---
Title: Tape recorder – AudioLab
---

# Tape recorder

![A reel-to-reel tape recorder with full reels of tape, its record level meters lit](figures/tape_recorder.jpg)

Wiring (*Stereo*): OUT 1 → tape line input L and IN 1; OUT 2 → tape line input R and IN 3; tape monitor output L → IN 2, R → IN 4. With *Stereo, references first* the references are IN 1 and IN 2 (a Y-cable from each output), the tape outputs IN 3 and IN 4. The machine must monitor from the playback head while recording (Source/Tape or Input/Repro switch). With a test tape, connect only the tape outputs and leave *Play test signal* off.

## What to measure

| What | Task | Set | Good value |
| --- | --- | --- | --- |
| Speed | Speed & flutter | 3150 Hz (3000 Hz NAB); with a test tape *Play test signal* off | within ±0.5 %, well adjusted ±0.2 % |
| Wow & flutter | Speed & flutter | run at least 10 s | compare the weighted figure with the catalogue: DIN 2σ or W.RMS (JIS) as quoted |
| Head azimuth | Speed & flutter | 10 kHz, both channels shown | L/R time difference below about 10 µs |
| Playback response from a test tape | Test tape | RUN, then play the tape from the start | to the tape's standard, typically ±1–2 dB |
| Playback level and treble trimmers | Level meter, **REL** | the tape's reference tone; REL on 1 kHz, then 10 kHz | nominal output; treble to the tape's standard |
| Record level | Level meter, *gain* | 1 kHz at the reference level, monitor from tape | gain 0.00 dB |
| Bias by overbias | Level meter, **REL** on the peak | the tone and drop of the service manual, about 30 dB below the 0 VU level | *rel* at the drop the service manual gives, e.g. −4 dB for a common tape at 19 cm/s |
| Record → playback response (bias, EQ) | Tape response | about 20 dB below the 0 VU level, live while adjusting | 30 Hz – 16 kHz ±2 dB (speed and tape dependent) |
| Distortion and MOL | THD vs level | 333 Hz or 1 kHz, the tape delay is measured and waited for automatically | 3 % THD = MOL |
| S/N | Noise & S/N | tone at the reference level | 55–70 dB (more with Dolby) |
| Crosstalk | Crosstalk | – | −40 to −60 dB |
| Modulation noise, dropouts | Spectrum → **Spectrogram** | a recorded tone, over a minute | a steady, narrow tone; no bursts or holes |

Order of adjustment: playback first – azimuth (**Speed & flutter** at 10 kHz), level and treble (**Level meter** with **REL**, then **Test tape** over the whole series) – then recording: record head azimuth, bias (**Level meter** with **REL**), record level (*gain*), record EQ (**Level meter** with **REL**, then **Tape response**). The sections below show each step on a reel-to-reel deck. See [REL and Max hold](spectrum.html#rel-and-max-hold-turning-a-trimmer).

## Speed & flutter

1. Wire as above (stereo: left on IN 2, right on IN 4). Azimuth is measured when both channel keys are on.
2. SIGNAL → *Test tone*: **3150 Hz** (IEC/DIN wow and flutter), **3000 Hz** (NAB), **10000 Hz** for azimuth. Record and play: tick *Play test signal*, record, monitor from tape. Test tape: untick it; the tone is found automatically, and the speed is read against **Test tape tone**: *Auto* takes the nearest standard test tape tone (315, 400, 1000, 3000, 3150, 6300, 8000, 10 000, 12 500, 14 000 or 16 000 Hz) when it lies within 6 % of one; pick the tape's tone, or type its own, when the tape carries another frequency or the deck runs far off speed – 3000 and 3150 Hz lie only 5 % apart, so a 3150 Hz tone played 3 % slow is nearer 3000 Hz. A tone set is used while the tone heard lies within 10 % of it; another section of the tape falls back to *Auto*. Without a tone to read against, the bar gives no speed – wow and flutter, azimuth and level do not need one.
3. **RUN**. Results come from the last ~10 seconds (AES6 asks for at least 5 s). When a test tape moves on to another tone, only the new tone is read: from 2 s after the change, over a window that grows back to 10 s; read the figures once it is full, and press **STOP** while the tone still plays to keep them.

| Value | Meaning |
| --- | --- |
| speed | deviation from the nominal tone (meaningful with a test tape; record and play on the same machine cancels it) |
| drift | slow variation below 0.5 Hz |
| **wow** | variation 0.5–6 Hz (capstan, reels) |
| **flutter** | variation 6–200 Hz (rollers, bearings) |
| wow & flutter **weighted** | with the IEC 60386 / DIN 45507 curve (most weight at 4 Hz, where speed variations are most audible): **2σ** (the deviation exceeded for only 5 % of the time, the AES6 method) and **W.RMS** |
| wow & flutter unweighted | total 0.5–200 Hz: RMS, 2σ and the largest peak |
| **azimuth** | how many microseconds the right channel lags the left (tilted head); measure at 10 kHz |
| level, dip | level stability and the deepest dropout |

The weighted figures compare with catalogues: a *DIN* or *IEC* figure with the weighted 2σ, a *JIS*, *NAB* or *WRMS* figure with W.RMS (a DIN peak reads about 2–3× the W.RMS of the same machine). They are approximate: the curve is read from the standard's published graph, and the 2σ statistic stands for the quasi-peak meter of DIN 45507. The 2σ value ignores rare single excursions. The chart shows the speed deviation over time, unweighted and weighted.

![Wow and flutter weighting of IEC 60386: 0 dB at 4 Hz, −3 dB at 1 Hz, −11 dB at 0.5 Hz and −22.5 dB at 200 Hz; the bands drift (below 0.5 Hz), wow (0.5–6 Hz) and flutter (6–200 Hz) are shaded](figures/flutter_weighting.png)

The **Spectrum** key above the chart shows the deviation **by frequency** (0.15 Hz – 1 kHz, % RMS): every rotating part varies the speed at its own rotation frequency and stands out as a line – a turntable's platter at 0.56 Hz for 33⅓ rpm (0.75 Hz for 45, labelled), a capstan, a pinch roller, a motor; scrape flutter (the tape rubbing on heads and guides) is a broad floor above 100 Hz. The drift, wow and flutter bands are shaded and named at the top, and the IEC 60386 weighting is dashed on the right axis up to 200 Hz, so you see what the weighted figures count; the scrape band above 200 Hz is shown but not counted (a tone below 3 kHz shows less of it). Details list the strongest lines. Run at least 10 s: the lowest lines need the long window.

The **Azimuth** key above the chart turns the view into a readout for adjusting the playback head (below).

To have the lines of a tape deck named, enter the tape speed and the capstan and pinch roller diameters (a caliper is enough) in **Settings → Analysis → Tape transport**. A roller of diameter d turns at speed / (π·d): a 2.4 mm capstan at 4.76 cm/s at 6.3 Hz; the line within 4 % is labelled *capstan* or *pinch roller*. Reel lines move as the tape winds, so measure at the start and at the end of a reel. Recording and monitoring on the same deck counts the transport twice, with the head gap delay in between: a line can read stronger, weaker or vanish. For the playback transport alone use a test tape.

### Adjusting the azimuth

**Azimuth** (key above the chart, Speed & flutter, both channel keys on) reads how much later the right channel arrives than the left from the **last second**, four times a second – fast enough to follow a turn of the azimuth screw. Wow and flutter do not disturb it: both tracks pass the head together, so a speed change moves both channels alike and cancels in the comparison. What remains is the head's tilt and the tape wandering across the head.

**The goal:** the two channels arriving together – the **mean within ±2 µs of 0** on the tape's 10 kHz tone. At the same setting the treble of both channels is highest.

**What the chart shows:**

| In the chart | Meaning |
| --- | --- |
| **Blue line** | every reading, four a second, over the last 30 s; now is at the right, the latest reading written large beside it |
| **Orange dashed line** | the **mean of the last 10 s** (its value in the legend and the bar): adjust by this, as the blue line wobbles round it while the tape wanders |
| **Green band** | the target, ±2 µs: well adjusted |
| **Thin green line** | exactly 0: both channels at the same moment |
| **Vertical axis** | how many µs the right channel comes later: **+** right later, **−** right earlier |

**The bar** says the same in figures: *azimuth* the latest reading (blue), *mean* the 10 s mean (orange, to end between −2 and +2), the range in those 10 s (how much the tape wanders – the narrower, the steadier the transport), *mono* what the offset costs the sum L + R (towards 0.00 dB when set), and the tone it reads on.

**Step by step:**

1. Clean and demagnetise the heads, put right anything that makes the tape wander (a reel rubbing on its flanges), let the deck warm up.
2. Play the tape's **1 kHz** tone. If the reading is beyond about ±30 µs, bring it near 0 here first: at 1 kHz the reading is unambiguous up to ±500 µs.
3. Play the **10 kHz** tone. Turn the azimuth screw a little (an eighth of a turn or less) and watch the **blue line**: it moves within a second. Going towards 0 – the right way; away – turn back the other way.
4. Go on in small steps and give each a few seconds, until the **orange line settles in the green band**. The blue line keeps wobbling round it. If it crosses 0 into the minus, you went too far: back a little.
5. Lock the screw and check again – locking often turns it a little.
6. Check on 1 kHz: it must read the same as on 10 kHz.

**A jump is not a change:** a tone tells offsets up to half its period, ±500 µs at 1 kHz, ±50 µs at 10 kHz. Beyond, the reading wraps round – from +45 to −45 µs at 10 kHz – and the status line warns in red near the limit. One period further there is a false zero, with the treble lower; that is why step 2 comes first. The levels L and R in Details are highest at the true zero (the classic method adjusts for them).

Then the record head: switch *Play test signal* on at 10 kHz, record and monitor from tape, and adjust the record head's screw the same way. With the generator on, the tone is read near its frequency, so hum on a channel recorded weakly (its bias far off while it is set) does not restart the course.

**The head's tilt in degrees.** Details give the offset also as the phase at the tone (*−2.7° at 9.81 kHz*, an electrical angle). With **Settings → Analysis → Tape transport → Track spacing** – the distance of the L and R track centres, about 4.3 mm half track, 3.5 mm quarter track, 0.9 mm a cassette – they add how far the head is turned: tilt = arctan(speed × time / spacing). At 19.05 cm/s with 4.3 mm, 1 µs is 0.0025°: the +8 µs below is 0.02°, a fraction of what one turn of an azimuth screw moves.

Export saves the course: time and azimuth in µs.

![Azimuth over 30 s on a reel-to-reel deck playing a test tape's 10 kHz tone: mostly +8 to +9 µs, one dip to +3.7 µs about 8 s ago, the mean of the last 10 s dashed at +7.2 µs, the band ±2 µs shaded green; the latest reading +8.3 µs](figures/azimuth_reel.jpg)

The reel-to-reel deck of the examples on its test tape's 10 kHz tone, before adjustment: the right channel about 8 µs late, the orange mean at +7.2 µs far above the green band – a turn of the azimuth screw to do. The readings hold within about ±1 µs despite the deck's wow; the dip to +3.7 µs, with nothing touched, is the tape skewing in its guides for a few seconds – uneven tension, or a reel flange the tape rubs against (this deck had a warped reel that rubbed audibly). Put such a cause right first, then adjust for the mean. An offset of 8 µs costs a fraction of a dB per channel; it shows in mono and in the stereo image, not as a treble loss of each channel.

![Azimuth over 30 s on the same deck after adjusting the playback head: the readings within −1.4 and +2.7 µs, nearly all inside the ±2 µs band, the mean of the last 10 s dashed at +1.1 µs](figures/azimuth_reel_adjusted.jpg)

The same deck after a few small turns of the playback head's azimuth screw, on the test tape's 10 kHz tone: the orange mean at **+1.1 µs**, inside the green band, the readings wandering between about −1.4 and +2.7 µs. That wander is the tape moving in its guides, not the head – no screw takes it away, and 1 µs is under 4° at 10 kHz, a few thousandths of a dB in mono. Judge the setting by the mean, then lock the screw and read again, and check that 1 kHz reads the same.

### Reading the lines

A part that touches the tape and does not slip – a roller, a shaft, the tape pack on a reel – rolls its circumference π·d along the tape once per turn. With v millimetres of tape passing each second it turns v / (π·d) times a second. A flaw that comes round once per turn (out of round, a hard spot, a bump, a worn bearing) jerks the tape at that rate, and the spectrum shows a line there.

```text
f = v / (π · d)      line frequency [Hz] of a part of diameter d [mm]
d = v / (π · f)      diameter [mm] of a part turning at f [Hz]

v = tape speed in mm/s: 4.76 cm/s = 47.6 mm/s, 9.53 cm/s = 95.3 mm/s, 19.05 cm/s = 190.5 mm/s, 38.1 cm/s = 381 mm/s
```

So each line answers one question: **what diameter turns at this rate?** With a diameter set in Tape transport, Details work it out for every line up to 50 Hz – *2.2 Hz (Ø 27.6 mm)* – and mark harmonics of a stronger line – *4.4 Hz (2× 2.2 Hz)*; a line that fits a part set there is named instead. Compare the diameters with the parts of the transport:

1. Read the line's frequency from Details or the chart label.
2. Work out d = v / (π · f), and measure the parts that touch the tape with a caliper; the one of about that size is the candidate.
3. A flaw is rarely a pure sine, so a line often has **harmonics** at 2f, 3f …: they belong to the same part, not to parts half or a third the size.
4. A **reel** line moves as the tape winds (its pack diameter changes), so it is broad and drifts between a measurement at the start and at the end of the tape. A roller or shaft line stays sharp and put.
5. Rubber deforms in the nip and the tape adds its thickness on a thin shaft, so a measured diameter and the line can differ by a few per cent.

To confirm a candidate, mark the part with a felt pen and count its turns in 30 s (turns / 30 = f), or change the tape speed: the line of a rotating part moves in proportion, a disturbance of other origin (for example a speed servo hunting) does not.

### Example: a reel-to-reel deck with a test tape

*A real device, measured as described: what the measurement shows and how to read it.*

A reel-to-reel deck at 19.05 cm/s (v = 190.5 mm/s) playing the 10 kHz azimuth tone of a NAB test tape, *Play test signal* off. The bar reads *speed −0.53 %* (against the tape's 10 000 Hz), weighted W&F ±0.183 % (2σ) and 0.109 % WRMS, azimuth 14.4 µs (Details: +51.7° at 9.95 kHz – the same lag as a phase, 14.4 µs × 9950 Hz × 360°). The wow does not depend on the tone: the 1 kHz tone of the same tape gives the same lines.

![Speed deviation over 8 s: bursts of about ±0.3 % repeating every 0.45 s; the level (orange) stays flat apart from one dropout](figures/flutter_time_reel.jpg)

Over time the deviation comes in bursts every 0.45 s – once per turn of something turning 1 / 0.45 s ≈ 2.2 times a second.

![The same by frequency: a sharp line at 2.2 Hz (0.108 %) with its double at 4.4 Hz, a line at 15.5 Hz and lines at 117, 130 and 142 Hz; the drift, wow, flutter and scrape bands named at the top](figures/flutter_spectrum_reel.jpg)

With a pinch roller diameter set in Tape transport, Details list the lines as:

```text
Lines (% RMS): 2.2 Hz (Ø 27.6 mm) 0.108 · 4.4 Hz (2× 2.2 Hz) 0.027 · 15.5 Hz (Ø 3.9 mm) 0.022 · 142 Hz 0.019 · 129.7 Hz 0.017 · 117.3 Hz 0.014
```

The lines, and what size of part each one fits:

| Line | Shape | d = 190.5 / (π · f) | Reading |
| --- | --- | --- | --- |
| 2.2 Hz, 0.108 % | sharp | 27.6 mm | the largest share of the wow: a part of about 27.6 mm turning at a steady rate |
| 4.4 Hz | sharp | – | twice 2.2 Hz: a harmonic of the same part |
| 15.5 Hz | sharp | 3.9 mm | a thin part, the size of a shaft |
| 117, 130 and 142 Hz | sharp | – | at the top of the flutter band: scrape at the heads and guides, or a motor |

Which parts they are is then checked on the deck itself, as above.

## Recording: bias, record level and record EQ

The figures in this section come from setting up a four-track deck by its service manual; another deck's manual gives its own tones, levels and drops.

The service manual's trimmers are turned with the **Level meter**, recording and monitoring from tape at the same time (the wiring above). The tone must be well below tape saturation: 10 kHz in particular, as the record EQ lifts it. Set the generator about 30 dB below the level that reads 0 VU (on the deck below, 0 VU is −40 dBFS, bias was set at −69 dBFS); if the 10 kHz peak plays back in volts rather than hundreds of millivolts, the tape is saturating and the peak is flattened – go lower. Leave the record level and EQ trimmers mid-way while setting the bias.

**Bias by overbias**, each channel on its own: the bias trimmer fully down, then up in small steps (a second each – the playback head is a fifth of a second behind and the reading averages a second) until the playback level stops rising; at the peak switch **REL** off and on, then turn on until *rel* reads the drop the manual gives for the tape. REL stores every input at once, so press it again on the other channel's peak; the two channels share the oscillator, so check both a second time. **Max hold** shows the same drop, but it also catches the jump when the tone or level is changed – REL on the peak is the surer way.

![Level meter on a reel-to-reel deck setting the left channel's bias: a 10 kHz tone at −65.5 and −65.6 dBFS (2.78 and 2.75 mV) on the references In 1 and In 2, the tape outputs at 110 mV (In 3) and 90.7 mV (In 4); REL on, In 3 reads rel −3.70 dB past the peak, gain +31.87 dB](figures/tape_bias_rel.jpg)

Here REL was pressed on the left channel's peak and the bias turned on: In 3 reads *rel* −3.70 dB, on its way to −4 dB, the drop this deck's service manual gives for this kind of tape at 19 cm/s. The references (In 1, In 2) read rel 0.00 dB – they carry the generator, which does not change.

**Record level:** 1 kHz at the reference level; switch the deck's monitor to its input, REL, then back to tape and turn the record level trimmer to *rel* 0.00 dB – the tape then plays back what goes in. The *gain* against the reference input shows the same in absolute terms.

**Record EQ:** REL on 1 kHz played back from tape, then the generator to the treble tone the manual names (12 kHz on the deck below) and the EQ trimmer to its figure, e.g. 0 to +1 dB.

![Level meter with 12 kHz recorded and played back after the record EQ was set: the references at 2.75 and 2.73 mV, the tape outputs at 205 mV (L) and 213 mV (R), gain +36.99 and +37.37 dB](figures/tape_record_eq.jpg)

Then **Tape response** over the whole band.

## Tape response

Record and monitor from tape at the same time, *Level* about 20 dB below the level that reads 0 VU on the deck (below tape saturation; the generator's dBFS figure depends on the deck's input – on the deck below 0 VU is −40 dBFS, so −59 dBFS), **RUN**. The chirp repeats continuously and the tape delay is removed by itself. The magnitude comes from the power ratio, so speed variations do not disturb it; the absolute phase is not drawn – the tape delay wanders with the speed (0.1 % of 178 ms turns 10 kHz by over 600°), so it says nothing about the deck. The L−R entry of the bar gives the channel balance and the azimuth: the tape delay is found for each channel apart, and what the two miss (several samples when the speed wanders) is read back from the slope of their phase difference over the coherent points; with none, no time is given. For adjusting a head use **Speed & flutter → Azimuth**.

![Tape response of a reel-to-reel deck at 19 cm/s after setting bias, record level and EQ: L and R within about ±1 dB from 50 Hz to 20 kHz around 34 dB of gain, falling below 50 Hz; a spike at 50 Hz and a small one at 150 Hz are hum](figures/tape_response_reel.jpg)

The same deck after bias, record level and record EQ, at −59 dBFS: within ±1 dB from 50 Hz to 20 kHz (−3 dB at 33 Hz L, 24 Hz R). The spikes at 50 and 150 Hz are hum on the playback, not the response; the ripple in the treble is the tape's contact with the head. The L−R time (44.6 µs here) is unreliable where the coherence is low – read the azimuth in Speed & flutter.

## Test tape

The playback response from a test tape (or a test disc with a sequence of tones): connect the playback output to the device-output inputs, pick **Test tape**, press **RUN** and play the tape from the start; **STOP** when it has played. Every steady tone (about a second of the same frequency and level) becomes a step at the frequency actually heard, so the tape's speed error does not matter. A short dropout does not split a tone; a tone the tape plays again (after a spoken announcement) is a step of its own in Details and the export. The chart draws the frequency series alone – the longest stretch of tones played in rising (or falling) frequency, a tone repeated within it as one point, their mean; tones before or after it (the azimuth tone, the playback EQ tones) are listed in Details but not drawn, as they are often recorded at another level. Reference tones, more than 6 dB louder than most steps, are listed apart in dBFS; a loud tone within 3 dB of the series tone played next to it stays in the series (the bass of a full-track tape on a half-track head reads a few dB high, and more than 6 dB above most steps when the deck loses the treble); the frequency series is shown relative to its tone nearest 1 kHz, for L, R and L−R (if the tape was started past its 1 kHz tone, relative to the caught tone nearest it – the bar, the axis and the export say which). Speed, flutter and azimuth from the tape's 3150 Hz and 10 kHz tones: **Speed & flutter**. Tapes with a continuous sweep instead of steps are not caught.

![Playback response of a reel-to-reel deck from a NAB test tape, L and R relative to 1 kHz: rising to +3.3 dB (L) and +6.4 dB (R) at 31.5 Hz, falling to −6.5 to −8 dB at 10–16 kHz](figures/test_tape_reel.jpg)

The same deck as in the Speed & flutter example, playing the frequency series of a full-track NAB test tape (−10 dB, 31.5 Hz – 16 kHz): 13 tones, each one point. The bass rise is largely the tape's: a full-track recording read by a narrower track picks up flux from beside the track at long wavelengths (*fringing*), and this tape does not compensate it, so it is not adjusted out. The treble falling to −6.5 to −8 dB is the deck's own – a worn or dirty head, poor tape contact or the playback equalisation; azimuth (11–14 µs between the channels here) accounts for under 0.1 dB of it.

## Other tape measurements

- **Distortion and maximum output level:** THD vs level record → playback with *Distortion limit* 2 or 3 % – see [THD vs level](distortion.html#thd-vs-level-thd-vs-frequency). The limit is read on THD, not THD+N: on tape, wow and modulation noise keep THD+N at about 9–25 % across the whole range.

![THD vs level of a reel-to-reel deck, 500 Hz recorded and played back from −50 to −22 dBFS: THD+N flat at about −21 dB (wow and modulation noise), THD rising from about −52 to −56 dB to −13 dB as the tape saturates; the bar reads 2 % THD at 2.01 V (L) and 2.05 V (R)](figures/tape_thd_vs_level.jpg)

The THD lines rise from about −52 to −56 dB as the tape saturates; the bar gives 2 % THD at 2.01 V (L) and 2.05 V (R) – the deck's maximum output level. THD+N stays near −21 dB throughout: that is the tape's noise and wow around the tone, not distortion. Set the deck's playback volume so its output stage does not clip before the tape does.

- **S/N as in a service manual:** see [Noise & S/N](noise.html#tape-recorder-sn-as-in-a-service-manual).

![Noise & S/N of a reel-to-reel deck playing a tape recorded with no signal: the tape outputs (L blue, R orange) around −115 dBFS/√Hz with hum at 50, 100 Hz and a line near 25 Hz, the references 40 dB lower; the bar reads S/N A-weighted 66.3 dB (L) and 65.4 dB (R) against the 2 % THD level](figures/tape_sn.jpg)

S/N against the 2 % THD level from THD vs level: A-weighted 66.3 dB (L) and 65.4 dB (R), unweighted 62.2 and 58.3 dB – the hum at 50 Hz counts there.

- **Crosstalk:** hiss hides it below a few hundred Hz – see [Crosstalk](crosstalk.html#filled-points-and-open-circles). Between the tracks of a recorded tape, the **Level meter** is quicker: record a 1 kHz tone on one channel only and read the *tone* of the other.

![Level meter with 1 kHz recorded on the left track only: In 3 (L out) at −20.6 dBFS tone, In 4 (R out) at −71.8 dBFS tone – 51 dB of crosstalk; the RMS of In 4 is higher, as hiss and hum add to it](figures/tape_crosstalk_level.jpg)

Read the *tone* lines, not RMS: the tone of In 4 is 51.2 dB below that of In 3, while its RMS (−68.3 dBFS) is mostly hiss.

## The tape recorder, interactively

How three heads record and play back at once, how the bias lets the tape remember a signal, where the capstan and pinch roller lines in the flutter spectrum come from (f = v / (π·d), as in Settings) and how a tilted head shows as an L/R time difference can be tried in the browser: [Tape recorder simulator](tape-simulator.html). It is an illustration of cause and effect, not targets for a particular recorder.
