TL494 magic chip

From time to time, in newsgroups, in www forums and in www pages, people speak about
converting DC to DC, or DC to AC, charge-pumping, building inverters (step up, step down
converters) for various devices. From high power car audio amp to handheld stazer. There are
lots of schematics (especially in Russian web) using simple TTL logic, transistor and other
devices. But why invent bicycle? There is special, very cheap and very easy to find, chip. It is
TL494 (or any other analog). You can found it in any PC power supply. According data-sheet,
TL494 is PULSE-WIDTH-MODULATION CONTROL CIRCUIT.
The TL494 incorporates all the functions required in the construction of a pulse-widthmodulation (PWM) control circuit on a single chip. Designed primarily for power-supply
control, this device offers the flexibility to tailor the power-supply control circuitry to a
specific application.
The TL494 contains two error amplifiers, an on-chip adjustable oscillator, a dead-time
control (DTC) comparator, a pulse-steering control flip-flop, a 5-V, 5%-precision regulator,
and output-control circuits.
The error amplifiers exhibit a common-mode voltage range from 0.3 V to VCC 2 V.
The dead-time control comparator has a fixed offset that provides approximately 5% dead
time. The on-chip oscillator can be bypassed by terminating RT to the reference output and
providing

sawtooth

input

to

CT,

or

it

can

drive

the

common

circuits in synchronous multiple-rail power supplies.

The uncommitted output transistors provide either common-emitter or emitter-follower
output capability. The TL494 provides for push-pull or single-ended output operation, which
can be selected through the output-control function. The architecture of this device prohibits
the possibility of either output being pulsed twice during push-pull operation.
The TL494C is characterized for operation from 0C to 70C. The TL494I is characterized
for operation from 40C to 85C.
Before doing any useful device, I deeply recommend to examine how this chip works. This will
be some sort of TL494 tutorial

Take your breadboard, prototyping board or anything you like and connect wires according this
schematics:

If everything is connected right and detail are similar to used in the picture, the schematics will
be working. Leave 3 and 4 open. Use your oscilloscope to test is oscillator is working- on CT
(pin 6) you must see saw. The output will be null. Now connect FEEDBACK (pin 3) and DTC
(pin 4) to ground (GND). You must find square pulses on the output. In the next message Ill
explain the pins in more details.

TL494 in the breadboard

TL494 magic chip, Part 2

Now lets speak about the pins.
GND (pin 7) and Vcc (pin 12) are usual power pins. Max Vcc is about 40V.
RT (pin 6) and CT (pin 5) are used for setting clock (1 300kHz). The frequency of the clock
is calculated using table from data sheet and changing resistor (1.8 500 k) and capacitor
(0.47 10000nF). Or the on-chip oscillator can be bypassed by terminating RT to the
reference output and providing a sawtooth input to CT, or it can drive the common circuits in
synchronous multiple-rail power supplies.
REF (pin 14) is 5V reference output. Low power output- 10mA.
Pin FEEDBACK (pin 3) is used to control the pulse width. Is the input is zero, the width is
maximum, if it is ~5V or more (compared to REF) than output is disabled. Input must be less
than Vcc.

Output vs Feedback

DTC (pin 4) is Dead Time Control. This pin is similar to feedback, but it is used to make soft
start. Regardless to FEEDBACK it controls the width of the pulse. If it is more than REF
(~5V) the output is disabled, if zero- output is maximum and controlled by FEEDBACK. Here
is example:

Simple soft start

There is small difference in DTC and FEEDBACK. It can be understood by reading data
sheet.
C1 (pin 8 ) is collector of output transistor #1. E1 (pin 9) is emitter of output transistor #1.
C2 (pin 11) and E2 (pin 10) are like C1 and E1, but for the transistor #2.
The output is max 200mA on each transistor.
OUTPUT CTRL (pin 13) is setting the type of output. If it is high (=REF), the output is two
phase. If input is GND, the output transistors are working in same phase, so you can can
connect them in parallel and thus, get more output current.
The last pins are used in error control unit. They are used to detect various errors and make
protections from short circuit and under-, over-voltage protection. Ill describe them in the
other post.

TL494 magic chip, Part 3

the output of TL494 is quite weak, and we want more power. So we need some powerful
transistors attached. The most easy to use (and very easy to get- from old computer
motherboards) are n-channel power mosfets. We must invert output of the TL494 because if
we connect n-channel mosfet to output, in shut-down stage mosfet will be open and lots of
current will pass threw then in DC. This may burn everything here Also all PWM (pulse width
modulation) will be inversed! So get generic npn transistor and connect according this
schematics:

Output stage with power mosfet and passive drive.

The powerful mosfet is driven in passive on mode. It is not very good, but for testing purposes
and low power are useful. R1 in the schematics are npn transistor load. Select it according max
current the transistor can handle. R2 is the load of our power stage. In next experiments it will
be replaced with transformer coil.

Now, lets look at real oscillograms

Sawtooth signal on the CT pin.

Soft start animation. Oscilloscope is connected to C1, TL494 output pin.

Soft start animation. Oscilloscope is connected to the drain of power mosfet.

TL494 magic chip, Part 4

Now lets make some high voltage! As the schematics is done on the breadboard, I cant
produce high power application. This is because of very long wires used in the experiment. Also
there is no any heat sink on the power mosfets. Connect transformer to the output stage
according this schematics:

Push-pull load
The primary windings of the transformer are identical windings with about 10 turns. The
secondary winding is about 100. So the transforming ration is 1:10. If you put 10V to primary,
you must get about 100V in the output. The core is made from ferrite. I used some middle sized
core from PC ATX power supply.
Be careful, the output on the transformer in under high voltage. The current is very low and
will not kill you. But youll get nice zapp. One more danger- if you place big capacitor on the
output, it can store fatal energy value.
Nixies are working from DC voltage. This nixie needed about 160V to start. (The power supply
of the device is about 15V- small wall cube)
(Other parts used in schematics: PHB55N03LT, 500 and 1K resistors, IN539 diode, unknown
value HV capacitor, -14 nixie, C1685 transistors from old Panasonic TV, 1000mkf
capacitor)

Nixie on the output.

TL494 magic chip, Part 5

Everything is working with simple passive mosfet drive. But if we increase switching frequency
or place powerful mosfet (or even several mosfets)? There is no load in gate circuit, but there is
gate capacitance. And if we need fast switching we must charge and discharge gate very
quickly. With passive drive there are limitations. We can use active charge-discharge circuit.
Also this circuit uses inverter transistor for proper signaling:

Totem pole mosfet drive (push-pull)

Few comments about schematics. R5 is TL494 output load- anything about 1K. R1 is Q1 load,
not critical- 1K. Q1 inverts signal from controller. R2 is used to prevent overload of totem pole
bases. Q2 and Q3 are complementary pair. R3 are useful if are few mosfets on the output, only
few ohms. Q1 and Q3- C945, Q3-A733. All transistors from old PC power supply. R4 load
simulation.
(C945- npn transistor, ~50V, 100mA, 200mW; A733- pnp, ~50V, 100mA, 200mW)
Connect pair of these totem poles and connect to transformer like in older post. You can notice
small increase of the performance. Dont overload. There are lots of parasitic oscillations in this
circuit.

TL494 magic chip, Part 6

While testing this device, do not overload, do not load asymmetrically. For checking your
system, remove nixie and load secondary winding of transformer with simple resistor.

Download small (~500kb XVid AVI): with oscillograms. And now take a note, that on the
drain oscillogram you see other phase pulses. Even when mosfet if in off stage, the induction
from other primary winding cause voltage to appear. This may kill mosfet. And if load is
unbalanced or not ideal, you can find lots of parasitic oscillations. So design engineer must take
care of them. Parasitic oscillations can be killed with snublers- simple RC network on
transformer windings. This devices takes some of the high frequency power and dissipates it.
And now lets look in internet:

Half-bridge and
full-bridge

supply voltage plus a

safety margin
minimum
mosfet voltage
rating

230VAC*1,414 + 50V
=> 400V mosfets

medium
voltage 400V fets, so
other mosfet
properties

a <0.2 Ohm channel

resistance is typical
=>
high current, low loss

Push-pull

Forward converter

double

double

the supply voltage plus a

large safety margin

the supply voltage plus a

large safety margin

2*230VAC*1,414+100V
=> 800V mosfets

2*230VAC*1,414+100V
=> 800V mosfets

high
voltage 800V fets, >2.0 Ohm

high
voltage 800V fets, >2.0

channel resistance is typical

=> low

Ohm channel resistance is

typical => low

current, high loss

current, high loss

can
be ignored

can
be ignored

some
100W

some
100W

must
be disabled,
otherwise mosfets
explode
if supply <50VDC:
one reverse schottky
in parallel is ok
mosfet body
diode:

if supply >50VDC:
more complicated
one series low voltage
high current
schottky, one
parallel high voltage
ultrafast recovery
diode (<250ns)

realistic power
levels

many
kW

base-feed
transformer core
power handling
what limits the
power level

capability
(saturation, induced
currents
causing core heating)

primary
leakage inductance, huge
voltage spikes (up to kV
range) at increasing
power levels, makes use of
snubber circuits imperative
(=>high heating
losses and low efficiency, and

(same
as for push-pull)

high circuit complexity)

mosfet current
ratings (paralleling
more than two fets,
the right way,
is tricky)

>=800V fets are expensive

and cant handle much
current

mosfet switching and

conduction losses

critical
design, only one primary
needs
only one primary
transformer
design

transformer design
non-critical

needs
two identical and well
coupled primaries, critical
design requires
skills! ;o)

only the first quadrant of

the ferrite cores B-H curve
is used, i.e.
"transformer core running
at only half of what it could
handle".

full-bridge:
minimal danger of
saturation, Vs
imbalance mainly

base feed
transformer
volt-seconds
(Vs) imbalance:

due to slight
differences

major
problems with Vs imbalance

in mosfet channel onresistances

as fully identical pri windings

are almost

half-bridge: if the
primary has a series
coupling capacitor,
then Vs

impossible to make. The

driver circuit absolutely must
have pulse-by-pulse
current limiting.

imbalance is no big
problem

tuneable down
to DC / 0Hz

yes,
by using a primary
series coupling
capacitor
grief
with gate drive

problems

transformers or
floating channel
mosfet driver ICs or
optocouplertweaking

no
(short-circuit at 0 Hz)

grief
with mosfets constantly
dying on overvoltage, gate
drive noise

no
(short-circuit at freq
towards 0 Hz)

(same
as for push-pull)

TL494 magic chip, Part 7

Alternative output connection.
In other post I used classic output connection- collector load. Browsing in the internet I found
Russian schematics of electric shock gun. It is based on the TL494 chip. But the output was
connected in different way- using emitters while collectors were connected to Vcc. Here is the
part of the schematics:

On the testing board everything is working. Quite cool, as we can eliminate one or two
transistors- not need to invert signal to mosfet. Why this type of schematics is not in datasheet?
On

original

Texas

Instruments

there

is NO such

connection.

But on

ON

Semiconductors same device (but with much more bigger values on maximum ratings)
datasheet this is the first schematics! So, RTFM twice

TL494- Power supply for VFD

Typically in self made small hardware (DIY), power from mains transformed and rectified
using simple wall cube. But this cube produces typically 12V. A bit too low for VFD. If I would
like to use some higher voltage cube for 25V it will be hard to find, and to much power will be
dissipated in 5V supply for logic parts. There is one chip in AT/ATX power supply: TL494 (or
analog), pulse width modulator, generator, etc. etc. Switching power supply controller. Using
this chip you can build any switching supply, or PWM regulator. Using typical parts from AT
power supply and powerful mosfet from mainboard I constructed boosting regulator.
Schematics are here.

Press on image of here to see the schematics.

The main parts are: T1, Q1, L1, D1. T1 is used to drive powerful mosfet. Mosfet is connected in
simplified way, so called passive on. It must be driven using totem pole schematics, but who
cares if it works. L1 is some inductor from old HP printer (about 50 turns, 1cm height, 0.5 cm
width with windings, open inductor). D1 is fast schotky diode from some other device. It is
SMD part from Motorola- the device is unknown. (Motorola just loves to encrypt
devicessome top secret paranoia). The TL494 is connected in alternative way. It can be
connected in typical way described in previous tutorials.