon-cable between the audio-board and the front panel:

12 - amp mute/stand-by
11 - +8v for the tuner (coming from a 78L08 regulator)
10 - ground
9 - aux in right
8 - aux in left
7 - ground
6 - rec out right
5 - rec out left
4 - [audio from the front board]
3 - [ditto]
2 - post-volume out left
1 - post-volume out right

The last two signals go through the NE5532 headphone amp, then to the switched headphone output jack, and from there the two signals get mixed through a pair of 4.7k resistors, after which the mono signal goes into the TL084.

Inserting a headphone jack into the designated socket did nothing to influence the noise, so the source is somewhere between that, and the amplifier. Now, the suspicions are starting to point towards the TL084 and the surrounding circuitry...

Said TL084, from what i can tell, does indeed do some frequency-filtering things, likely to get the diddy little speaker to sound not-too-awful. There's some really convoluted feedback paths in there, complete with what look like some factory bodges - they might've replaced or changed the values of a few of the capacitors and resistors in the area with a smear of hot-glue over it, since they don't all look quite as neatly placed and soldered as all the others do.

Took a little while, but i managed to "decode" the circuitry around the TL084. All the ceramic capacitors are unmarked, and since they're part of the uncertainty / likely fault, i'm not 100% confident measuring them will yield much of a (useful) result. Most of the component designators match what's on the board.

First we have a mixer / buffer stage (mixing the stereo headphone-out signal down to mono). Following that is a 2nd order (12dB/octave) high-pass filter, and then a really funky EQ / filter stage, mostly rolling down some of the top end. The last stage is a 2nd order low-pass filter, in parallel with a 680 ohm resistor, resulting in a high-shelf attenuation, i believe (potentially a baffle-step compensation).

Ah, what the hell - off come all the ceramics in that area! But of course, not before doodling up a "cap map", to make sure each goes back where it should, assuming it shows a credible value upon measurement. The last one to get removed, C13, came off a liiiittle bit too easily, as if the solder joint at one end was already cracked or something. A clue, perhaps?

Aaaaand nope. Although it's starting to look (or rather, sound like) the amp chip might be at fault, after all. The voltage on the mute / standby pins is a healthy 8v, so it should be well out of both those states (over 3.5v on the mute pin, and over 1.3v to come out of standby, according to the datasheet). So the symptoms are, at least as it stands, some odd noises as you connect the power (with the radio turned on), as well as the volume going down after a handful of seconds.

The signal right before the last attenuator pad ahead of the amp (2.7k series, and 1k going to ground) peaks at a good 2Vpp or so, but the speaker output (even though the amplifier is spec'd to have 26dB of voltage gain, so about 12x) is way down, well under 1Vpp. Although since i can't see anything on the oscilloscope between those two resistors... They're now looking as the even greater suspects.

Gotcha, you little bugger! While the 1k (R34) measures fine, the 2.7k (R35) is up in the stratosphere, showing a really wobbly 6-7Meg on the meter. Right then, time for swapsies... Couldn't find an SMD 2.7k on short notice, but a 4.7k and a 5.6k in parallel should be close enough, giving 2.55k or so.

And sure enough, that did the trick. No more weird noises on power-up, and now max volume is LOUD - 10-15% of the volume knob's travel is already louder than it was when maxed out before. Chalk up one more live one!..