Handmades

5.7 DC offset, at output
DC offset is the name for one kind of unwanted signal at an amplifier’s output. It is
worth considering in depth, as the book makes the case for direct coupling, and DC
offset is the Imp of this realm. Whether DC offset (voltage and the resulting cur-
rent) is a small quirk, a slight nuisance or the sign of a major breakdown depends on
the size of the DC voltage. In conventional amps with high NFB and internal direct
(‘DC’) coupling, DC output offset voltage is assumed to be reduced to a low level
by the very high feedback factor at DC (‘0’ Hz), often 110dB or more. This is true
to an extent. But transistor mismatches, and other production variables, as well as
many kinds of fault conditions that disrupt NFB, can cause offset voltages that are
non-trivial. With inadequately protected BJT output stages with high global NFB,
small DC offsets can cause transistor fatalities (see section 5.5).
When to dial 999
There are no standards for DC offset (error) voltage at the output of an audio amplifier.
A large offset, above say 1 volt (for >50w/8Ω), constitutes a DC fault. This is a
serious condition – see the next section. Never plug an unprotected speaker into an
amp suspected of having a DC fault. The telltale sign is a loud thump usually fol-
lowed by a buzzing, and no music.
Minor effects
Offsets below 1V at the output, are considered medium to small and are almost
harmless to the drive-unit per se. Dependent on program symmetry, a small offset
may slightly reduce or increase headroom in one polarity. Amplifier clipping on
any symmetrical large signals will be slightly asymmetric, changing the distortion
spectra just before hard clip. If the amplifier’s output is direct coupled, the speaker
cone or diaphragm will also be slightly off-centered, with similar effects.
Irritating Effects
DC offsets small enough to be considered harmless may still cause disturbing clicks
and moderate plops when speakers are connected-up. If the amp is wholly direct
coupled or has a slow acting servo, such clicks may well originate from DC offsets
in preceding sources.

Acceptable levels
The approximate sound level of DC ‘clicks’ and ‘blahhts’ can be ascertained from
rough calculation. For example, if an amplifier drives sensitive large monitor speakers
giving about 104dB(c-wtd) rms SPL for 1 volt rms at the input, then an offset of
0.01v (1/100th) applied by the preceding equipment will cause an SPL that is about
40dB lower, or 64dBSPL. This is as loud as room conversation, hence potentially
disturbing, but not too shocking. But an offset of +0.63v giving 100dBSPL would be
more ugly.
Occurrence
A DC offset across the output terminals does not normally occur in amplifiers with
output transformers − which are usually those with valve output stages. But it might
occur if shorted transformer windings or bad connections are at large.
In amplifiers with a series output capacitor (usually single-rail transistor types, or
OTL valve amps), offset can occur but is limited by electrolytic capacitor leakage
and any bleed resistor.
Servo-fix?
Some amplifiers use servo circuits to continually ‘auto-null’ any output offset volt-
age. Servo circuits have a finite range and rate of adjustment. They are no substitute
for DC protection. But if there is no overt DC fault, they can ensure that offset is
kept below say +/− 5 mV over time and independent of temperature. Because the
integration (ie. averaging) time of a servo’s feedback loop needs to be longer than
the longest repetition period of music signals, namely low bass, any such system
will take time to settle after switching on.
Some practical sonic snags of servo circuits are detailed in section 3.4.1.
5.7.1 DC (Fault) protection (DCP, DCFP)
DC fault protection circuits have long been a feature of professional power amplifi-
ers. The intention is to protect (i) the speaker(s) from damaging DC voltages and
(ii) preferably disable the amplifier channel. For users of domestic amplifiers which
have no DCP and which are connected to speakers costing ‘a pretty penny’, the fact
that all transistors will one day fail and that most will fail short-circuit, putting 50v
or more of DC on the speaker, should be sobering.
The reason why DC above a few volts is damaging to drive units is because some
part of the voice coil is pushed out of the gap. A high swing power amplifier can
push the entire voice coil out. Whatever is out of the gap looses the cooling effect of
the usual proximity (typically within 0.2mm) to a substantial heatsink, the magnet,
as well as cooling by air pumping. This part of the coil’s power handling is there-
fore reduced, and if driven normally, the insulation and glue in this section cooks.
With a large offset over say 20v, the coil is quickly burnt out, and either rubs or
breaks, whereas with smaller offsets, say 1 to 10v (according to the power rating)
the voice-coil’s insulation will be slowly degraded, over weeks or months.

The requirement for protection is simple enough: If more than between +/− 1v to +/−5v
DC appears (standards vary) on the output terminals, the output should be discon-
nected. This shouldn’t happen on any legitimate bass signals, however low and
loud. Yet it must happen with DC, within a few seconds.
In theory this can be accomplished simply enough if the amplifier already employs
an output muting relay. A one or two pole low pass filter is used to detect DC, not
signal. The snag is that the relay won’t usually open under ideal zero-current condi-
tions; there will be an arc as the contact opens. Thus the contacts are degraded as
soon as the relay performs in anger. Any gold-flashing or other precious metal will
then have been lost. For this reason, a relay in the output path has long been anath-
ema in high-end domestic power amplifiers.
Without an output relay, the next simplest way to curtail DC at the output would
seem to be to mute the input. But this presumes it comes from the source, and
assumes a healthy amplifier. It won’t help if one or more OPS transistors have
blown ‘short’, or are very leaky, two likely causes of a DC fault.
The next simplest way is to use a ‘crowbar’ (thyristor clamp) to short the output.
This will save the speaker (hence speaker makers fit them to their wares !) but it can
increase the amount of destruction in a partly zapped or sick amplifier − if it ever
operates.
The third approach is to switch off the supply. It is easiest and safest to do this at a
point where just one supply line need be switched. This is usually the AC mains
supply. The DC protection in this case must then be either passive, or else powered
separately. The apparent snag with this approach is that the DC supply in most
power amplifiers contains enough energy for it to take a few seconds for the supply
to discharge. Fortunately, the average bass driver (the driver most likely to be sub-
jected to DC) has no trouble with a power dissipation and voltage that is tapering
off rapidly. If the supply capacity is very large, a high surge-capacity dumper resis-
tor may be switched in.
One important practical feature is often absent from those amplifiers that do have
DC protection: A test facility. In some of my own designs, users can test that the DC
protection circuit works at close to the lowest detection voltage, by moving a header.
In future computer controlled systems, DC protection testing should be included in
power-up or daily test routines.