---
Title: Phono preamplifier – AudioLab
---

# Phono preamplifier (RIAA)

![A tube phono preamplifier beside a four-input audio interface, on top of a network player](figures/phono_stage.jpg)

**Why:** accuracy of the RIAA curve (the tonal balance of records), hum and noise (tiny signals), overload margin.

**Connect:** the input wants a **small signal** (MM about 5 mV, MC about 0.5 mV): interface output → **attenuator** −40 dB (MM) or −60 dB (MC) → preamplifier input. Take the **reference after the attenuator**, at the preamplifier input, so the attenuator's own value and tolerance drop out. Use 48 kHz or more. Run the laptop **on battery** and keep the input cables **away from the stage's mains transformer**: a charger and the transformer's field add more hum than most stages make ([what it looks like](#what-a-phono-stage-suffers-from-and-how-it-looks)).

| What | Task | Set | Good value |
| --- | --- | --- | --- |
| RIAA accuracy | **Stepped sine** (final figure), Gain & phase (adjusting) | ANALYSIS → *Compare with phono curve* = **RIAA** (or **RIAA + IEC**); 20 Hz – 20 kHz, 31 points; *Level* about **2 mV** at the stage's input for MM, 0.2 mV for MC (the bass comes out 19 dB louder) | bar *RIAA max* within ±0.5 dB, excellent ±0.1 dB |
| Hum and noise | Spectrum, Noise & S/N | input connected to the attenuator; read the A value and the noise density in Details | S/N (A) above 80 dB for MM |
| Overload over the band | THD vs frequency, THD vs level | SIGNAL → *Inverse RIAA* on (set before RUN); level at 1 kHz; watch the note under it | margin above 20 dB over the rated level, also at 10–20 kHz |
| THD+N | THD+N meter | 1 kHz, rated level | 0.005–0.05 % |
| Channel balance | Gain & phase, both channels | with *Compare with phono curve* on | L−R below 0.5 dB |

## How RIAA accuracy is measured

It is a frequency response: the preamplifier's gain over the whole band compared with the ideal curve. One tone gives the gain at one frequency only, which is not enough.

- **Stepped sine** plays **one tone after another** (31 frequencies) and reads the tone itself at the output and the reference. Hum and noise do not bias it, so it gives the **final figure** – the classic generator-and-voltmeter method, automated.
- **Gain & phase** plays a chirp and measures all frequencies at once, with phase, live – good while **adjusting**. In a humming preamplifier, points near 50/100/150 Hz may drop out (low coherence).

1. Interface output → attenuator → preamplifier; reference after the attenuator to IN 1 / IN 3, preamplifier output to IN 2 / IN 4 (*Wiring* Stereo, so the front inputs' bass roll-off cancels – see [Which wiring](wiring.html#which-wiring)).
1. CONNECTIONS → *Attenuator before the device* = **−40 dB (phono MM)** (−60 dB for MC), as the switch on the box. With the outputs calibrated the note under *Level* then gives the level at the preamplifier input, and the keys under it set it: **5 mV** for MM (0.5 mV for MC). Uncalibrated: *Level* about −14 dBFS on an M4. The bass comes out 19 dB louder: keep the output below input clipping.
1. ANALYSIS → *Compare with phono curve* = **RIAA**; **RIAA + IEC 7950 µs** for a stage with the IEC rumble filter (otherwise a false −3 dB at 20 Hz).
1. **Stepped sine**, 20 Hz – 20 kHz, 31 points, **SINGLE**; check with **Gain & phase** (they agree within about 0.1 dB).
1. Read the bar: gain at 1 kHz and **RIAA max** (e.g. *RIAA max −0.30 dB at 20 Hz*); with both channels shown, **L−R** is the channel match.

*Inverse RIAA* on the generator is not needed for accuracy (the gain is taken against the reference); it is for distortion and overload. The attenuator is a source of about 100 Ω, while an MM cartridge has hundreds of ohms and about 0.5 H, so with a cartridge the treble comes out slightly different; the attenuator measures the preamplifier alone, the usual catalogue figure. A test record measures cartridge and preamplifier together (task [Test record](turntable.html)).

## Hum and noise, step by step

S/N is quoted against 5 mV at 1 kHz at the input, A-weighted. Measure the signal once through the attenuator, then the noise without it, and subtract. Laptop **on battery** (a desktop computer through a [USB isolator](phono.html#a-desktop-computer-a-usb-isolator)), cables away from the stage's transformer.

1. **The signal.** Through the attenuator, task **Noise & S/N**, *S/N against a test tone* on, 1 kHz at 5 mV after the attenuator (the **5 mV** key with the attenuator set in CONNECTIONS; −14 dBFS on an M4 with −40 dB), **SINGLE**. Note the tone's level at the output in Details (*S/N re tone … at −25.1 dBFS*).
2. **The stage alone.** STOP, unplug the attenuator, **shorting plugs** into the stage's inputs. *S/N against a test tone* **off**, **SINGLE**. Details give the noise A-weighted in dBFS: **S/N (A) = tone level − noise A**. With the test tone left on, AudioLab says *No test tone at the output* and gives the noise only.
3. **With the turntable**, as it will be used: the turntable into the stage, its ground wire to the stage's ground terminal, stylus raised, motor **off**, then **on**. The same SINGLE. The cartridge's coils (hundreds of ohms, about 0.5 H) add noise in the treble; a rise of hum with the motor running is the motor's field or supply. The reference inputs may stay empty: hide their curves with **REF**; the interface floor in Details then means nothing.

Run **Spectrum** beside each step to see what the number is made of: hum lines, a broad hump from a charger, or hiss. What each looks like is [below](#what-a-phono-stage-suffers-from-and-how-it-looks).

## An attenuator box for MM and MC

![Phono attenuator: each channel from the interface output (M4 RCA 3 and 4) through a 10 kilo-ohm series resistor to a node that feeds both the preamplifier input and the reference input of the interface; a DPDT push switch shunts the node to ground through 100 ohm (MM, −40 dB) or 10 ohm (MC, −60 dB)](wiring/phono_attenuator.png)

- 10 kΩ in series and **100 Ω** to ground: **−40 dB** (MM); **10 Ω**: **−60 dB** (MC). The push switch **SW2** changes both channels at once.
- The **reference** (J8, J9) is taken at the same node as the preamplifier input (J6, J7), **after the attenuator**, so the exact ratio and the loading by the preamplifier do not matter. With *Stereo, references first*: J8 → In 1, J9 → In 2, the preamplifier outputs → In 3, In 4.
- **Switch only with the generator stopped** (STOP): while the switch moves, the shunt is open for a moment and the preamplifier gets the full generator level – 40 to 60 dB more than it expects, enough to overload it and anything after it.
- The preamplifier sees a source of 100 Ω (MM) or 10 Ω (MC). That suits MC; a real MM cartridge has about 0.5–1 kΩ and an inductance, so the **noise** of an MM stage measured this way reads lower than with a cartridge. The response and distortion are not affected.


## What a phono stage suffers from – and how it looks

A phono stage amplifies about 5 mV by 35–40 dB at 1 kHz and by almost 20 dB more in the bass, so anything that reaches its input – hum, a charger's noise, the other channel – comes out as large as the music. The figures below are one inexpensive tube MM stage, measured through the −40 dB attenuator; they show the faults most setups meet first, and how to tell them apart.

### Hum and noise: four causes, four pictures

Spectrum, generator off, noise as density (Settings → Chart). Read the **output** curves (blue, orange) against the **references** (violet, brown): what the references see too is already at the stage's input.

![Spectrum of the phono stage's output with the laptop's charger plugged in: hum at 50 Hz about −50 dBFS/√Hz and 150 Hz about −57, and between 500 Hz and 5 kHz a broad floor near −100 to −108 dBFS/√Hz, 15–30 dB above the floor on battery (below)](figures/phono_hum_charger.png)

**A laptop on its charger.** The switch-mode charger ties the computer, and with it the interface, to the mains in its own way: a ground loop through the cable shields. It shows as a **broad hump of noise** in the midrange – here 15–30 dB above the floor on battery – and stronger hum; in another state of charging the same charger gave a sharp line at 60 Hz and 50 dB more noise below 200 Hz. On battery the hump and the extra hum were gone.

![Spectrum with the attenuator's cables laid against the tube stage's mains transformer, on battery: 50 Hz at −37 and 150 Hz at −40 dBFS/√Hz, a comb of odd harmonics up to 10 kHz, seen faintly on the reference inputs too](figures/phono_hum_transformer.png)

**Cables at the mains transformer.** The transformer's stray field induces hum into the loop the input cables make. Its signs: **150 Hz nearly as strong as 50 Hz** and a **comb of odd harmonics** reaching into the kilohertz, the same in both channels, and visible on the references – it enters before the stage.

Why shielded cables do not help: a cable's shield stops the **electric** field – the capacitive pickup from a mains lead – and drains it to ground. The **magnetic** field at 50 Hz passes through a thin braid as if it were not there; only a thick layer of magnetic material, a mu-metal can, holds it back. The field induces a voltage in the loop the shields of all the cables close between the interface and the stage, the current it drives flows through the input cable's shield, and on an unbalanced input that shield is the signal's return. This is typical of **tube equipment**: a large mains transformer, often run near saturation, open on a chassis with no magnetic screening, radiates far more than the small, screened or toroidal transformers of most solid-state gear. Any unit without good magnetic shielding can do the same. The field falls roughly with the cube of distance, so moving the cables away and keeping them together, which shrinks the loop, is what works – a ground wire does not.

![The same with the cables moved away from the transformer: 50 Hz at about −52 and 150 Hz at −58 dBFS/√Hz, the comb weaker, the midrange floor down near −120 dBFS/√Hz](figures/phono_hum_cables.png)

**Cables in a good place:** run short, close together, on the side away from the transformer. The hum falls some 15 dB; what is left is still induced in the cables, as the next picture shows.

![Spectrum with the stage's inputs shorted: 50 Hz at about −67, 150 Hz at −75 and 250 Hz at −91 dBFS/√Hz, the floor falling from −110 in the bass to −155 at 20 kHz](figures/phono_shorted.png)

**Inputs shorted** (shorting plugs instead of the attenuator): the stage's **own** hum and noise – here from its power supply, the floor its hiss. This is the figure to judge the stage by. S/N (A) re 5 mV at 1 kHz was **63–66 dB** shorted against 45–47 dB through the attenuator: a good MM stage reaches 75–80 dB. Measure it with the Noise & S/N task in two runs – the tone through the attenuator, then the noise with *S/N against a test tone* off – and subtract.

**Against the record's own noise.** A record's groove noise lies about 60–70 dB (A) below its reference level – a new, good pressing at the top, a worn one at the bottom – so a stage of 63–66 dB hisses about as much as the record does. Noises add as powers, not as decibels: a stage as loud as the record raises the total by **3 dB**, one 6 dB quieter by about **1 dB**, one 10 dB quieter by **0.4 dB** – a good stage of 75–80 dB adds next to nothing. During the music both are masked; the difference shows in the quiet: between tracks, in the run-out groove, in soft passages played loud. Two reservations: **hum is not hiss** – a tone at 50 or 150 Hz is heard as a buzz under the music even where its figure equals the record's hiss, and the A-weighting, some 30 dB down at 50 Hz, makes it look smaller than it sounds; and with a **cartridge** in place of the shorting plugs the stage's noise rises, as its coils add noise in the treble.

### A desktop computer: a USB isolator

A desktop computer cannot run on battery: its power lead ties it to the mains earth, and the interface's USB cable carries that earth to the cable shields. Break the loop with a **USB isolator** between the computer and the interface. It must be rated for **USB 2.0 High Speed (480 Mbit/s)**: the common inexpensive isolators (ADuM4160 and alike) carry only Full Speed (12 Mbit/s), too little for an interface such as the M4, which is then not found or drops out. If the isolator gives the interface too little current, power the interface from its own supply on the isolated side. Check the result as above: the hum and the hump should drop as they do on battery.

### RIAA: a rumble filter is not an error

![Stepped sine against RIAA, L and R: flat from 200 Hz to 10 kHz, the bass falling to −3 dB at 20 Hz with the phase rising to +65°, the treble rising by up to 0.8 dB at 20 kHz](figures/phono_riaa_subsonic.png)

![The same steps against RIAA + IEC: within about ±0.5 dB from 20 Hz to 10 kHz, the treble still up to +0.8 dB at 20 kHz](figures/phono_riaa_iec.png)

Against the plain RIAA curve the bass falls 3 dB at 20 Hz and the phase leads: the mark of a high-pass. Against **RIAA + IEC** (its 7950 µs rumble filter) the same steps are flat: the stage has the filter, and it works. Before you call a bass roll-off an error, compare with both – switching *Compare with phono curve* redraws the measured steps at once. The rise of up to 0.8 dB at 20 kHz (more in R) is a real, small deviation of this stage.

### Crosstalk: the shape tells the path

![Crosstalk of the stage, both ways alike: −40 dB from 20 to 40 Hz, falling with the RIAA curve to −57 dB at 1 kHz and −70 dB at 5 kHz, then rising 6 dB per octave to −56 dB at 20 kHz](figures/phono_crosstalk.png)

Up to 5 kHz the crosstalk **follows the RIAA curve**: −40 dB in the bass, −57 dB at 1 kHz – 17 dB apart, as RIAA is 18 dB between 30 Hz and 1 kHz. Such a leak runs from one channel's output back to the other's **input** and is amplified with the RIAA there: a shared power supply (a tube stage's common B+) or ground. Above 5 kHz it rises 6 dB per octave: capacitance between the channels. For a turntable it is ample – a cartridge's own crosstalk is about −25 dB – but the shape shows where to look.

### Distortion: when THD+N is not distortion

![THD vs frequency with Inverse RIAA: THD+N flat at −25 to −28 dB in both channels, THD from −80 to −92 dB above 300 Hz, with peaks at 25, 50 and 125 Hz](figures/phono_thd_vs_frequency.png)

With *Inverse RIAA* the output stays at its 1 kHz level at every frequency. **THD+N** sits flat at −25 to −28 dB: that is the hum and noise of the attenuator setup (its unweighted S/N was 28 dB), not distortion. **THD** – the harmonics alone – is excellent, −80 to −92 dB. Its peaks at 25, 50 and 125 Hz are not the stage either: there a harmonic of the tone falls on a hum line (2 × 25 = 50 Hz, 2 × 125 = 250 Hz). With hum in the setup, read THD, not THD+N, and distrust single points on mains harmonics.

![THD vs level at 1 kHz, harmonics H2–H5 over the output level from 20 mV to 1.05 V: below 300 mV all of them fall as the level rises; above it H2 rises again, about 6 dB per doubling, to −78 dBc at 1.05 V, while H3–H5 stay near −102 to −113 dBc](figures/phono_thd_vs_level.png)

THD vs level, the harmonics one by one (Y → *Harmonics H2–H5*). Below about 300 mV out all of them fall as the level rises: there they are not distortion yet but the noise in their bins, which a stronger tone leaves behind. Above it the stage's own distortion shows – almost only the **second harmonic**, rising about 6 dB per doubling of the level, with H3–H5 25–35 dB below: the signature of a tube stage working well inside its range. At 12 mV in (2.5 × the rated 5 mV, 1.05 V out) THD is 0.012 %, no clipping in sight. That is as far as this setup goes: the generator stops at −6 dBFS (Settings → *Generator output limit*), about 12.5 mV after the −40 dB attenuator, so it shows at least 8 dB of headroom. A good MM stage takes 20 dB and more; to find its limit, feed it through a weaker attenuator, for example 10 kΩ with 330 Ω to ground (−30 dB, about 40 mV).

The RIAA curves, the deviation chart and *Inverse RIAA* in detail: [Frequency response and RIAA](response.html#phono-preamplifier-deviation-from-riaa).
