Five decibels hot at 10 kHz

The calibration article has carried the reproduce procedure since August. Today the deck actually went through it — 3-5-(1) to 3-5-(6), factory order, on the 7.5 ips MRL tape whose tone program was pinned down earlier the same day. Not the machine's first alignment — it is fifty years old and will have been set up in service more than once — but its first since the recap, and that turns out to matter more than it sounds.

The switch in the wrong place

The setup step, 3-5-(1), asks for the SRL switch in its "off" variable position. Mine was on SRL for the whole of the first pass, including the azimuth adjustment. I noticed after the fact and had to work out whether the morning was wasted.

It wasn't, and the reasoning is worth keeping. Azimuth is a maximization: you turn the screw until output peaks. R65 and R165, the pots the SRL position brings into circuit, sit downstream in the output stage. A fixed gain after the head changes the height of the peak but not its position, so the screw ends up in the same place either way. The only real risk was clipping flattening the top of the curve, and a clean peak with both channels tracking together said that hadn't happened.

The theory got confirmed with a number later in the session. More on that below.

Azimuth

3-5-(3) on the ½-track playback head — the rightmost of this deck's four heads, which run erase, reproduce 4T, record, reproduce 2T along the tape path. The MRL tape carries its azimuth group at −10 dB: 500 Hz to set level, then 8 kHz and 16 kHz. Coarse against 8 kHz first, fine on 16 kHz last, which is the order that keeps you out of the false peaks that 16 kHz has and 8 kHz doesn't.

Both channels peaked at the same screw position. That is the result you want — it says the head gap is straight and the stack isn't worn unevenly. Had they peaked in different places there would have been nothing to do but split the difference and note it.

The imbalance that wasn't

Partway through, the two meters disagreed at 1 kHz and agreed at 500 Hz. It looked like a frequency-dependent channel fault. It was two things, neither of them a fault.

The first is that the 1 kHz reference sits at 0 dB on this tape and the 500 Hz azimuth tone at −10 dB. Near 0 VU the needle is in the accurate part of its scale and a decibel or two of difference is obvious; ten decibels down it is in the compressed end, where the same difference is invisible. The imbalance was present at both tones and only legible at one.

The second is that the SRL switch was still engaged, so both channels were being read through R65 and R165 — pots left wherever they were last set, which was before this board had fifteen new capacitors in it. Any mismatch between them shows up as exactly this kind of flat channel difference. Setting them at step 6 brought the channels together.

Reproduce equalization, and the number that matters

3-5-(4), R60 for CH-1 and R160 for CH-2, on PA-286A. The manual prints "R161" for the second one; it is a misprint, and the digital edition marks it.

The manual's procedure assumes an alignment tape whose tones sit at operating level, and asks you to trim until the meter reads 0 dBm. This tape's response run is at −10 dB, so the whole measurement runs relative instead: bring the run's own 1 kHz to a convenient point with the OUTPUT controls — 1.23 V, the +4 dBm this deck puts out at 0 VU with LINE OUT at HIGH — freeze the knobs, and compare everything against it.

10 kHz came in at 2.23 V. That is +5.2 dB.

Enough to pin the needle. Not enough to clip anything, but enough that every tape played since the recap came off this deck with its top end lifted by five decibels. Trimmed down to 1.23 V on both channels.

Tone	Before	After
10 kHz	2.23 V (+5.2 dB)	1.23 V
16 kHz	—	1.245 V (+0.1 dB)

The 16 kHz reading is the one that closes two questions at once. The manual asks for a recheck at 15 kHz after any adjustment; this tape's nearest tone is 16 kHz. Coming in a tenth of a decibel from reference means the whole high-frequency shelf moved together, which is what an intact equalizer network does — had the network itself drifted, correcting 10 kHz would have thrown 16 kHz off instead.

Which points at the likely cause. The replay half of PA-286A — the board these very pots sit on — took fifteen new capacitors on 17 July, six weeks before this session. A reproduce equalizer's response is set by those capacitors, and R60 and R160 were still trimmed for the ones that came out. Five decibels is a lot, but it is the kind of lot a recap produces: the network is healthy and the pot is pointing at the wrong place on it. That reading is what the 16 kHz check supports, and it is a better explanation than fifty years of drift, because the deck was not measured between the recap and today.

It also settles the azimuth worry from the morning. 16 kHz is the shortest wavelength on the tape and by far the most azimuth-sensitive tone, and misalignment shows there first and worst, as loss. A reading essentially at reference means the gap really is square to the tape, whatever position the SRL switch happened to be in when it was set.

One small thing worth recording: the 1 kHz reference read 1.23 V early in the pass and 1.20 V later, a drop of 0.2 dB. Part of that is real — moving a high-frequency shelf by 5 dB nudges a tone sitting an octave and a half below the NAB turnover — and part is ordinary variation across a reel. It is below the resolution of the tape's own calibration graph, whose scale is one millimetre per 0.2 dB, and it is not worth chasing. A tenth of that much again would have been.

Both channels finished inside the ±2 dB window.

The speed that cannot be checked

The procedure's last item asks for the response to be confirmed at the other tape speed. On this deck that cannot be done, for two independent reasons, and it is worth having both on the record.

There is no 3¾ ips alignment tape here. Playing the 7.5 ips reel at LO halves every reproduced frequency, and while the wavelengths line up with what a 3¾ tape would present, the recorded equalization does not — this reel is cut to NAB 7.5, a genuine 3¾ tape would be cut to NAB 3¾. No ±2 dB verdict survives measuring one curve through the other.

And the LO position is already known to be wrong. NAB is 50 µs at both 15 and 7.5 ips and 90 µs at 3¾. These electronics were built for 7.5 at the LO position, so they apply 50 µs where the 3¾ this deck now runs wants 90. That is a wrong time constant, not a misadjustment, and no trim pot reaches it — the consequence of the pulley swap the announcer gave away in August.

Marked not verifiable rather than passed or failed.

Maximum reproduce level, and SRL

3-5-(5), R63 and R163, with both OUTPUT controls hard clockwise — the step sets the chain's maximum gain, so the knobs have to be wide open while it happens. The manual reaches for its tape's first tone, noting that it is recorded 10 dB below operating level; this reel opens at operating level instead, so a −10 dB tone stands in.

Then 3-5-(6), R65 and R165, SRL switch finally in the position it spent the morning wrongly occupying, against the closing 1 kHz at 0 dB. The manual says the operating-level tone is the last tone on a 7½ ips tape, and this reel closes with exactly that.

No offset arithmetic was needed. The tape is 180 nWb/m at 1 kHz, which is the manual's 185 nWb/m-at-700 Hz under its 1 kHz name — the check made against the reel's own graph back in August. The reference tone goes straight to 0 VU.

Whatever those two pots were set to, it was set against the old capacitors. Now 0 VU means 185 nWb/m again.

Both steps went cleanly and neither produced a figure worth arguing with, which is the honest reason there are no numbers for them here. Everything above was measured; these two were set. Next time the before-values are worth writing down even when nothing surprising happens, because "clean" is not a measurement and a later session cannot check it.

Both reproduce heads

This deck carries four heads — erase, reproduce 4T, record, reproduce 2T — and 3-5-(3) was run twice, once against each of the two playback heads.

The manual covers aligning either of them. Its setup step says to put the reproduce selector to 2T position for models 2S or 2SD, to 4T position for models 4S or 4SD, so a factory 4S owner runs exactly this procedure with the selector in the other position. Section 4 puts it generically too: "Follow the procedure for Reproduce Amplifier Alignment given in Section 3 of this manual." And where erase head azimuth is dismissed as "not critical," the contrast tells you the reproduce heads' azimuth is not.

What the book never says is do both, because that selector line assigns 2T or 4T by model — one machine, one reproduce head. This one has two, so the instruction to align the second simply has no occasion to exist. Running it is not an extrapolation; it is the same procedure with the switch in its other position.

The ¼-track pass: selector to 4T, that head's own azimuth screw — second from the left along the tape path — and the same three tones in the same order, 500 Hz for level, 8 kHz coarse, 16 kHz fine, with the same one-turn ceiling. Nothing downstream needed touching. The changeover switch feeds whichever head is selected into the same reproduce amplifier, so R60, R63 and R65 and their CH-2 partners already served both.

One caveat worth keeping: the ¼-track alignment is unconfirmed. It wants a tape known to be quarter-track to check against, and the first tape tried since turned out to be half-track mono — a separate piece of work.

What is not done

The reproduce side is complete. Everything else on the page is still open.

• Record and bias, 3-6, all five steps. It needs blank bulk-erased tape and, before anything else, a decision about which stock this machine gets standardized on — bias is set for the tape you intend to use, and setting it for one and running another is worse than not setting it.
• The speed measurement, still outstanding since August despite the 3150 Hz flutter reel sitting on the shelf. The manual runs that tape spliced behind a full reel of blank stock, because tension affects speed.
• Wow and flutter, which needs the same reel and cannot be estimated from music.

The Results table on the calibration article still has empty rows. The reproduce frequency response column can be filled in now.