TL494

[Pimpom]

OK, we've all been bitching - silently or vocally - about the low
SNR at SED. Here's a real, albeit small, design question.

The venerable PWM controller TL494 was widely used in older
computer power supplies. It differs from the similar SG3524 in
that, in addition to two error amplifiers, it has a dedicated
dead time control (DTC) pin. It allows minimum dead time at 0V
and zero duty cycle at around +3V.

In the PSU designs I've analyzed, the DTC pin is held at 0V in
normal operation and is pulled high under fault conditions such
as overvoltage or failed 'power good' signal.

I'm considering using the TL494 in a minor project with the roles
of the DTC and error amps interchanged. The IC will provide free
running push-pull pulse outputs to drive a power stage with a
fixed duty cycle set by the DTC pin. There will be no feedback
from the output Under normal conditions and the error amps will
intervene only under fault conditions.

There's just one thing I'm not perfectly clear about from looking
at the datasheet: Can the error amps override the fixed duty
cycle setting in case of a fault?

[Tim Williams]

Both inputs (DTC and COMP) act to reduce the output pulse width.

COMP is equivalent to DTC, except it has 0.7V offset to CT (instead of
0.1V), and it has two error amps wired-OR on it. And a slight current sink.

Best, semi-traditional way to use it, is: half bridge forward converter (add
gate drivers and gate drive transformer; controller is secondary side, and
needs an aux supply to start up), current shunt senses output filter
inductor current (or rectifier/secondary ground return), error amp closes
loop on that. Err amp 2 is left disabled (strap inputs to REF/GND). Add
external op-amp for voltage error amp. This breaks the infamous
compensation problem that plagues voltage mode controls, and provides
implicit current limiting (average current mode control).

Otherwise if you just need some PWM, it's a fine starting choice. Easily
frequency modulated, too (a somewhat less convenient basis for a resonant
SMPS, perhaps?). I've used it for induction heaters before.

Tim

--
Seven Transistor Labs, LLC
Electrical Engineering Consultation and Contract Design
Website: http://seventransistorlabs.com

"Pimpom" <Pim...@invalid.com> wrote in message
news:op30k3$pep$1...@news.albasani.net...

[legg]

If you look at the data sheet, you'll see that the DTC and error
amplifier outputs are orred, to turn off the outputs.

What is not normal about the TL494 is that these controls actually
set the threshold of the cycle turn-on; At the start of the clock
period, switches are off until the oscillator output rises
sufficiently to reach the threshold set by both the DT and the outputs
of error amplifier control functions.

The switches turn off at the end of the clock period, when the voltage
on the CT pin external capacitor reaches a maximum control threshold,
set by the internal oscillator.

A turn-on event can be over-ridden by the error amplifiers' internally
diode-orred output (visible at FB pin) during the clock cycle, turning
switches off prematurely, before the end of the clock cycle. Be aware,
however, that the turn-off is not latched in the internal logic. A
dominant error amp output or a controlling dead-time threshold, if
removed, can again allow the switches to turn on before the end of the
clock cycle.

Similar direct control of the output switches can be effected through
the Vz pin.

This lack of cyclic latching can complicate any externally-developed
protection function that is expected to operate on a cycle-by-cycle
basis. In order to do so, this protection signal must be maintained
externally, for the full clock cycle.

If the protection signal is continually maintained, there is no issue,
save the lack of synchronism or slow-restart precautions, on release.

RL

[Pimpom]

On 9/10/2017 7:56 PM, Tim Williams wrote:
> Both inputs (DTC and COMP) act to reduce the output pulse width.
>
> COMP is equivalent to DTC, except it has 0.7V offset to CT (instead of
> 0.1V), and it has two error amps wired-OR on it. And a slight current sink.
>
> Best, semi-traditional way to use it, is: half bridge forward converter (add
> gate drivers and gate drive transformer; controller is secondary side, and
> needs an aux supply to start up), current shunt senses output filter
> inductor current (or rectifier/secondary ground return), error amp closes
> loop on that. Err amp 2 is left disabled (strap inputs to REF/GND). Add
> external op-amp for voltage error amp. This breaks the infamous
> compensation problem that plagues voltage mode controls, and provides
> implicit current limiting (average current mode control).
>
> Otherwise if you just need some PWM, it's a fine starting choice. Easily
> frequency modulated, too (a somewhat less convenient basis for a resonant
> SMPS, perhaps?). I've used it for induction heaters before.
>
> Tim
>

I had more than one application in mind, none of which is fully
thought out yet. One of them is a small simple inverter at mains
frequency instead of the tens of kHz these devices usually
operate at.

In this application, the IC will serve as an oscillator with a
push-pull output at a preset duty cycle. The output will go to a
pair of transistors which in turn drives a step-up transformer.
There will be no attempt to regulate the final output voltage by
PWM except to limit the max current drawn by the transistors.

There are many designs for a basic inverter on the internet but
most of them operate at near 50-50 duty cycle and I don't want
that. This is where the TL494 comes in. It's cheaper than an MCU,
can be 'programmed with a couple of resistors and can run
straight off a 12V battery.

[tabb...@gmail.com]

Such an invertor would suffer deregulation due to battery voltage & transformer copper drops, which added together makes a fair amount of swing. If your app can tolerate that why not go simpler & use an oscillating relay.


NT

[Pimpom]

Thanks. I can see those things in the internal block diagram now
that you've pointed them out.

To reiterate, there will be no attempt to regulate the output
voltage by PWM under normal load. I thought I'd feed a signal to
the error amp proportional to the current drawn by the power
stage and that will have an effect only if the current crosses a
preset threshold..

[Pimpom]

Yes they can. Those effects are fully anticipated.

> why not go simpler & use an oscillating relay.
>

You mean like those electromechanical vibrators from the vacuum
tube days? That's a bit too crude for my purpose. I want a fairly
stable frequency, presetable duty cycle and no mechanical contact
with its attendant problems.

[tabb...@gmail.com]

On Sunday, 10 September 2017 20:45:04 UTC+1, Pimpom wrote:

Yes, those. FWIW duty cycle was adjustable by changing the spacing, at least on some units.


NT

[Martin Riddle]

I've done the exact same thing you are planning to do, it was for a
isolated 485 bus implemtation. Fixed frequency push pull. Works well.
One observation in the differences of the 494 and 3525 is that 3525
has a shutdown pin , soft start, totempol outputs, and sync pins.
But for the price, it's an easy push pull oscillator.

Cheers

[Tim Williams]

"Martin Riddle" <marti...@verizon.net> wrote in message
news:qiibrcp4ip6bqkdjm...@4ax.com...

> I've done the exact same thing you are planning to do, it was for a
> isolated 485 bus implemtation. Fixed frequency push pull. Works well.
> One observation in the differences of the 494 and 3525 is that 3525
> has a shutdown pin , soft start, totempol outputs, and sync pins.
> But for the price, it's an easy push pull oscillator.

If you're looking for the TI equivalent: TL598. Or Unitrode/TI UC3525 (or
UCC, I forget).

Fine for direct driving transformers, too, but add schottky clamp diodes (to
+V/GND) to handle reactive current -- they're bipolar, not like today's CMOS
gate drivers.

[Tim Williams]

"legg" <le...@nospam.magma.ca> wrote in message
news:mbrarcpttaupenukg...@4ax.com...

> This lack of cyclic latching can complicate any externally-developed
> protection function that is expected to operate on a cycle-by-cycle
> basis. In order to do so, this protection signal must be maintained
> externally, for the full clock cycle.

The appnotes show a solution or two, for implementing peak current mode
control. It's a hack at best -- not so much testament to the versatility of
the part, as it is "if all you have is a hammer" syndrome.

I mean, it doesn't take many more 555s to do the same job, either. :)

Also, one of the app note implementations is blatantly wrong. But what do
you expect, the intern probably didn't breadboard, let alone SPICE, the
thing.

[legg]

On Mon, 11 Sep 2017 22:13:18 -0500, "Tim Williams"
<tmor...@gmail.com> wrote:

>"legg" <le...@nospam.magma.ca> wrote in message
>news:mbrarcpttaupenukg...@4ax.com...
>> This lack of cyclic latching can complicate any externally-developed
>> protection function that is expected to operate on a cycle-by-cycle
>> basis. In order to do so, this protection signal must be maintained
>> externally, for the full clock cycle.
>
>The appnotes show a solution or two, for implementing peak current mode
>control. It's a hack at best -- not so much testament to the versatility of
>the part, as it is "if all you have is a hammer" syndrome.
>
>I mean, it doesn't take many more 555s to do the same job, either. :)
>
>Also, one of the app note implementations is blatantly wrong. But what do
>you expect, the intern probably didn't breadboard, let alone SPICE, the
>thing.
>
>Tim

As the part type has been in the field for 40+ years, you'd have to
give a reference to the app note you're talking about. Most issues
relevant to the era are addressed by the mfr using solutions that made
sense for the applications of that time.

The original mfr's app notes pre-date any electronic format (or the
non-academic dissemination of Spice modeling) and may have suffered in
the conversion process. Though they are unlikely to have been written
by an 'intern' (author's names and sources are typically preserved),
the usual precautions for typos in anything passing through the sales
department are required.

Industry and consumer print articles employing the part may come from
just about anywhere and demonstrate a wide range of competence and
accuracy. Differing versions of 'improved' part have also been
released by both the original mfr and secondary sources, over the
years, but these usually have significant part numbering differences
and should be investigated independently.

RL

[Tim Williams]

"legg" <le...@nospam.magma.ca> wrote in message

news:951grchn8d9nmn235...@4ax.com...

> As the part type has been in the field for 40+ years, you'd have to
> give a reference to the app note you're talking about. Most issues
> relevant to the era are addressed by the mfr using solutions that made
> sense for the applications of that time.
>
> The original mfr's app notes pre-date any electronic format (or the
> non-academic dissemination of Spice modeling) and may have suffered in
> the conversion process.

On the contrary; TI's recently updated their TL494 datasheet, in fact!
(They also added a layout section to the LM555 datasheet. Go figure?)

They've been updating a lot of their resources, it seems!

Anyway, as we're talking TLxxx, I'm mainly referring to TI's appnotes. But
there are others, yes (and other errors :) ).

> Though they are unlikely to have been written
> by an 'intern' (author's names and sources are typically preserved),
> the usual precautions for typos in anything passing through the sales
> department are required.

There are relatively few appnotes with authors on them, actually. At least
that I've seen. LT seems to be a standout in that regard (indeed, their
appnotes even have character, from time to time, especially the Jim Williams
ones). TI not so much.

Pretty sure I've never seen an appnote that shows the author's name AND
title / qualifications (not that you couldn't look them up separately, of
course).

And, yeah, I honestly doubt most appnotes are written by interns, but it's a
lie I tell myself to make myself feel better, that a real engineer couldn't
be so ignorant as to write such things.

[legg]

I'm still interested in what was 'just wrong'.

The earliest TL494 app note that was digitized at TI as SLV001 in 1998
dates from 1989 print media, but it originated as Bulletin CA-198 by
John Spencer published in 1978. The external latch is shown as fig27.

The 1998 digitization for SLV001 was image-only, so there was little
chance of 'extra' typographical issues being generated. This was fully
re-edited for later revs to be text-searchable and was still basically
the same information, at revD, in 2005.

Double-pulse suppression was claimed in the part's print data sheet as
early as 1983, but an effective latch is not present in the functional
block diagram of the TL494's (SLVS074) specification sheet, to this
day.

RL

[Jim Thompson]

Tim just being his usual supercilious self >:-}

...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| STV, Queen Creek, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

I'm looking for work... see my website.

Thinking outside the box...producing elegant & economic solutions.

[Tim Williams]

"legg" <le...@nospam.magma.ca> wrote in message

news:hm9grcl72nklrdhb6...@4ax.com...

> I'm still interested in what was 'just wrong'.

Remembered what it was:
http://www.ti.com/lit/an/slva666/slva666.pdf

(Which, speaking of, does in fact have an author. The name shows up on
prabook.com, "Analog applications engineer Texas Instruments India Private
Ltd., since 2007".)

The schematic should speak for itself, but if you'd like to know what I
spotted, let me know.

[legg]

On Tue, 12 Sep 2017 18:00:28 -0500, "Tim Williams"
<tmor...@gmail.com> wrote:

>"legg" <le...@nospam.magma.ca> wrote in message
>news:hm9grcl72nklrdhb6...@4ax.com...
>> I'm still interested in what was 'just wrong'.
>
>Remembered what it was:
>http://www.ti.com/lit/an/slva666/slva666.pdf
>
>(Which, speaking of, does in fact have an author. The name shows up on
>prabook.com, "Analog applications engineer Texas Instruments India Private
>Ltd., since 2007".)
>
>The schematic should speak for itself, but if you'd like to know what I
>spotted, let me know.
>
>Tim

Looks functional.

You may think he's incorrectly grounded the OC pin, but he's handling
maximum duty cycle and slow-start via DT pin bias.

The fault limiting will have issues, however.

Fighting the output emitter for control of the gate drive under
peak-current-fault conditions is a bad idea. If the gate drive
turn-off was latching, rather than just a pnp pull-down, or the
control supply wasn't bootstrapped, it could conceivably work; but
you'd be right in trashing that as a useful fault limiting circuit.

It wouldn't prevent basic operation of a breadboard, though, and he
has pictures to prove it.

RL

[Tim Williams]

"legg" <le...@nospam.magma.ca> wrote in message

news:curgrc1p29fq757or...@4ax.com...

> Looks functional.
>
> You may think he's incorrectly grounded the OC pin, but he's handling
> maximum duty cycle and slow-start via DT pin bias.
>
> The fault limiting will have issues, however.

A circuit that doesn't have functional limiting or fault protection, is not
a functional circuit, sorry. :-)

Not if I'm reviewing it.

> Fighting the output emitter for control of the gate drive under
> peak-current-fault conditions is a bad idea. If the gate drive
> turn-off was latching, rather than just a pnp pull-down, or the
> control supply wasn't bootstrapped, it could conceivably work; but
> you'd be right in trashing that as a useful fault limiting circuit.

It would've been slightly more okay, if the clamp transistor were connected
just after the series resistor. It would have some hope of fighting the
emitter output, then.

But you missed the best part: the LM339 is open collector. There's no
pull-up!

> It wouldn't prevent basic operation of a breadboard, though, and he
> has pictures to prove it.

Again, if there's a limit function, then you must also perform a test, to
prove it is functional!

Short the output and watch the smoke rise. :)

(The implied irony being, if there isn't a limit function, you don't need to
test it! Well, that's actually fine -- as long as the end user _fully_
understands the consequences. But, accident-prone users (which is everyone)
aren't a good match for unprotected circuits, so I cannot abide such a
conclusion.)

[legg]

On Thu, 14 Sep 2017 06:19:38 -0500, "Tim Williams"
<tmor...@gmail.com> wrote:

>"legg" <le...@nospam.magma.ca> wrote in message
>news:curgrc1p29fq757or...@4ax.com...
>> Looks functional.
>>
>> You may think he's incorrectly grounded the OC pin, but he's handling
>> maximum duty cycle and slow-start via DT pin bias.
>>
>> The fault limiting will have issues, however.
>
>A circuit that doesn't have functional limiting or fault protection, is not
>a functional circuit, sorry. :-)
>
>Not if I'm reviewing it.
>
>
>> Fighting the output emitter for control of the gate drive under
>> peak-current-fault conditions is a bad idea. If the gate drive
>> turn-off was latching, rather than just a pnp pull-down, or the
>> control supply wasn't bootstrapped, it could conceivably work; but
>> you'd be right in trashing that as a useful fault limiting circuit.
>
>It would've been slightly more okay, if the clamp transistor were connected
>just after the series resistor. It would have some hope of fighting the
>emitter output, then.
>
>But you missed the best part: the LM339 is open collector. There's no
>pull-up!
>

Yeah, well, the protection cct doesn't function, period.

Odd to see a 2A schottky on a 3A flyback output.

RL