TL494

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?

Reply to
Pimpom
Loading thread data ...

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 
 Click to see the full signature
Reply to
Tim Williams

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

Reply to
legg

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.

Reply to
Pimpom

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

Reply to
tabbypurr

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..

Reply to
Pimpom

Yes they can. Those effects are fully anticipated.

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.

Reply to
Pimpom

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

NT

Reply to
tabbypurr

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

Reply to
Martin Riddle

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

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

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

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

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

Reply to
legg

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 :) ).

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.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

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

Reply to
legg

Tim just being his usual supercilious self >:-} ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

Remembered what it was:

formatting link

(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

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

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

Reply to
legg

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

Not if I'm reviewing it.

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!

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.)

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

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

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

RL

Reply to
legg

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.