I have the same problem with audible noise (albeit only on one side!)
Then the other side may have a problem. I would check whether the Teac has a protection relay in the output that could have contact issues. (With the
schematic at hand, yes it does. Two, actually, RY51 for the left and RY52 for the right channel output, with a pair of contacts in parallel. That relay must be in fairly bad shape, but you can always unsolder it, pop the lid, inspect the contacts and see whether some contact cleaner on blotting paper or metal polish would restore them to a sufficient degree. If all fails chances are you can adapt a standard 12V DPDT relay.)
There's more relays involved with CD Direct switching, so toggle that a few times as well. The classic screwdriver handle tap test may expose potential suspects.
You may also be hunting for a dead coupling cap, of which there are a few in the signal path as well. Sky-high ESR may mean little degradation in signal level but high effective source impedance driving up the noise.
Side note, 41 dB of gain sounds about right for a 50 wpc integrated. That your average '90s SIP package 4558 preamp and tone stage isn't the last word in SNR is another problem.
on my highly sensitive speakers.
That would be about how much in dB SPL / 2.83 V / m?
For starters, can you live without tone and balance controls? Then use the CD input and enable CD Direct, which bypasses that section entirely and uses a discrete preamp section instead (Q201-Q213, Q202-Q214). That alone should bring down the noise a fair bit. (Does it have any influence on your one noisy channel? That should help narrow down the list of suspects.)
So this is the CD Direct pre (left channel):
The input stage is being run decently hot at about 450 µA per transistor. Not an ideal match for a 50k volume pot (worst-case output impedance 12k5 + 120 ohm R201) though I've seen worse. I assume this thing would fundamentally benefit from source impedances down to the low hundreds of ohms. A 2SA1268G is rated for a beta of 200-400 at 2 mA, so I would expect to be looking at around -1.5 µA of input bias current. Note how source resistances at DC are mismatched (one leg sees 22k, the other 1.8k||22k ~= 1.66 k), giving a ~30 mV input offset, and DC gain equals AC gain as they skimped on the big capacitor in series with R219 so that's (1+22k/1.8k) ~= 13.2, for an expected DC offset of +0.4 V. So the +0.3 V shown seems realistic.
I noticed that this preamp has super high supply rails and could fundamentally swing fairly high output levels close to 20 Vrms. Meanwhile the power amp is a 50 wpc affair on +/-36 V with a gain of about 12.83, so even if we're being generous it's barely ever going to need 2 Vrms. I would consider giving the preamp like 10 dB more gain and attenuating the output by about 20 dB. So R219/220 = 560 ohm MF, R231/232 = 20k MF, R233/234 = 2.0-2.2k MF. I would actually prefer an R231/233 (R232/234) divider of more like 9.1k/1k to keep output impedance low, but then C217/218 should be paralleled or replaced by something like a 10µ/50-63 V (+ towards R231 or R232 respectively). You can also try bypassing the caps and seeing whether you can live with any pop noises this may cause when switching between normal and CD Direct modes.
If you want to convince the input stage to take a chill pill, R285/286 (68k, maybe 100k) and C211/212 (maybe 5p NP0?) would be the ones to address.
Stock pre+power gain looks like about 44.5 dB, yeah that's starting to get a bit on the hot side.
BTW, I think that the block diagram contains a mistake: The tone amp seems to be a line-level affair that ought to be shown connected after the discrete preamp, not before it. (This position is generally good for so-so noise levels, although the power amp's higher than average input levels would be helping here.)
If you want to mod the tone amp as well, I'd think replacing R301/302 by 10k MF and R305/306 by 11k (alternatively 10k/12k) MF would be a good start, as their stock values are degrading noise levels quite needlessly. I would also be looking out for a somewhat better opamp though the SIP case means pretty slim pickings. Maybe you can get an NJM4565L or NJM/NE5532L.
