1

The John Hardy Company

1728 Brummel St.
Evanston, IL 60202
USA
Phone: 847-864-8060
Toll Free: 866-379-1450
Fax: 847-864-8076
www.johnhardyco.com
October 1, 2011
The 990 discrete op-amp is the finest op-amp available for audio applications. If you want
superior sound uality, the 990 can provide it. The 990 is used in the most critical audio
applications. !eadin" recordin" studios, remote recordin" companies, disc and #$
masterin" facilities, T% and radio stations use the 990 for superior performance. &everal mic
preamps and other products usin" the 990 are available from the 'ohn (ardy #o.
Features
Circuit design of the 990 is by $eane 'ensen
of 'ensen Transformers. $eane was awarded
).&. patent *+,2,-,+-9 for this desi"n.
.very aspect of the 990/s desi"n and
performance was optimi0ed throu"h
e1tensive computer aided desi"n and
analysis. .ach component of this discrete
op-amp was carefully chosen for its specific
tas2, providin" superior performance
compared to monolithic op-amps and other
discrete op-amps. 3or complete desi"n
theory, circuit details and specifications,
please see the 'ensen en"ineerin" report.
Packaging and production design of the 990
is by 'ohn (ardy of the 'ohn (ardy #o. The
desi"n enables this +1-component circuit
4++ components for the 9905 and +-
components for the 990#6 to be constructed
on a circuit board 17 suare, with final
module dimensions after encapsulation of
1.1287 suare by 0.9007 hi"h. The
dimensions and pinouts conform to the 5:I-
2820 pac2a"e, allowin" direct replacement
in most applications.
Discrete vs. monolithic op-amps. 5n op-amp
typically consists of do0ens of diverse
components includin" transistors, diodes,
resistors, capacitors and occasionally,
inductors. The fundamental difference
between a discrete op-amp and a monolithic
op-amp is the way these diverse components
are brou"ht to"ether to ma2e a wor2in" op-
amp.
A discrete op-amp is made from individual
4discrete6 transistors, diodes, resistors,
capacitors, and occasionally, inductors that
are brou"ht to"ether on a circuit board or
other substrate to create the final circuit.
.ach diverse component is fabricated on a
manufacturin" line that is fully optimi0ed
for that specific part. Therefore, each
component is the very best it can be. !ow-
noise input transistors are fully optimi0ed
for their uniue reuirements, hi"h-power
output transistors are fully optimi0ed for
their uniue and very different
reuirements. :recision resistors come from
manufacturin" lines that are dedicated to
ma2in" precision resistors. #apacitors come
from optimi0ed capacitor lines. Only after
these fully optimi0ed components are
fabricated are they brou"ht to"ether on a
circuit board or substrate.
A monolithic op-amp starts with a sin"le chip
4monolith6 of silicon that is typically 1;197
suare. This chip is the substrate upon
which the do0ens of diverse components are
created. <ote that all components are
created on the same chip, and you simply
cannot have the world/s best input
transistors, and the world/s best output
transistors, and precision resistors and
capacitors on the same tiny chip. There are
unavoidable compromises due to limitations
in the fabrication process. If the process is
optimi0ed for low-noise input transistors
there will li2ely be a compromise in the
hi"h-power output transistors, etc. It is
impossible to incorporate the two inductors
used in the 990 4!1 and !2 on the 990
schematic, pa"e =6 into a monolithic desi"n.
.ach of them is 0.2=87$ 1 0.+007(. (ow
do you wor2 that into the 1;197 suare chip
of silicon of a typical monolithic op-amp>
?ou don/t@
.ven the small si0e of the typical silicon
chip is a limitin" factor. To fit all of the
parts on such a small chip they must be
made much smaller than mi"ht otherwise be
desired. The reduced si0e causes a reduced
ability to dissipate heat. The closer spacin"
of components and circuit traces reduces the
ma1imum volta"e levels that the circuit can
tolerate.
Aonolithic op-amps are marvels of
technolo"y, but when performance is
critical, they cannot match a discrete op-
amp. 5 discrete op-amp costs more and is
lar"er than a monolithic op-amp, but it
offers superior performance in many waysB
Lower noise. The 990 is an e1tremely uiet
op-amp, particularly with low impedance
sources. This can provide as much as ,dC of
improvement in si"nal-to-noise ratios in
summin" amp applications, compared to the
popular 88=+ monolithic op-amp.
The 990 also provides e1tremely low noise
when used in mic preamp applications. The
'ohn (ardy #ompany manufactures the A-1
Aic :reamp, the A-2 Aic :reamp, the
'ensen Twin &ervo
D
990 Aic :reamp and
several mic preamp cards usin" the 990
discrete op-amp. :lease refer to the data
pac2a"es for those products. The application
notes later in this pac2a"e include a
schematic of the mic preamp circuitry of the
A-1 and a discussion of circuit details.
One of the reasons the 990 is so uiet is its
use of the <ational &emiconductor !A-=9+
supermatched transistor pair for the input
pair of transistors 4E1 and E2 on the 990
schematic6. &ince the input pair of
transistors in any op-amp should be as
closely matched in performance as possible,
the !A-=9+ is ideal because both transistors
of the !A-=9+ are fabricated on the same
chip of silicon, thus "reatly reducin"
performance differences that would e1ist
between separate chips of silicon. This is a
uniue situation where the monolithic
process is superior to discrete. The
performance reuirements are identical for
both transistors, so they should be made on
990 Discrete Op-Amp
T(. 'O(< (5F$? #OA:5<?
2
the same monolithic substrate.
Cut <ational "oes much farther by
fabricatin" 100 transistors on the same chip
instead of Gust two. 28 transistor pairs 480
transistors6 from across the chip are
connected in parallel to form E1, and the
other 28 pairs from across the chip are
connected in parallel to form E2. This
causes even sli"ht variations across the chip
to be avera"ed out, typically reducin" errors
by an order of ma"nitude. :lus, havin"
multiple transistors in parallel reduces
noise.
The silicon chip of the !A-=9+ is about
1;197 suare, the same si0e as the entire
chip of a typical monolithic op-amp@ <o
doubt <ational &emiconductor decided to
ma2e the finest possible supermatched
transistor pair, usin" whatever si0e chip was
reuired.
High output power. The 990 provides much
hi"her output power than monolithic op-
amps. This is because the A'.-1,1 and
A'.-1-1 discrete output transistors 4E, and
E96 are much lar"er than the ones found in
monolithic op-amps 4and some other
discrete op-amps6, so they can handle much
more power. They were desi"ned from the
"round up as power transistors. They use a
silicon chip that is as lar"e as the chip in a
typical monolithic op-amp. The chip is
attached to a metal bac2-plate for improved
heat-sin2in" and heat dissipation. .ach
transistor is about as lar"e as an ,-pin $I:
op-amp.
Then the 990 pac2a"e comes into play. The
metal bac2-plates of the A'.-1,1 and A'.-
1-1 transistors are bonded to the aluminum
shell of the 990 usin" a hi"h thermal
conductivity epo1y. This provides
e1ceptional heat-sin2in" of the transistors.
The 990 pac2a"e has about 1+ times the
surface area of a typical ,-pin $I: op-amp,
"reatly increasin" its ability to dissipate
heat. It is easy to see how the 990 can
handle much hi"her power levels than the
typical monolithic op-amp. In fact, the 990
can drive -8H loads to full output level,
while monolithic op-amps are limited to
loads of 900H at best, and more typically
22H. &ome discrete op-amps use much
smaller output transistors than the A'.-1,1
and A'.-1-1. The transistors have smaller
chips and are lac2in" a metal bac2 plate
critical for heat dissipation. They cannot
handle as much power as the A'.-1,1 and
A'.-1-1.
The ability to drive lower-impedance loads
is important for two reasons. 3irst, the 990
can easily drive multiple power amps, or
pots, or other devices, with less concern for
cumulative loadin". &econd, the resistors,
capacitors and other parts that are connected
around the 990 to determine the function of
the circuit 4preamp, .E, etc.6 can be scaled
down to much lower impedances than those
of a monolithic desi"n. This can result in
lower noise. &ome monolithic op-amps are
theoretically capable of very low noise
performance, but they cannot drive low
impedances without increased distortion or
decreased headroom, compromisin"
performance.
Low noise and high output power. Ihen you
have the combination of low noise and hi"h
output power in the same op-amp, you can
often eliminate e1tra op-amp sta"es in
euipment. Join" bac2 to the A-1 mic
preamp as an e1ample, the 990 provides the
e1tremely low noise that is reuired in a mic
preamp, and the hi"h output power that is
reuired in a line driver or main output
sta"e. There is no need to have two sta"es K
one for low noise and one for hi"h output
power. ?our euipment is simpler with
fewer sta"es. The si"nal path is shorter,
resultin" in less si"nal de"radation. $iscrete
op-amps cost more than monolithics, but
when you use fewer of them, the hi"her cost
is less of a factor.
Higher voltage ratings. The components of
the 990 discrete op-amp are capable of
handlin" hi"her volta"es than those in most
monolithic op-amps. This allows the 990 to
operate with L2+% power supplies, while
the typical monolithic op-amp is limited to
L1,% power supplies. It is very common for
monolithic op-amps to be operated at L18%,
sometimes even L12%. In audio terms, this
means that the monolithic op-amps have
reduced headroom. The 990 with L2+%
power supplies is capable of output levels of
"reater than M2+dCu, while most monolithic
op-amps clip at least several dC below that
due to the reduced power supply volta"es.
Precision passive parts. The 990 uses 1N
100ppm metal film resistors and ultra-stable
#OJ;<:O ceramic capacitors with much
better specifications than the ones typically
found in monolithic op-amps. &ee the
special report about ceramic capacitors on
pa"e , for important information about this
superior type of capacitor.
It sounds better! Aost important of all is the
fact that the 990 sounds better than
monolithic op-amps. The 990 does not
suffer from the many compromises of the
monolithic manufacturin" process. &ome
people thin2 that solid-state euipment is
cold and harsh soundin". <ot so with the
990@
Applications. The 990 offers the finest
performance in summin" amps, mic
preamps, phono preamps, tape-head
preamps, 5;$ and $;5 converters,
euali0ers and line drivers. The sensitivity
of measurement euipment can be increased
by the low noise of the 990. 5pplication
notes follow on pa"e +.
odels. The 990 is available in three circuit
versionsB the ori"inal 990, the 9905 and the
990#. The ori"inal 990 has been available
since 19-9 and has established itself as the
finest op-amp available for audio
applications. The 9905 adds three
components to the ori"inal 990 circuit to
provide protection in the rare event the
positive power supply is lost while the op-
amp is drivin" an e1tremely low $#
impedance such as the primary of an output
transformer. )nder those conditions the
ori"inal 990 circuit would consume hi"her
than normal current from the ne"ative
supply, but the O57 modification prevents
the e1cess current flow. The 990# is a
further development of the O57 version,
allowin" the op-amp to operate over a wide
ran"e of power supply volta"es. Other
additional components provide reduced
offset volta"e. &ee the schematic on pa"e =
for details.
Model# Application
990-12V 12V power supplies.
990-15V 15V power supplies.
990-18V 18V power supplies.
990-24V 24V power supplies.
990A-12V 12V power supplies.
990A-15V 15V power supplies.
990A-18V 18V power supplies.
990A-24V 24V power supplies.
990C 12 to 24V power supplies.
!ote" The 990# is the only model that is in
re"ular production. #ontact the (ardy
#ompany if you need one of the other
models.
Package details. The 990 is pac2a"ed in a
blac2-anodi0ed aluminum pottin" shell
filled with a soft silicone material to
encapsulate the components. The soft
silicone "reatly reduces physical stresses
that could e1ist if the circuitry were
encapsulated in a hard epo1y. (ard epo1ies
e1pand and contract at a hi"her rate than
most electronic components as the
temperature rises and falls. The hardness
and hi"her e1pansion rate of the epo1y can
cause stress fractures in some types of
electronic parts. The soft silicone virtually
eliminates the problem.
The metal bac2 plates of the power
transistors are bonded directly to the
aluminum shell usin" an epo1y with hi"h
thermal conductivity, assurin" ma1imum
heat sin2in" of the transistors. The blac2
anodi0ed finish of the shell provides
ma1imum thermal emission. The pac2a"e
measures 1.1287 1 1.1287 1 0.9007
4!1I1(6, not includin" the pin e1tension of
0.2==7. The pac2a"e is fully compatible
with the 5:I-2820 op-amp. :ins are
0.0+07$, "old;nic2el plated.
#eliabilit$. To ensure lon"-term reliability at
temperature e1tremes, 1N tolerance
L100ppm or L80ppm metal film resistors are
used for F1 throu"h F12. The capacitors in
the si"nal path are ultra-stable 4L=0ppm6
monolithic ceramics, #OJ;<:O
formulation. <OT.B :lease see the special
report on ceramic capacitors on pa"e , for
important information on this very special
and superior formulation. 5ll modules
receive a total of +, hours of active burn-in
at 100P# 4212P36.
=
Component upgrades and other
in%ormation. Aany of the components
listed in the 'ensen en"ineerin" report
have been up"raded in the 990s made by
the 'ohn (ardy #ompany to ensure lon"-
term reliability at temperature e1tremesB
F1 throu"h F12 are up"raded from 8N
carbon film resistors to 1N tolerance
L100ppm or L80ppm metal film resistors.
#ertain critical 1N resistors are
individually trimmed to a hi"her de"ree of
accuracy usin" proprietary trimmin"
procedures.
#1 throu"h #= are ultra-stable 4L=0ppm6
#OJ;<:O type ceramic capacitors. &ee
the special report on ceramic capacitors
on pa"e ,. #+ and #8, which are not in
the audio si"nal path, are up"raded from
the ?8% type ceramic to the Q-F
ceramic, offerin" much "reater stability
over a wider temperature ran"e.
#F= 41<91+C diode6 is replaced with a
diode-connected :<+2805 transistor as
su""ested in the 'ensen en"ineerin"
report. This provides better matchin" with
E=, also a :<+2805.
The !A-=9+< is sometimes used in place
of the !A-=9+( for E1 and E2. It
provides identical performance to the !A-
=9+(, and is pac2a"ed in the much easier
to use ,-pin $I: pac2a"e.
Thermal couplin" aids as listed in the
'ensen en"ineerin" report are unnecessary
because components reuirin" thermal
couplin" are in direct contact with each
other. (i"h thermal conductivity epo1y is
used in critical areas to complete the
couplin" process.
The metal bac2 plates of the output
transistors 4E, and E96 are bonded
directly to the wall of the aluminum shell
for ma1imum heat sin2in". The aluminum
shell distributes heat evenly to all points
of the 990 circuit.
F18 and != 4Ooutput isolator76 are not
part of the basic op-amp Otrian"le7 and are
not included in the 990 as manufactured
by the 'ohn (ardy #o. They are available
separately and are recommended in many
applications for best results. &ee the
'ensen en"ineerin" report for details.

&&'C (peci%ications )'d*u + '.,,-./
easurement (pec. 0nits
Open-loop gain, DC to 30! 125 "#
$ain error at 100"# gain 0.4 "#
%oise-&oltage spe'tral "ensit(,
ea') transistor 0.8 nV*+!
"i,,erential pair 1.13 nV*+!
%oise 'urrent spe'tral "ensit( 1 pA*+!
%oise in"e-,
1./ sour'e resistan'e 0.0 "#
12ui&alent input noise &oltage,
20.! 3an"wi"t), s)orte" input 100 nV
Correspon"ing &oltage le&el -133.4 "#u
5a-i6u6 input &oltage at unit( gain 13.8 V
Correspon"ing &oltage le&el 725 "#u
8nput i6pe"an'e, non-in&erting input 910 5/
8nput 3ias 'urrent 72.2 :A
5a-i6u6 output &oltage,
;< = 45/ 13.8 V
Correspon"ing &oltage le&el 725 "#u
5a-i6u6 pea. output 'urrent 200 6A
>otal )ar6oni' "istortion at 20.!,
VO?> = 724"#u
;< = 45/, gain = 40"# 0.00 @
;< = 45/, gain = 20"# 0.005 @
;< = 000/, gain = 40"# 0.015 @
Alew rate, ;< = 150/ 18 V*:A
Alew rate, ;< = 45/ 10 V*:A
<arge-signal 3an"wi"t),
;< = 150/ 145 .!
A6all-signal 3an"wi"t),
at unit( gain B,tC 10 5!
$ain-3an"wi"t) pro"u't,
10.! to 100.! 950 5!
D)ase 6argin at 105! 938 "eg
D)ase 6argin at E25! 900 "eg
;esponse ti6e at unit( gain E20 nA
Auppl( 'urrent wit) no loa" 25 6A
990 Discrete Op-Amp
+
Application Notes
3ollowin" are several circuits for use with the 990 discrete op-amp. Iith
proper attention to detail, you should achieve e1cellent results.
1igure 2" 3raditional mic preamp. 3i"ure 1 shows a traditional transformer-
input mic preamp, adGustable from 11.9 to 90dC of "ain includin" the input
transformer step-up of 8.9dC. The circuit has a bandwidth of 1802(0 4-=dC6.
The 'ensen 'T-19-C mic-input transformer was desi"ned specifically for the
990.
F1, F2 and #1 provide proper termination for the 'T-19-C input transformer.
F=, F+ and F%1 determine the 5# volta"e "ain of the 990.
#= is used for two reasons. 3irst, it 2eeps the input bias current 4thus $#
volta"e6 of the invertin" input of the 990 from reachin" the "ain-adGust pot
4F%16 where it could cause noise durin" adGustment of the pot. 5ll op-amps
have small amounts of bias current flowin" at their inputs. &mall $#
volta"es develop as these currents flow throu"h whatever $# resistance path
is available 4.RI1F6. <oise could occur durin" adGustment of the "ain pot if
more than about 1m% were to develop.
#= also 2eeps the $# "ain of the 990 at unity so that a small difference
between the $# volta"es at the invertin" and non-invertin" inputs of the 990
will not be amplified into a lar"e offset volta"e at the output.
5n optional offset compensation circuit is shown. The diode re"ulator and
filter circuit supplies a current to the invertin" input which compensates for
the uneual $# resistances seen at the inputs. The offset volta"e at each
input is found by multiplyin" the input bias current 4typically 2.2S56 by the
$# resistance seen at that input. 3or the non-invertin" input, the $#
resistance is the input transformer secondary resistance in parallel with F1
49.192H6. 3or the invertin" input F= is the only $# path. &ince the closed
loop $# "ain of the amplifier is unity, the $# offset at the output is eual to
the difference of the offset volta"es at the two inputs. The compensatin"
current reuired into the invertin" input is the offset volta"e divided by F=
4102H6. This compensation will si"nificantly reduce the $# offset at the
output for applications with no output couplin" capacitor.
#2 provides phase-lead compensation with a hi"h-freuency cut-off of
1-82(0. #+ 5#-couples the output of the 990 to remove any $# offset from
the output.
The use of capacitors #= and #+ to control various $# problems is
traditional. 3or a superior approach that eliminates these capacitors and the
sonic problems they can cause, see the application note for the A-1 mic
preamp on pa"e -.
1igure 4" Phono preamp. 3i"ure 2 shows a phono preamp with related
component values and theoretical FI55 response fi"ures. Jain is +1.-dC at
12(0. The circuit provides FI55 response accuracy of L0.1dC. The values
are ta2en from a paper by !ipshit0 T1U which covers FI55 euali0ation
networ2s and their proper desi"n.
#olumn 1 shows the e1act calculated resistor and capacitor values. The
nearest 1N resistor values are in column 2. #olumns = and + show the values
scaled by a factor of 10 to ta2e advanta"e of the 990/s lower noise fi"ure at
lower source impedances.
#= 5#-couples the 990, causin" $# "ain to be unity. #= could be eliminated
if offset compensation were performed. &ee fi"ure 1 for one method. &ee the
A-1 mic preamp application note for superior methods. The ferrite beads at
the input are optional to reduce F3I.
REFERENCE: 1. Lipshitz, S., On RIAA Equalization Networs!, "ournal, Au#io
En$ineerin$ So%iet&, 'ol. (), *+, +,)-, pp. ./01.01.
1igure 5" 3ape-head preamp. 3i"ure = shows a tape-head preamp. #omponent
values for =.-8 and -.8 ips <5C euali0ation and a "ain of 80dC at 12(0 are
listed. Other "ains and euali0ations can be achieved usin" the formulas
provided. Tape head specs and characteristics vary widely, so the values
listed will probably reuire trimmin". The results should be carefully
e1amined.
Tape heads with e1tremely low output levels will reuire additional "ain. 5
2nd op-amp should be considered for that purpose. It should have flat
response. .ach op-amp should be set for eual "ain at hi"h freuencies
4202(06.
This circuit is similar to the phono preamp, e1cept it is tunable. The F2-#2
networ2 is at =002(0 performin" phase-lead compensation rather than FI55
euali0ation. &ee :hono preamp for comments on #= and ferrite beads.
8
1igure 6" 3wo-stage mic preamp. 3i"ure + shows a two-sta"e transformer-coupled
mic preamp, recommended for situations where e1tremely hi"h "ain is reuired.
The first sta"e is the same as the sin"le-sta"e preamp of fi"ure 1 e1cept the
ma1imum "ain is about +0dC. 5 switchable second sta"e with 20dC of "ain "ives
a choice of sin"le-sta"e operation with up to +0dC of "ain 4includin" the
transformer step-up6, or two-sta"e operation with up to 90dC of "ain. The 2nd
sta"e could be chan"ed to adGustable "ain. Ideally each sta"e would have the same
amount of "ain.
Offset volta"e compensation can be performed on the first sta"e as described in
the sin"le-sta"e preamp te1t, or as shown in the A-1 application note. The second
sta"e will have a low offset volta"e because the invertin" and non-invertin" inputs
see identical $# resistances 4102H6. The techniues in the A-1 application note
can be applied here too. &ee the data pac2a"e for the 'ensen Twin &ervo
D
990 Aic
:reamp, a superior two-sta"e mic preamp usin" the 'T-19-C input transformer
and 990# op-amp. It eliminates all couplin" capacitors by usin" $# servo
circuitry and input bias current compensation circuitry.
1igure -" (ockets. Aany types of soc2ets for 0.0+07$ pins are available from
several manufacturers. The 'ohn (ardy #o. uses and stoc2s the soc2et shown in
fi"ure 8, reprinted from the #oncord catalo". The same part is also available from
#ambion, a very similar part 4but with less retention force6 from Aill-Aa1. It can
be soldered in place, or swa"ed 4tool reuired6. (ere are three sourcesB
CONCORD ELECTRONICS CORP. 212-777-6571 800-847-4162
30 Great Jones St. Part #09-9035-2-03
New York, NY 10012
CAMBION 617-491-5400
445 Concord Ave. Part #450-3756-02-03
Cambridge, MA 02238
MILL-MAX 516-922-6000
190 Pine Hollow Road, P.O. Box 300Part #0344-2-19-15-34-27-10-0
Oyster Bay, NY 11771
1igure 7" (umming amp. 3i"ure 9 show a summin" amp with several optional
features. &ome applications reuire si"nals to be combined at unity "ain, others
reuire different "ains. 3or e1ample, the si"nal from channel = is attenuated by a
potentiometer 4typically 10dC of attenuation6 before it enters the summin" circuit.
To restore the 10dC lost throu"h the pot, a lower value is used for FI<, in this case
=.192H 4see formulas6. Iith many channels bein" summed, the output of the
summin" amp could become e1cessive. The final value for FI< is chosen based on
the number of channels, si"nal levels, pot settin"s, etc.
The non-invertin" input may be "rounded directly, or throu"h a resistor. The
value of the resistor should eual the $# source resistance seen by the invertin"
input, which is the parallel resistance of all the input resistors 4assumin" they are
not 5#-coupled6 and the feedbac2 resistor 4FI< and F3C6. Ihen both inputs of the
990 see identical $# source resistances, the output offset volta"e will be the
lowest. This resistor can result in increased noise when compared to a "rounded
input. This problem can be overcome by addin" a capacitor in parallel with this
resistor. The capacitor has infinite impedance at $#, so the $# specs are
unchan"ed. The impedance is much lower above $#, so the noise performance of
the 990 is not si"nificantly compromised. The value of the capacitor is not critical,
with 0.1S3 bein" a "ood startin" point. If the non-invertin" input is "rounded a
compensation current can be provided to the invertin" input as shown in the A-1
application note. This provides the lowest $# offset at the output of the 990, the
lowest noise, and without the potential de"radation caused by the capacitor.
The actual terminatin" point for the non-invertin" input is critical@ In lar"e
consoles with many inputs, much noise can appear on the "round bus. .ven a
heavy "round bus will have a measurable resistance, with volta"e appearin"
across the resistance. These volta"es can be in the form of power supply noise,
return currents 4thus volta"es6 from other modules, etc. 5lthou"h each input of the
summin" amp may be at unity "ain, the overall "ain of the summin" amp is
hi"her. The "reater the number of inputs, the hi"her the overall "ain will be. 3or
e1ample, 2+ inputs with FI< of 102H results in a final parallel resistance of +1-H,
for a volta"e "ain of 2+ 42-.9dC6 4see the formulas6. That is how much the "round
bus noise would be amplified if the non-invertin" input were terminated far from
the si"nal sources bein" summed. The 990 is much uieter than most other op-
amps, but poor layout or "roundin" can defeat this advanta"e@
!on" summin" busses cause increased stray capacitance at the invertin" input,
resultin" in phase-shift of the feedbac2 si"nal. In sufficient uantities, this can
cause oscillation at ultra-hi"h freuencies. #apacitance can be added in the
feedbac2 loop to compensate. 5n isolator 4F!26 can be inserted between the
summin" bus and the invertin" input. It maintains normal audio performance by
providin" less than 1H impedance throu"hout the audio bandwidth, while
isolatin" stray capacitance by providin" =9.2H impedance at ultra-hi"h
freuencies.
9
-
M-1 Mic Preamp with Input Bias Current Compensation and DC Servo Circuitry
3i"ure - shows the complete circuit of the
A:#-1 mic preamp card used in the A-1 and
A-2 mic preamps, state-of-the-art mic
preamps manufactured by the 'ohn (ardy #o.
This circuit eliminates all couplin" capacitors
traditionally used in mic preamp circuits, and
the de"radation in si"nal uality that they can
cause. The main difference between the A-1
and the A-2 is the type of "ain controlB a 2-
section potentiometer in the A-1, a 19-
position rotary switch in the A-2. &ee the A-1
and A-2 data pac2a"e for further details.
5t first "lance capacitors seem li2e ideal
components to use when tryin" to eliminate
the $# volta"es that always mana"e to creep
into audio circuits. #apacitors have essentially
infinite impedance at $#, and 0ero ohms
impedance throu"hout the audio bandwidth if
the value is lar"e enou"h for the application.
(owever, capacitors also have problems. &ee
the special report about ceramic capacitors on
pa"e , for a discussion of one problem.
5nother problem is dielectric absorption. This
is a condition where a small portion of the 5#
volta"e that passes throu"h the capacitor is
temporarily absorbed by the dielectric of the
capacitor, then released a short time later,
causin" a smearin" of the sound. The severity
of the problem depends on the type of
dielectric in the capacitor, and other
construction details.
The problem tends to be unmeasurable with
normal test methods, but can be audible. &ome
film dielectrics such as polypropylene,
polycarbonate, polystyrene and Teflon
minimi0e the problem. Cut when a circuit
reuires several hundred microfarads, it is out
of the uestion to use them, both from a space
and cost standpoint. 5 compromise approach
has been to use electrolytic capacitors of the
reuired lar"e value, then add a 1.0S3 or
0.1S3 4or both6 film capacitor in parallel, the
theory bein" that low freuencies will be
handled by the lar"e electrolytic capacitor, and
hi"h freuencies 4where the smearin" would
be most audible6 will be handled by the small
film capacitors.
Traditional transformer-input mic preamps
typically have two couplin" capacitors in the
si"nal path. Feferrin" to the traditional mic
preamp circuit of fi"ure 1 they are #= and #+.
Their functions are discussed in that
application note.
Aic preamps with transformerless inputs have
two additional couplin" capacitors to 2eep the
M+,% phantom power supply volta"e from
reachin" the active circuitry of the preamp
where it would cause dama"e. 5n input
transformer inherently bloc2s $# volta"es,
but does not suffer from the problem of
dielectric absorption that capacitors have.
Aanufacturers of transformerless mic preamps
mi"ht say that these capacitors cause less
sonic dama"e than an input transformer. This
is true of some input transformers, but not
with the 'ensen 'T-19-C input transformer
used in the A-1@ This is 'ensen/s finest input
transformer, and it is truly superior. &ee the
A-1 data pac2a"e for details.
The A-1 ta2es a different approach. Father
than forcin" the audio si"nal to pass throu"h
various capacitors to bloc2 the $# volta"es
4and, in the process, smear the audio si"nal6,
the $# volta"es are nulled usin" special
circuitry. The couplin" capacitors are
completely eliminated.
The input bias current compensation circuit
4OICI5&76 on the A:#-1 mic preamp card
provides an adGustable current to each input
of the 990 op-amp. The current is the
opposite polarity of the normal input bias
currents of the 990 op-amp. Ihen F%2 is
properly adGusted, the input bias currents of
the 990 are nulled so that no $# volta"es
are developed at the inputs of the 990.
Traditionally a couplin" capacitor 4#= in
3i"ure 16 is used in series with the "ain
controls to 2eep $# volta"es from reachin"
the "ain controls where they could cause
noise durin" adGustment of the control. The
ICI5& circuit eliminates the need for this $#-
bloc2in" capacitor.
&ince all input-related $# volta"es have
been nulled by the ICI5& circuit, it is no
lon"er necessary to worry about a small
difference in the volta"es at the inputs bein"
amplified into a lar"e $# error or offset at
the output of the 990. Therefore it is not
necessary to limit the $# "ain of the 990 to
unity, a function that #= also traditionally
performs. 5"ain, #= can be eliminated by
usin" the ICI5& circuitry.
5 M18% reference volta"e is applied to the
top of F%=, a 28-turn trim pot. The trimmed
volta"e is applied to the inputs of the 990 as
a current throu"h F9, F10, F11 and F12.
#= and #+ act as noise filters.
The $# servo 4O&.F%O76 circuit
continuously monitors the output of the 990
for the presence of any $# offset, and
provides a correction to the invertin" input
of the 990 throu"h F18. The final $# offset
of the 990 is determined by the $# offset
characteristics of the servo op-amp 4)26.
The 5$-08' was chosen because it has
e1ceptional $# characteristics, with a
typical $# offset of 200 microvolts and drift
of 2 microvolts;P#. The $# offset
performance of the 5$-08' is further
improved by an order of ma"nitude throu"h
the use of trim pot F%=. 4#urrently the
O:9-3: is used as the $# servo op-amp.
The !T1012 and !A11#< op-amps were
used in earlier production6.
The servo circuit itself acts as an ultra-low
freuency low-pass filter. The 3-=dC
freuency is so low 4well below 1(06 that
essentially only $# is passed throu"h the
circuit and applied to the invertin" input of
the 990 as a nullin" si"nal. The two F;#
networ2s, F1=;#8 and F1+;#9, alon" with
F18, determine the operatin" freuency.
The capacitors have no detrimental effect on
the audio si"nal because they only affect
freuencies in the pass-band of the filter
4well below 1(06.
Input Bias Current Calibration
Ihenever a 990 op-amp is replaced, it
should be assumed that it has a different
input bias current than the previous 990.
The input bias current adGustment procedure
should be performed as followsB
1. Install the new 990, turn on the power and
allow the unit to warm-up for at least 18
minutes.
2. #onnect a $# voltmeter with at least 100
microvolt sensitivity to the circuit as followsB
The :ositive lead connects to test point *1
4OT:176. This is the output of the 990 op-amp.
5 lon" "old pin is provided for T:1, located
alon" the left ed"e of the p.c. board near the
rear. The ne"ative lead connects to "round. 5
lon" "old pin is provided for this "round
connection to the rear of the 990 op-amp.
=. Aove ':= to the 5$')&T 4O5$'76
position. ':= is located to the ri"ht of the 990
op-amp. This disconnects the $# servo circuit
so you can measure the $# offset of the 990
op-amp.
+. &et the "ain controls to minimum "ain and
ma2e note of the $# offset measured at T:1.
8. &et the "ain controls to ma1imum "ain and
adGust F%2 so that the $# offset readin" is
within 1 millivolt of the readin" ta2en when
the "ain controls were at minimum "ain. It
may ta2e several seconds for this measurement
to settle. F%2 is a 28-turn trim pot labeled ICI5&
located to the ri"ht of the 990 op-amp.
9. Fepeat steps + and 8 until the $# offset
measurements are within 1 millivolt of each
other at minimum and ma1imum "ain. <ote
that both readin"s may be several millivolts, or
even tens of millivolts. They mi"ht both be
positive, or ne"ative. The important thin" is
that they are within 1 millivolt of each other,
and the same polarity.
-. Aove ':= to the OF)<7 position. This
reconnects the $# servo. This should cause
the $# offset volta"e of the 990 op-amp to
drop to well below 1 millivolt. In fact, the $#
offset should drop to well below 100
microvolts if the $# servo circuit is
functionin" properly.
DC Offset Calibration
The final $# offset of the A:#-1 mic preamp
card is determined by the performance of the
$# servo op-amp 4)26. 5n 5$-08' or O:9-
op-amp is used because it has e1cellent $#
specifications. It is capable of providin" a $#
offset that is typically less than 200 microvolts
without any additional trimmin". This is
e1cellent, but can be improved by an order of
ma"nitude throu"h the use of trim pot F%=.
Ihen ma2in" $# measurements below 100
microvolts you will need a $# voltmeter with
a sensitivity and resolution of at least 1
microvolt. Aost meters will have $# offsets
of their own to deal with, as well as drift due
to time and;or temperature. .ven the cables
and test probes can introduce errors. 3ollow
the meter instructions very carefully re"ardin"
warm-up time and 0eroin" procedures.
Ihen you have properly warmed-up and
0eroed your meter, adGust F%= until the $#
offset measurement between T:1 and "round
is as close to V.FO microvolts as possible.
,
Ceramic Capacitors
#eramic capacitors have a bad reputation in audio circles. It is only partially deserved. Aany
en"ineers are unaware that there are several distinctly different "rades of ceramic capacitors, each
havin" a uniue formulation of ceramic dielectric, and a uniue set of properties. The three most
common ..I.5. T1U types areB
1. )ltra-stableB #OJ dielectric 4also called <:O T2U6.
2. &tableB Q-F dielectric.
=. Jeneral purposeB V8) dielectric.
The #OJ dielectric is a vastly superior performer. It is also more e1pensive, particularly in values
above a few hundred p3, and is usually dismissed as cost-prohibitive. 5 common mista2e is to
shop by price alone and buy the cheaper dielectrics, not reali0in" the serious performance
compromises. The en"ineer then condemns all ceramics based on the limited e1perience with only
the inferior types. Too Cad@ .1amination of the performance "raphs of fi"ure 1 reveals si"nificant
differences between the dielectrics. In each case K capacitance vs. temperature, capacitance vs.
time 4a"in"6, capacitance vs. applied 5# volta"e, capacitance vs. $# stress, and dissipation vs.
temperature K the Q-F and V8) dielectrics show si"nificant compromises when compared to the
#OJ formulation.
The Q-F and V8) formulations trade off electrical performance for increased volumetric
efficiency. To achieve this a ferroelectric material is used. 3erroelectric behavior is comple1. 5n
e1cellent te1t by #entre .n"ineerin" T=U provides a comprehensive discussion of this and other
ceramic properties. .ssentially, ferroelectricity causes capacitance to chan"e as the applied
volta"e to the capacitor is chan"ed. In audio applications the 5# volta"e passin" throu"h a
ferroelectric dielectric would modulate the capacitance. In resistor;capacitor networ2s in
euali0ers and crossovers this modulation causes distortion which increases as the si"nal
freuency approaches the cut-off freuency of the F;# networ2.
Tests were conducted with the #OJ, Q-F and ?8% ceramic dielectrics 4?8% is similar to the
V8) formulation6 to measure total harmonic distortion vs. freuency when used as hi"h-pass and
low-pass filters. 3i"ure 2 shows the specific (: and !: filter circuits and the test results. The
Q-F and ?8% formulations show si"nificant amounts of distortion, but the #OJ formulation,
bein" non-ferroelectric, shows distortion fi"ures at or very near the residual of the measurin"
euipment.
5n article by 'un" and Aarsh T+U presented the same test, but with the Q-F dielectric only,
providin" a ne"ative view of ceramic capacitors. The balance of the article is very enli"htenin", as
it enters into relatively une1plored areas of capacitors in audio applications. 5 more recent paper
by 'un" T8U ma2es the distinction between the various ceramic dielectrics and e1presses a
favorable view of the #OJ;<:O dielectric.
The #OJ ceramics were chosen for use in the si"nal path of the 990 op-amp as manufactured by
the 'ohn (ardy #ompany for several reasons. 3irst, their performance is e1ceptional, as noted.
&econd, thou"h the hi"her values are usually considered cost-prohibitive, the low values used in
the 990 492p3, 91p3 and 180p36 are cost-competitive with other dielectric types. Third, they are
the smallest capacitors available, e1tremely important when +- components must be pac2a"ed on
a 1W suare p.c. board.
(opefully this information will provide a better understandin" of ceramic capacitors. .ach
formulation has its proper place, and for audio applications, the #OJ dielectric is superior.
XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
References:
1. 1.8.A. = 1le'troni' 8n"ustries Asso'iation.
2. %DO = %egati&e-Dositi&e-Fero, in"i'ating a te6perature 'oe,,i'ient o, 'apa'itan'e t)at is neit)er plus nor 6inus, 3ut is &er(
'lose to !ero. >e6p'o = 30pp6, -55 to 7125GC.
3. H>1C%8CA< 8%IO;5A>8O%, Cera6i' Capa'itors.J Capa'itor 'atalog, Centre 1ngineering, 2820 1. College A&e., Atate
College, DA 10801
4. Kung, L., 5ars), ;., HDi'.ing Capa'itors - Dart 1J Au"io, 2*80M HDi'.ing Capa'itors - Dart 2J Au"io, 3*80
5. Kung, Lalter $., H>opolog( Consi"erations ,or ;8AA D)ono Drea6pli,iersJ. A.1.A. preprint N1419BD12.