more modern version of tl494?

[2strokeforever]

Hello, I am looking for a more modern and capable version of the tl494 or sg3525 that is still reasonably simple to use.

I would like to hear if there are better options, or should I stick with the tried and true.

Big picture im trying to settle on a PWM IC to be powered from pretty much any voltage over 10v (with dc dc isolator if needed)

The output of the IC is only to power an isolated gate driver, or 2, any current measuring will be done with an acs758 hall sensor.

Frequency will be below 100khz

Depending on the project the power level will be from 1kw to 20kw

Heres the wants:

Reasonably sized IC, dont like the vqfn style/ leadless case that is difficult to solder.
Single ended or push pull outputs
Access to both inverting and non inverting pins of the error amplifiers

Something with a few examples out there would be nice too since I am not an expert.

Let me know what you think.

[coromonadalix]

wide input voltage of ??   current ??    you miss some vital point  ....

[CosteC]

You mention 1 kW-20 kW PSU and you look for simple IC... This is rather contradictory as there are no simple things at 20 kW level.

Selection of IC is based on topology you chosen for your SMPS. I do not know if you would like to make AC->DC power supply or something else.
Infineon, among many others, has good portfolio of power ICs which may suit you. Now it is quite popular to use specialised DSPs to control such large power units.

[coromonadalix]

even at 1kw   ...  lots of things are involved    input stages,  PFC  if there is,  main power section ...

not a project for beginner   

[2strokeforever]

My applications will always be DC input, could be ac or dc out, usually a buck converter.

I could see it being extra complicated to make a multi KW converter cheap, compact, accurate, and efficient.

But for a one off where none of that is a concern had some extremely crude designs work well enough by simply oversizing the fets.

[strawberry]

buck with synchronous rectification for output voltage below ~20V
can use current mode controller , TL494/7500/3525 is voltage mode controller
small transistor silicon chip mean larger thermal resistance or silicon temperature = lower reliability

[Doctorandus_P]

What's wrong with the tl494?

[xavier60]

I read somewhere that newer TL494 ICs now have multiple pulse suppression to prevent noise causing unwanted toggling of the steering flip-flop. Has anyone confirmed this?

[2strokeforever]

What's wrong with the tl494?

The output is slow archaic junk.

I have heard other people say fairly often that there are much better PWM IC out there, but nobody seems to be able to point to one with access to both inverting and non inverting pins of the error amplifiers.
I have spent many hours on digikey looking, but havent had much luck.

Voltage feed-forward would be nice

and pulse skipping too

The Tl598 seems to be the tl494 with a better output section.

[xavier60]

What's wrong with the tl494?

The output is slow archaic junk.

I have heard other people say fairly often that there are much better PWM IC out there, but nobody seems to be able to point to one with access to both inverting and non inverting pins of the error amplifiers.
I have spent many hours on digikey looking, but havent had much luck.

Voltage feed-forward would be nice

and pulse skipping too

The Tl598 seems to be the tl494 with a better output section.

And states "Internal Circuitry Prohibits Double Pulse at Either Output"

[T3sl4co1l]

Not much I think, that is still simple enough to have sort-of-uncommitted error amps, perhaps the UC(C) relatives -- but they're not exactly much newer, and are legacy priced.  (At least cost doesn't matter for a "one off".)  UC3525 does go much faster (500kHz osc).

Can always make one yourself, it's nothing more than a couple op-amps, comparators, current mirrors, and logic.  An embedded version might even be feasible, if you feel confident enough to chance the survival of a 20kW inverter upon the reliability of software...

(I've written an embedded 1kW resonant control, myself; I'm fairly meh on its overall quality: it seems to work, it's not the smoothest (i.e. low noise, at least in part due to quantization but also probably sampling dither), but I don't exactly have any way to prove its correctness, or [begin to] fuzz its immunity to other software glitches or input errors.)

Tim

[2strokeforever]

Thanks, thats exactly what I needed to know.

Software control is way above my pay grade, and I would rather have more reliability.


UC3525A has:

• Pulse-by-Pulse Shutdown
• Latching PWM to Prevent Multiple Pulses

TL598 has "Internal Circuitry Prohibits Double Pulse at Either Output"

from what I can tell the TL598 can be used to limit pulse current

and an extra error amplifier


So far I am leaning towards the TL598 unless someone has good reasons to that the UC3525A might be better

[T3sl4co1l]

AFAIK, 598 is just 494 with totem pole outputs -- better for gate driving.  But you might want a proper driver on it, still, anyway; it's old bipolar stuff.

Tim

[2strokeforever]

AFAIK, 598 is just 494 with totem pole outputs -- better for gate driving.  But you might want a proper driver on it, still, anyway; it's old bipolar stuff.

Tim

From the data sheets the rise and fall times were much better than the 494, some of my preferred gate drivers require a sharp edge to trigger on and this will help.

I never drive anything without a proper modern gate driver

[xavier60]

I have a strong preference for the UC3525.
It is reassuring to see in the internal block diagram, the steering flip-flop being clocked directly by the oscillator and also the latch at the comparator's output.
Not so for the TL598 despite the claims. The extra opamp might seem useful but can be difficult to achieve fast current limiting with it.
With the UC3525, I implement fast pulse-by-pulse limiting by pulling down the soft-start pin.

[David Hess]

Unitrode had advanced replacements for the TL494, but after Texas Instruments bough them their selection guides were removed.  I think they had a current mode control version of the TL494.

[2strokeforever]

I have a strong preference for the UC3525.
It is reassuring to see in the internal block diagram, the steering flip-flop being clocked directly by the oscillator and also the latch at the comparator's output.
Not so for the TL598 despite the claims. The extra opamp might seem useful but can be difficult to achieve fast current limiting with it.
With the UC3525, I implement fast pulse-by-pulse limiting by pulling down the soft-start pin.

Great info, I have heard a few people saying their problems went away after switching from the TLxxx to the 3525 series, this could explain it.

From the  UC3525 datasheet
All transitions of the voltage on the shutdown pin should be within the time frame of one clock cycle and not
repeated at a frequency higher than 10 clock cycles.

Does this pose a problem to fast current limiting, or do you hold the voltage low for 10 cycles?

Thanks

[xavier60]

I have a strong preference for the UC3525.
It is reassuring to see in the internal block diagram, the steering flip-flop being clocked directly by the oscillator and also the latch at the comparator's output.
Not so for the TL598 despite the claims. The extra opamp might seem useful but can be difficult to achieve fast current limiting with it.
With the UC3525, I implement fast pulse-by-pulse limiting by pulling down the soft-start pin.

Great info, I have heard a few people saying their problems went away after switching from the TLxxx to the 3525 series, this could explain it.

From the  UC3525 datasheet
All transitions of the voltage on the shutdown pin should be within the time frame of one clock cycle and not
repeated at a frequency higher than 10 clock cycles.

Does this pose a problem to fast current limiting, or do you hold the voltage low for 10 cycles?

Thanks

I only use the soft-start pin 8. It only acts on the comparator input but still terminates the present half cycle quick enough.
A capacitor is normally placed from pin 8  to ground. I include a series resistor so that the pin can be quickly pulled to ground without completely discharging the capacitor with a small transistor or possibly a comparator's open collector output.
The transistor or comparator will be monitoring the rectified, loaded and lightly filtered output of a current transformer sensing primary current.
My application is forward half-bridge converters.
 It provides short term pulse-by-pulse limiting for each half-cycle until a slower main control loop can take control of output current. A battery charger for example.

[T3sl4co1l]

Unitrode had advanced replacements for the TL494, but after Texas Instruments bough them their selection guides were removed.  I think they had a current mode control version of the TL494.

UCC3808 is kinda like that, but a stripped-down (8 pin) version, not as flexible.

Tim

[Hiemal]

You can use any PWM IC if you just short the FB pin to ground, and then drive the COMP pin with an external error amplifier.

This is exactly what most isolated switchers do actually. The good ol' TL431 acts as the error amplifier on the secondary side, which drives an optocoupler that exerts itself on the COMP pin, which changes the duty cycle output. The comp pin is usually high impedance, and thusly will follow pretty much whatever voltage you command into it with an op amp. Most (good) datasheets will have a parameter or a note somewhere telling you what voltage on the comp pin corresponds to what duty cycle.

So, you can achieve the same exact thing just using an op amp to drive the comp pin directly, using whatever PWM chip you want.

[profdc9]

I hacked high-side pulse-by-pulse current sensing into a TL494 buck converter power supply

https://www.github.com/profdc9/FlexibleCharger

I have a load resistor on the high side switch connected to a PNP transistor, which pulls up on the dead time control pin to shut off the transistor if the current exceeds a threshold (the load resistor multiplied by the threshold current equals 0.7 V).  This is especially needed because if you short the output, once the short is removed, a large inrush current occurs through the transistor to recharge the output capacitor.  Perhaps a high-side Hall effect current sensor would be better to eliminate the power loss due to the extra resistor.  Most ATX power supplies that used the TL494 simply shut down using a SCR and had to be power cycled before they would be turned on again, which would reenable the soft start, and so didn't have this problem.

I made this power supply as something that would be made out of basic generic parts but be programmable as a battery charger with an Arduino, as TL494 is available in generic form (so is UC3846, SG3525, etc.).

[T3sl4co1l]

Yeah, Hall effect is fine, or one of those current shunt amp / level shifter devices.  Bandwidth a bit over Fsw is sufficient.

Peak current is easier done with a UC3842/3, and level shifters; including a high side gate driver (BJTs are slower, MOSFETs preferred) probably with a charge pump to allow operation from 0-100% duty.

Or make it flying, run the UC384x from bootstrap power and use an opto to do voltage feedback.  Harder to do adjustable current as a nice CC/CV supply though.

For a CC/CV, I'd rather average current mode control, and a 494 or even a discrete solution is fine.

Tim

[2strokeforever]

Does anyone have experience with the SG3524 and SG2524? It looks to be a SG3525 with current limiting.



Ti datasheet says this, but is it true?
In addition to constant-current limiting, CURR LIM+ and CURR LIM– also can be used in transformer-coupled
circuits to sense primary current and shorten an output pulse should transformer saturation occur.

[T3sl4co1l]

Temperature!

A hint: at least among older devices, or from some manufacturers (National for instance), the first digit describes temperature rating.  LM117/217/317 for example, the regulator for mil/industrial/commercial temperature ratings.  Most parts these days are already made to run up to 125 or 150°C so it's not a big deal, but they were a bit more limited back then.

SG2524 happens to be rated 85°C; 3524, 70°C.  Unitrode also used this scheme, with UC1842 rated 125°C; 2842, 85; 3842, 70.

The SG prefix was SGS Thompson, bought by, various I think?  ST Micro (well, ST is "SGS Thompson" renamed..!) and on semi at least.  Several second-sources have made them over the years, and TI remains among them today.

It doesn't always hold, of course; LM2903 is basically a LM193, in plastic package, with moderately degraded characteristics.  After they used up the first few hundred numbers, National I think got a bit more free-form with their numbering scheme.

Tim

[2strokeforever]

Anyone use the UC28023 or UC28025?

They look to be more modern, with voltage feed forward, much higher frequencies, and access to both inverting and non inverting pins of the error amplifiers. (except when using double ended mode)

Unfortunately there isnt any info about them, not even an eval board, and only 1 example circuit.