This is only a preview of the October 2023 issue of Practical Electronics. You can view 0 of the 72 pages in the full issue. Articles in this series:
Items relevant to "Stewart of Reading":
|
AUDIO
OUT
AUDIO OUT
L
R
By Jake Rothman
Discrete audio op amp – Part 2
L
ast month, I introduced a
Current mirror
Note: resistors and
capacitors have
been renumbered
from Fig.9 and
Fig.12
Special compensation network
R3
330Ω
Output stage bias
C1
220nF
R19
1kΩ
OA+ Noninverting
input
C8
150pF
TR5
BC556
1mA
R1
180Ω
+
R6
2.2kΩ
C2
1µF
10V
+
R19*
1kΩ
R8
220Ω
OAO
R17 C6
47kΩ 1µF +
35V
+
C10
100µF
25V
Output
C15
100pF
Mute
(Pull to
0V or V–)
R25
560Ω
or link
TR14
BD139
V–
–6 to –25V
OAO
High-power option
For low-power operation
R24/R25 are links
for high-power operation
use 560Ω resistors
C10
100µF
+
+
0V
+
R23
100kΩ
R15
43kΩ
R16
5.6kΩ
VAS current sink
Output
–
V–
C8*
150pF
R22
47Ω
0V
Output isolator
R18
22kΩ
R12
12Ω
TR10
BC556
L1**
3.9 to 10µH
V+
OA–
C5
100nF
C11
10pF
Feedback components
OA+
TR11*a
BC556
R14
1.5kΩ
R21
22kΩ
Discrete op amp
C7
4.7µF
4.5mA
TR12
BC546
R20
4.3kΩ
R11
12Ω
Thermal link
TR7
BC546
LED
red
low I
Output stage
L1
70mV 3.9 to 10µH
R13
2.7kΩ
Iq set
PR2
5kΩ
C9
100µF
10V
Current sink voltage reference
TR8
BC546
R7
10kΩ
+
LTP current sink
R10
3.9kΩ
0V
OA–
Inverting
input
C4
10µF
+ 10V
R9
6.8kΩ
C13
680pF
R2
180Ω
R5
470Ω
TR13
BD140
6mA
TR3
BC556
C12 R26
47pF 3.3kΩ
1mA
TR6
BC546
Long-tailed pair
R24
560Ω
or link
TR9
BC546
TR1
TR2
BC546 BC546
R18
22kΩ
46
V+
High-power option +6 to +25V
(symmetrical)
Short-circuit protection
PR1
5kΩ
TR4
BC556
C14 +
C7
4.7µF 10µF
25V
C3
39pF (6-10 gain)
82pF (unity gain)
R4
330Ω
DC
Offset
Input
VAS
R27
330Ω
Input
+
We are so used to thinking of op amps
as little black boxes that we rarely
consider what we could achieve by
tweaking component types or values
month, I want to further explain my
thoughts on this circuit, especially
the many options open to builders
thanks to its non-integrated format.
novel project – my design for a
discrete op amp (DOA) primarily
intended for audio applications. This
R21
R20
22kΩ
4.3kΩ
*RF filtering on input
(gain set x6)
**Isolation from cable capacitance
C9
100µF (non-polar preferred)
R23
100kΩ
For high-power operation
R11/R12 are 2.2Ω
For high-power operation
fit C15 if using BD139/140
Fig.15. The DOA circuit with each stage
defined by blocks. (Note, a few gremlins crept
into last month’s diagram of the overload part
of the circuit in Fig.13. TR11 had its base and
collector flipped; C6 was reverse polarised
and must go to ground not V–. Also R17
should be 47kΩ, not 4.7kΩ.)
Practical Electronics | October | 2023
Discrete op amp
V+
Input
OA+
+
OA–
–
OAO
Output
V–
R21
R20
22kΩ
4.3kΩ
(gain set x6)
+
C9
100µF
Gain = 1 + (R21/R22)
Divide R21, R22 by
10 for lower noise
0V
Fig.17. DOA configuration of non-inverting amp.
n S
erviceable
n U
pgrade possibilities – there are endless tweaking and
experimenting opportunities
n G
ood-quality discrete PNP transistors are available, whereas
Fig.16. PCB for the DOA – we’ll construct it next month.
– this is the great opportunity offered by a discrete version
of the classic op amp.
Figure 13 update
The basic DOA circuit diagram was given in Fig.13 last month.
In fact, there are a number of components to be added to
enable various amplifiers (eg, inverting or non-inverting) to
be built around the op amp on the PCB. The complete circuit
is shown in Fig.15. There are specific inputs and outputs
for the DOA (OA–, OA+ and OAO), and also for complete
amplifiers with added feedback and coupling components.
Each of the DOA’s internal stages are highlighted in Fig.15.
These are the:
n L
ong-tailed pair difference amplifier (LTP)
n L
TP current mirror load
n L
TP current sink
n V
oltage reference for LTP current sink
n V
oltage amplifier stage (VAS)
n V
AS constant current (sink)
n O
utput stage
n O
utput bias stage
n H
igh-power output stage
n S
hort-circuit protection system
A quick note about the short-circuit protection system, which
I didn’t cover last month. This operates by sensing the voltage between the emitter (TR9 and TR10) resistors R11 and
R12. When the voltage across TR11’s base-emitter junction
exceeds 0.6V, it turns on. This pulls the base of TR12 towards
0V turning it on, thereby shutting down the DOA.
Winners and losers
Before going any further I’d like to summarise the pros and
cons of taking the discrete route, since it played a big role
in how I designed the circuit.
Pros
n L
ower noise if special input transistors are used
n H
igher possible power rail voltages, giving more headroom
n L
ower distortion into a low impedance load (<600Ω) – the
result can effectively be a small power amplifier
most chip designs only have low-gain lateral PNPs unless
expensive proprietary processes are used
n C
urrents and voltages can be adjusted and optimised to
suit the application
n L
ow-cost, commodity parts can be used – some special
chips cannot be imported into some locations/regions
n ‘
Discrete’ has fashionable connotations in high-end audio
– potentially higher profits!
n L
arge, high-quality NP0 capacitors can be used for compensation. On the other hand, ICs can only use poor-quality
oxide-dielectric capacitors. In ICs the maximum capacitance
value available is greatly reduced by the (relatively) huge
amount of chip real estate consumed by capacitors
n D
iscrete designs are fun to build and highly educational
for the home constructor
n Y
ou can put your name on it via the silk screen!
Cons
n H
igh parts count
n P
oorer reliability compared to ICs
n H
igh labour cost, unless you build it yourself
n L
arge size, so not a good idea for systems using half a dozen
op amps. The basic PCB is shown in Fig.16.
n L
ower levels of protection, more vulnerable to overheating
and rail-to-output shorts
n H
igher power consumption
n L
ess consistent unit-to-unit performance
n H
igher offset and input bias currents, so generally less
suitable for precision DC instrumentation
Configuration options
One of the nice advantages of a PCB-based DOA design is
that we can include on the same board space for standard op
amp circuit ‘external’ components – for example, input and
feedback resistors. This means on one board we can easily
create some of the most common op amp circuits.
Non-inverting amplifier
This will possibly be the most popular option used by most
audiophiles for pre-amps and headphone amplifiers. It is ideal
as a pre-amp following a volume control. The configuration
of the non-inverting op amp is shown in Fig.17. A finished
PCB is shown in Fig.18.
n L
ower noise when being driven by a low source impedance
Inverting amplifier
– eg, moving coil transducers
n L
ower cost than specialist audio ICs
The inverting amplifier configuration is shown in Fig.19. This
gives lower distortion than the non-inverting amplifier but at
Practical Electronics | October | 2023
47
R21
All resistors same
value, eg 1kΩ
V+
Omit C9, R18
R20
Input–
OA–
OA+
Input+
–
Output
+
OAO
R19
This resistor
in C8 position
V–
Discrete op amp
0V
Fig.20. DOA configuration of differential amplifier.
Fig.18. Finished PCB for low-power DOA. Note unused extra
output transistor holes.
R21
V+
R20
OA–
Input
–
OA+
Omit C7, C9
Link R18 and R19
R20 beomes Rin
Output
+
Gain = –R21/R20
OAO
V–
Discrete op amp
Fig.22. (above) Distortion plot of DOA with a load of 600Ω, ±25V
PSU and 12Vpk-pk output. (This includes noise.)
0V
Fig.19. (left) DOA configuration of inverting amplifier.
the expense of higher noise from the input
resistor R20. The inverting configuration is
useful for mixing and distortion-cancelling
negative output impedance transformer
drivers (see Audio Out, February 2022).
Differential amplifier
The configuration for this is shown in
Fig 20. It’s useful for interfacing with
low-impedance sources, such as moving-coil microphones.
Buffer
To be honest, this is just included for
‘completeness’ and I do not recommend using it! There’s not much point
using a full DOA as a follower, since the
massive open-loop gain of an op amp
is not needed. A simpler circuit using
several transistors could be used, such
as a diamond buffer, which can give
Discrete op amp
V+
Input–
Input+
OA–
+
OA+
–
OAO
Output
V–
0V
Link R21
Omit C9, C11 and R20
Set C3 to 82pF and C13 to 1nF
Fig.21 DOA configuration of buffer amplifier.
48
0.001% total harmonic distortion (THD).
If you really want to make one then the
configuration is shown in Fig.21.
Discrete op amp specifications
Now that we have our DOA design,
what is its performance like? Table
1 below shows a mini data sheet for
the basic DOA using the BC546B and
BC556 transistors. This is a preliminary
specification – more refined data will
emerge after I’ve tested multiple PCBs
with different devices.
Transistor options
Many transistors are available with different Hfe grades, or put another way
– ordinary, good and excellent versions.
Table 1. Discrete op amp ‘data sheet’
Max supply voltage.............................±30V (recommended ±25V)
Output swing total..............................power rail minus 1.4V loss with no load
Total current consumption.................12-18mA quiescent (output Iq set by user)
Max current, clipping into 330Ω........40mA
Absolute max current output.............±100mA limit = 60mA
Input bias current...............................2.7µA
Noise voltage.......................................1.5nV/√Hz with low-noise input transistors
Current noise......................................1.2pA/√Hz with low-noise input transistors
Optimum source impedance..............1kΩ
Offset voltage, un-trimmed.................±150mV, trimmed ±2mV
Using dual transistors (TR1/2/4/5).....<±10mV untrimmed
Max dissipation..................................600mW
Continuous power output...................300mW into 330Ω
Open loop gain....................................>100,000x (no emitter resistors, R1 and R2)
Gain bandwidth product....................13MHz
Slew rate.............................................10V/µS proportional to transistor currents
Input impedance.................................>1MΩ
Output impedance (open loop)..........20Ω
Min load impedance...........................330Ω (This can be reduced by a factor of 10
by adding extra output transistors.)
Power supply rejection ratio PSRR....−80dB negative rail
−74dB positive rail (no decoupling network)
Distortion............................................<0.002% THD at 20Hz-20kHz into 600Ω (see
Fig.22)
Practical Electronics | October | 2023
Toshiba dual transistors
6
6
5
4
D
1
G
Q1
1
2
3
HN1A01F
PNP
5
4
Top view
C
E
B
C
E
BC546
BC556
BC549
BC550
BC212
Q2
B
Japanese pin out
2SA970BL
BC212L
(L suffix for centre collector)
Type name
1
2
3
6
5
4
hFE rank
G (best) = 200 to 400
6
5
4
C
1
Y
1
2
3
Instaling ‘Japanese’ transistors in
PCBs (top view)
B
C
E
C Rotate 180°
B ‘Cross legs’
Use sleeving
E
Type name
HN1C01F
NPN
Fig.24. The output transistors and bias
transistor are all thermally coupled
together on a metal pad on the PCB.
Q2
Q1
1
2
3
tary NPN/PNP pair. High current and
voltage capability are the main requirements, along with reasonable Hfe (>200).
BC546 and BC556 are only rated at
100mA. BC639 and BC640 are rated at
80V 600mA, but their low Hfe of 80-120
means there may be a need to increase
the VAS current. ZTX651 and ZTX751
are similar alternatives, rated at 65V 2A.
hFE rank
Y = 120 to 240
Fig.23. Transistor packs and pin outs, and how to install centre-collector TO92 devices.
For audio work, the higher Hfe values are
preferred since they provide lower distortion. On graded BC series transistors (eg,
BC549) the suffix ‘A’ is for the lowest Hfe
values of 100-200; for ‘B’, 200-350; and
for C, >350-600. If no grade is quoted then
the gain can vary from about 100 to 600.
On Japanese Toshiba leaded small-signal
devices , such as the 2SA970, ‘BL’ denotes
the highest H fe group. Unfortunately
there seems to be a multitude of suffixes
and consulting the data sheet is the only
option. Toshiba, Hitachi and Sanyo generally make the best audio transistors if
you can get them, but the circuit still gives
excellent results with generic European
types. Note that most Japanese transistors
have their collector leads in the middle
rather than the more common centre base,
as shown in Fig.23. Once I’ve built several
PCBs with different devices I’ll have a
better feel for the what the real benefits
of Japanese transistors are.
The component option list below is
in order of quality – in other words,
upgrading the LTP circuit has the most
beneficial effect.
For a small current mirror noise improvement, the BC550 has specified lower
noise at 1-3dB NF, but a lower 45V Vce
(only an issue for high-power rails) The
2SA2362K is better still with 0.5dB NF.
Output stage bias generator
In this circuit for TR8, it’s best to use
the same NPN device as in the output
pair (TR9). This will give better thermal
tracking since their V be values will
be similar. So, if you are using ZTX
output devices, use a ZTX651 for TR8.
Note how the output transistors TR9
and TR10 are all thermally coupled
together with the bias transistor TR8,
as shown in Fig.24.
The remaining transistors (TR8/11/12)
are uncritical (low voltage, low current,
low Hfe). Old timers like the NPN BC108
and BC182 can be used for TR6 and
Dual devices
Dual devices are worth considering
because they automatically come with
well-matched transistors. I was pleased
to discover these dual devices from
Toshiba: HN1C01F-GR NPN for TR1/
TR2, and HN1A01F-GR PNP for TR4/
TR5. They can be mounted either way
round (see Fig.23) which makes soldering a little less error prone. They are
available from Mouser in SOT26-6 6-pin
packs, which are ‘easy’ to solder. Last,
but not least, they are cheap.
VAS
For TR1 and TR2, low noise and matching is critical. BC546B is recommended
for normal use. The BC549C (max rail
±25V) will give lower noise. For best
results use 2SC2240BL, which have a
very low 0.5dB noise factor (NF).
So long as the rail voltages do not
exceed ±18V, a good choice would be the
2SC3068, which has an Hfe of 1600. This
device is almost a possible substitute for
JFETs, having a quarter of the normal
bipolar transistor input bias current.
The VAS PNP transistor TR3 is subjected to the full rail voltage, so it
will need to be rated at a minimum
Vce of 65V. It also needs a high Hfe
to avoid loading the current mirror.
Higher voltage types generally have
lower Early effect distortion, but
with lower Hfe, which then increases the loading on the current mirror.
Ultra-low noise is less important,
since it mainly amplifies the first
stage noise. A BC556 is acceptable,
but the BC556B has higher Hfe. The
Japanese 2SA2362K device is rated
at 120V with an Hfe of 350.
The associated current sink transistor (TR7) needs a high voltage
rating, but its Hfe and noise specifications are not important. A BC546
is fine here.
Current mirror
Output stage
In a standard DOA current mirror, use
BC556 PNP transistors for TR4 and TR5.
For the low-power output DOA,
TR9 and TR10 are a complemen- Fig.25. Installing FETs on the input LTP stage.
Long-tailed pair (LTP)
Practical Electronics | October | 2023
R3
330Ω
R4
330Ω
DC
Offset
PR1
5kΩ
TR4
BC556
C1
220nF
TR5
BC556
1mA
1mA
TR1
BF244A
Non-inverting
input
Top view
Inverting
input
TR2
BF244A
Replace R1, R2
with links
R6
2.2kΩ
TR6
BC546
D/S
G
S/D
R5
470Ω
+
C2
1µF
10V
Use centre pin gate (G) device such as BF244A
Note JFET drain (D) and source (S) are interchangeable
49
TR11. BC178 and BC212 could
be used for PNP TR12.
Audio amplifier with
10kHz input
Amplifier
output
Slight fuzziness (low-level oscillation)
Amplifier
output
Audio amplifier with
10kHz input
Oscillation burst often
at one particular level
Square wave 10kHz input
Settling time
Output waveform
Overshoot
Fig.26. Amplifier on the verge of oscillation.
Input sinewave 20kHz
Output slewing-induced distortion
Large
voltage
swing
FET input
The most important possible
tweak/upgrade to the standard
DOA (Fig.15) is using FETs for
the LTP input transistors (TR1/2)
shown in Fig.25. It’s worthwhile
substituting N-channel JFETs here
to produce high input impedance
(>1MΩ) and low bias currents.
The gate becomes the base
connection and since JFETs are
symmetrical devices, it does not
matter which way the source and
drain pins go into the emitter and
collector holes. The voltage noise
is twice as much for generic JFETs
compared to bipolar transistors,
but the current noise and bias
currents are very low.
It’s more difficult to match
JFETs, since their spreads are
much higher than bipolar transistors. The distortion is also higher,
although using BF244A FETs on
a ±25V supply the difference becomes negligible.
FET inputs are particularly good
for active filters and circuits using
potentiometers without coupling
capacitors, such as amplifiers following high-impedance volume
controls. The 2SK170 is the best
device. Older European low-noise
JFETs, such as the BFW10 are also
good. If you are rich, consider dual
matched JFETs from InterFET.
High power output option
Ramps from capacitor charging and discharging. Often asymmetric
due to different current in postitive and negative direction
Fig.27. Effect of overloading the input stage by
too large a value for C3.
Standard ‘Miller’
compensation
C3
39pF
VAS
C12
82pF
C3*
82pF
2-pole
compensation
Vary resistor from
100kΩ (no filtering effect)
to 2kΩ (maximum filtering);
typical value is 3.3kΩ.
R26
0V
Tune for minimum
distortion at 10kHz
without oscillation.
VAS
*For C3 connection
details see next month
Fig.28. Second-order compensation.
50
A pair of extra output power transistors (TR13 and TR14) can be
added to the existing output stage
on the PCB, which then becomes
the driver stage. This enables more
difficult loads to be driven down
to 50Ω. This will be covered next
month.
Stability tip
Remember that output stages can
oscillate with capacitive loads,
such as screened audio leads and
scope probes. This is normally
compensated for by inserting a
resistor (39Ω to 120Ω) in the output
(R22). This unwanted additional
output impedance can be mitigated by putting an inductor (L1) in
parallel with the resistor.
Another stability tip involves
adding a small ‘phase lead’ capacitor of 4.7 to 47pF (C11). This is
often placed across the feedback
resistor (R21), and compensates
for any stray capacitance to ground
from the inverting input. It is best
optimised by looking for any overshoot
on a high frequency (>10kHz) square
wave. Too much capacitance will cause
low-pass filtering and round-off the
square wave.
Compensation options
In all amplifiers using negative feedback,
a capacitor (Cdom, C3 on the DOA board)
is needed to stabilise the loop by setting
a dominant high-frequency roll-off. The
more feedback, or lower the closed-loop
gain, the bigger the capacitor. The single
compensation capacitor C3 has to be 82pF
for a unity gain (buffer), such as in a Baxandall tone control, but 39pF for a gain of
6-10 and scaled accordingly:
Gain
1 (unity)
6-10
10-20
20-50
C3
82pF
39pF
27pF
15pF
Do remember, these are guidelines only,
you still need to check stability for your
board, application and batch of transistors.
C3 can be optimised for different gains
and transistor spreads by using a 5.5 to
65pF trimmer capacitor instead of C3.
Remember to increase it by at least 30%
beyond the point where oscillation stops,
just in case changes in temperature and
loading cause more compensation to be
needed. Instability often occurs at only
some points in a sinewave, as shown in
Fig.26. Stability compensation includes so
many unknown variables it often becomes
a fiddle fest. This is why you see odd ceramic capacitors soldered under the board
in many commercial amplifiers.
Another thing that has to be watched
is that this capacitor does not get too big
(over compensation), because it may then
overload the input stage. There has to be
enough current available to cleanly charge
or discharge it on rapidly changing signals.
This can be tested by driving the amplifier
to almost full output voltage at 20kHz and
is revealed if the output sinewave suffers
from slew limiting and becomes an asymmetrical ramp wave, as shown in Fig.27.
Second-order compensation
If you want an amplifier with the lowest
overall high-frequency distortion, then to
achieve more overall feedback at high frequencies, C3 is upgraded to a second-order
compensation network. This increases the
slope of the compensation curve, so it can
be moved to a higher frequency before
oscillation occurs, allowing the feedback
factor to be higher at higher frequencies. This
gives lower distortion at high frequencies.
It was used in the Tiny Tim Amplifier (PE,
January 2015). It normally adds another
capacitor in series with Cdom (C3) with
the junction tied to earth via a resistor to
Practical Electronics | October | 2023
V+ 10 to 30V
High-pass
filter
Second output inclusive
compensation path
C12
47pF
C3 (Cdom)
33pF
10kΩ
R26
3.3kΩ
5.6V
BZY88
0V
+
Input
–
VAS
Output
*May need to be
adjusted Metal film
2.7kΩ
C13
680pF
10µF+
10V
Tant
Output
0V
Fig.29. Inclusive Baxandall second-order compensation.
You couldn’t do this on a chip.
provide the extra high-pass slope. This
scheme is shown in Fig.28.
Inclusive Baxandall secondorder compensation
Normally, in amplifiers the compensation
capacitor feedback loop cannot include the
output transistors. This is because large
transistors, such as the TIP31, are rather
slow, often with a transition (unity-gain)
frequency (Ft) of only 3MHz. This phase
lag from the output transistors can cause
oscillation. In DOAs, the smaller output
devices used usually have an Ft at least 20x
higher, allowing the compensation loop to
include the output devices. This enables
crossover distortion to be more effectively
reduced by the overall negative feedback,
minimising the distortion rise often seen
around 10kHz. It also seemed a good idea
to combine this with second-order compensation.
This compensation scheme is shown
in Fig.29 and it can be seen that a signal
from the output is fed into C13, rather than
from the output of the VAS. This signal is
high-pass filtered by C12 and R26 to get
the second-order slope. Since there is now
feedback summed from two points, the VAS
feedback can be reduced by decreasing C3
to 33pF. This reduces the loading on TR1.
The overall effect is to provide inclusive
Baxandall second-order compensation.
Note that for gains other than 10, the
values of C3, C12 and C13 will all have to
be scaled; more capacitance for less gain. I
will cover this in more detail next month.
Remember to only use low distortion dielectric capacitors such as NP0 ceramic,
polypropylene or polystyrene for C3, C12
and C13. If you use low voltage SMT X7R
ceramic types, they cause the distortion
we’re trying to reduce.
Optional off-board input bias
current compensation
For a bipolar transistor version of the DOA,
bias current flows into the base of the LTP
input transistors (DOA input terminals).
This causes a small negative offset which
Practical Electronics | October | 2023
2.7MΩ*
Bias current
2.1µA
Discrete op amp
2.7MΩ*
V+
OA+
+
OA–
–
OAO
V–
Fig.30. Input current bias compensation – take care with noise injection.
is why the electrolytic capacitors (C7, C9
and C10) in Fig.15 are all oriented with the
negative terminal facing into the DOA. If
you have an application that requires cancellation of these input bias currents then
you can add on the circuit shown in Fig.30
(an offboard add-on). It works by injecting
a positive current from a low-noise source.
With PNP transistors the current has to go
the other way.
Off-set adjustment
The voltage input offset trimmer (PR1)
can either null the output (OAO) offset or
balance the LTP collector (TR1/2) currents
for minimum distortion. It can’t optimise
both at the same time. Sonically, even-order
distortion is much more subtle than the
effect of earlier asymmetrical clipping, so
trim for minimum OAO voltage offset. It
should be possible to trim this to within
±2mV.
Quiescent current adjustment
The output stage quiescent current (Iq)
must be set carefully with PR1. If set too
high the output transistors will burn out.
Always set for maximum resistance (fully
clockwise) before turning on. Gradually
increase until 70mV is obtained across each
12Ω emitter resistor (R11 and R12) giving
6mA. This current can be optimised for
lowest distortion and current consumption for various loads. More details will
be provided next month.
Input and output protection
formers and loudspeakers. This protection
is shown in Fig.31.
Normally a DOA is built into a system
and doesn’t need these diodes, but in the
less common situation where you are directly driving an inductive load or are using
phantom power then consider adding them.
Recommended power supply
I view an amplifier as a power supply modulator. If the power supply is poor quality,
then so will be the amplifier’s output. Unlike
power amps, a regulated supply is always
used for op amps, so the voltage ratings
of the transistors are less critical because
the rails are better specified. For the DOA
I recommend my special low-noise audio
power supply described in Audio Out,
April 2022. A suitably scaled-up version
providing ±25V is shown in Fig.32. Note
that to set the regulators to 25V, unusual
resistor values (R7 and R8) are needed. E96
series 1% metal-film resistors are now cheap
and are available from Mouser and Farnell.
If you want, 2.2kΩ 0.5W and 82Ω resistors
can be connected in series which will come
closer to 25V than the E96 resistors.
All amplifiers have a power supply rejection ratio (PSRR), that is how much noise on
the power rail is attenuated at the output.
This parameter is normally specified in dB
at a particular frequency. In this case, it is
around −75dB at 1kHz. With this design,
the PSRR is 6dB worse on the positive rail
because signals from the rail can be directly
injected into the emitter of TR3. This is why
an additional power supply decoupling
capacitor is provided: C14 on the DOA.
Another optional off-board add-on is backto-back diodes across the
DOA inputs to protect TR1
V+
and TR2 from reverse biasing
of their base-emitter juncD1, D2
D3
1N4148 OA+
1N4001
+
tions. This can permanently
damage them, increasing
Inputs
Output
OA–
–
OAO
noise levels. Also, reverse-biD4
ased diodes from output to
1N4001
V–
rails will protect the output
Discrete
V–
op amp
transistors from reverse voltage spikes from inductive
loads such as output trans- Fig.31. Protection of inputs and outputs from voltage spikes.
51
L
F1
1A A/S
N
Fig.32. New values for ±25V power supply using
LM317/337 regulators. Obtaining a 22-0-22V
transformer may be tricky. Cricklewood electronics
do a big 80VA one and RS and CPC do smaller
ones. Using the more common 25-0-25V type could
exceed the 37V input voltage of the regulators and
stress the 35V capacitors.
T1
22-0-22V
7VA (minimum)
S1
E
IEC
filtered
mains
connector
120V
VDR1
275V
0V
22V
120V
22V
0V
0V
0V
Con1
Screen
R5
1Ω
2.5W
+34V
R1
10Ω
C1
100nF
D1
SB350
D3
SB350
R3
10Ω
C5
1nF
D5
UF4004
+34V
Unreg
Transformer
connector
Mains earth
connect to
metalwork
C3
100nF
R2
10Ω
+
C2
100nF
Dirty
Gnd
star
D4
SB350
R4
10Ω
C4
100nF
R13
1kΩ
0.5W
C6
1nF
C9
2200µF
35V +
–34V
–34V
Unreg
C17
10nF
+
C10
2200µF
35V +
Clean
Gnd
+
R7*
2.26kΩ
C13
22µF
35V
C14 +
22µF
35V
R10
120Ω
Adj
–33V
+25V
R9
120Ω
R8*
+
2.26kΩ
D8
UF4002
C12
100nF
C8
2200µF
35V
R6
1Ω
2.5W
D7
UF4002
C11
100nF
C7
2200µF
35V
D2
SB350
IC1
In LM317 Out
Adj
+33V
R11
1Ω
+
D20
1N4001
C15
22µF
35V
C16
22µF
35V
D21
1N4001
Con2 (x3)
Op amp
power
R12
1Ω
In IC2
Out
LM337
Adj
–25V
*R7/R8 are Vishay MSR25 0.6W
D6
UF4004
To work effectively this needs an associated decoupling resistor
(R27). (On the early PCB shown in Fig.16, this was inadvertently
omitted but is present in the ones supplied by PE.)
Instrumentation use
Instrumentation normally demands DC precision with low offset.
DOAs have high unit-to-unit spreads, especially if they use random
unmatched input transistors for TR1 and TR2. If the transistors are
closely matched for Vbe at a low current and reasonably matched
for Hfe, the DC performance will be adequate. Matching of the
current mirror transistors TR4 and TR5 is less critical, but the Vbe
values need to be close. If you don’t have access to a transistor
analyser you can just put 3-pin sockets in the board and pick out
pairs giving the lowest offsets and closest currents. The simplest
way to do this is to solder in turned-pin socket strips, as shown
in Fig.33.
Mike Grindle (our PCB designer) has positioned the transistor
pairs so they are thermally coupled with their flat sides facing
each other to reduce drift. He has also provided pads for the dual
SMT Toshiba devices (shown in Fig.34). Even untrimmed, these
devices will give much better DC performance than separate
unmatched transistors.
JTAG Connector Plugs Directly into PCB!!
No Header!
No Brainer!
Next month
In Part 3 next month we will build the PCB and explore the higher
power DOA option.
Our patented range of Plug-of-Nails™ spring-pin cables plug directly
into a tiny footprint of pads and locating holes in your PCB, eliminating
the need for a mating header. Save Cost & Space on Every PCB!!
Solutions for: PIC . dsPIC . ARM . MSP430 . Atmel . Generic JTAG . Altera
Xilinx . BDM . C2000 . SPY-BI-WIRE . SPI / IIC . Altium Mini-HDMI . & More
Fig.33. Trying different
transistors is easy if you install
sockets. In this case I was
trying different types for TR3.
52
Fig.34. Toshiba dual transistors –
note TR1 marking is ‘C1Y’ (for TR2
it is ‘D1G’). Solder these devices
first. They can go in either way.
www.PlugOfNails.com
Tag-Connector footprints as small as 0.02 sq. inch (0.13 sq cm)
Practical Electronics | October | 2023
|