Hi everyone!

I'm trying to make a simple 200W 24VAC to 12VDC buck converter based on TI TL494 IC and IR2110 MOSFET driver, for educational and (partly) practical purposes. I designed the circuit and simulated it in LTSPICE (where it works flawlessly), drawn schematics and designed the PCB, etched it and soldered everything together, but I ran into a peculiar problem. The switching doesn't work and all I get on output is some 9VDC. After scoping some signals, I've been able to pinpoint the problem to the weirdly distorted waveform on the pin 5 of the TL594 (I'm using it instead of the TL494 due to more precise reference voltage output, but it shouldn't be a problem). The waveform should be a clean sawtooth from 0 to 3V, of desired frequency (25 kHz in my case, although I tried 50kHz with the same result), but it is distorted and biased as in the attached scope screenshot. Naturaly, as the result of that the PWM signal from the TL594 emitter output is also distorted and biased (another screenshot), so the IR2110 can't be driven properly, and consequently also the MOSFETs.

I tried changing the combination of the timing capacitor and resistor, with the same result. So, I'm clueless as what to try next. Could it be the parasitic capacitances and inductances, even at this relatively low frequency?

I attach the circuit schematics and photos of front and back sides of the finished circuit.

Any help or even a clue will be appreciated. Thanks in advance!

TL594 Pin 5 (Timing Capacitor):


TL594 Pin 9/10 (Emitter Output):


Schematics:

Front:






Back:

 

There seems to be something peculiar about how you have connected the IR2110.
HO goes to the high-side gate
VS goes to the high-side source
The bootstrap capacitor goes between Vs and Vb (Vb is the boosted supply)

 

You're right, of course, thank you for your remarks! I don't know what was I thinking, I just automatically drew wires although I studied the IR2110 datasheet beforehand, and did not do the final check. Anyway, I corrected the connections with the help of a scalpel and some jumper wires on the PCB, so my IR2110 is now connected as in the attached image (of course, it's much uglier now, and I'm aware I introduced new parasitics). I also replaced the TL594 and IR2110 ICs with new ones (in case I fried the old ones). But to no avail - the timing waveform is unchanged, still distorted with identical bias and peak values as in the former case.

 

From the pics it seems you used a 1X scope probe. Do you see the same waveforms using a 10X probe?

 

I don't understand why you don't just drive the Mosfet directly from the TL494 outputs pins/10, like most switch-mode psus do?

 

Maybe because it has to be high-side switched for a buck regulator.

 

I don't understand why you don't just drive the Mosfet directly from the TL494 outputs pins/10, like most switch-mode psus do?

All the TL494-based switch-mode PSUs I've seen use BJTs as switching devices. As the datasheet states:
"Two output transistors are available on the TL494. Both transistors are configured as open collector/open
emitter, and each is capable of sinking or sourcing up to 200 mA. The transistors have a saturation voltage of
less than 1.3 V in the common-emitter configuration and less than 2.5 V in the emitter-follower configuration.
The outputs are protected against excessive power dissipation to prevent damage, but do not employ sufficient
current limiting to allow them to be operated as current-source outputs."

So, I presume that TL494 couldn't provide enough current to drive two MOSFETs, and maybe it would even fry if I tried to drive a MOSFET directly.

Also, as already noted, it's high-side switching with n-channel MOSFETs, so level shifting should be employed. Could have used the p-channel ones, though...

I also prototyped the previous version of this circuit with a discrete MOSFET driver (several BJTs and a bootstrap capacitor between TL494 outputs and FET gates), but it didn't work as planned due to excessive ringing and BJT overheating/blowing up, so I reworked it to use a dedicated IC driver.

Will try scoping with the 10x probe later when I get home. I'll report.

 

What is the design requirement for max current output?

Also, I suggest re-arranging the Vsense (FB) signal to this:

 

What is the design requirement for max current output?

Also, I suggest re-arranging the Vsense (FB) signal to this:

View attachment 305425

Thanks for the suggestion regarding the Vsense voltage divider. I see that this would give me greater adjustment sensitivity around the nominal output voltage, which is a good idea.

As for the max current output requirement, it is not too strict, but it's some 20-22A. The reason for not stating the exact value is the fact that I'm aware I can't be too precise in sensing the current by means of the smallish 10mOhm shunt that I'm using for the purpose, before amplifying the Isense voltage to around 2.5V, to be able to compare it with the voltage reference in the TL494 error amplifier. I chose to use such small value to minimize losses.

 

From the pics it seems you used a 1X scope probe. Do you see the same waveforms using a 10X probe?

OK, so this is so very embarassing. You were right to suspect the probe attenuation ratio as the source of my initial problem. It seems that several days ago I inadvertently flipped the attenuation switch on the probe to 10x while the scope channel was indeed set to 1x. After I adjusted it back, the waveforms of the TL494 timebase and the PWM output show as correct.





However, this does not solve the problem of the missing voltage on output, so now I should dig in some more.

I suspect a faulty IR2110, since it was the only component I bought some time ago from an unverified source (AliExpress), while everything else is from Mouser. Frankly it would not surprise me if the whole batch was faulty or fake. I'll have to check them individually on breadboard.

 

Is your bootstrap capacitor getting charged?
Normally, it would be charged when the low-side MOSFET switches on, but you have no low-side MOSFET because it is not a synchronous converter.
Have a look at the LT7001 because it doesn't need the low-side MOSFET to charge the boostrap capacitor.

 

After I adjusted it back, the waveforms of the TL494 timebase and the PWM output show as correct.

Much better!

 

Hi everyone!

I'm trying to make a simple 200W 24VAC to 12VDC buck converter based on TI TL494 IC and IR2110 MOSFET driver, for educational and (partly) practical purposes. I designed the circuit and simulated it in LTSPICE (where it works flawlessly), drawn schematics and designed the PCB, etched it and soldered everything together, but I ran into a peculiar problem. The switching doesn't work and all I get on output is some 9VDC. After scoping some signals, I've been able to pinpoint the problem to the weirdly distorted waveform on the pin 5 of the TL594 (I'm using it instead of the TL494 due to more precise reference voltage output, but it shouldn't be a problem). The waveform should be a clean sawtooth from 0 to 3V, of desired frequency (25 kHz in my case, although I tried 50kHz with the same result), but it is distorted and biased as in the attached scope screenshot. Naturaly, as the result of that the PWM signal from the TL594 emitter output is also distorted and biased (another screenshot), so the IR2110 can't be driven properly, and consequently also the MOSFETs.

I tried changing the combination of the timing capacitor and resistor, with the same result. So, I'm clueless as what to try next. Could it be the parasitic capacitances and inductances, even at this relatively low frequency?

I attach the circuit schematics and photos of front and back sides of the finished circuit.

Any help or even a clue will be appreciated. Thanks in advance!

TL594 Pin 5 (Timing Capacitor):
View attachment 305350

TL594 Pin 9/10 (Emitter Output):
View attachment 305351

Schematics:
View attachment 305352

Each one of the mosfets is good for 80-110A max, so why are there two in parallel?

 

Each one of the mosfets is good for 80-110A max, so why are there two in parallel?

The main reason for two mosfets instead of one was to distribute the heat dissipation losses and consequently utilize smaller heatsinks. The first version of the circuit used a driver made from discrete components which (at least in simulation) weren't able to provide rise and fall times short enough for mosfets to minimize the periods spent in linear region. With driver IC there's probably no need to use two mosfets, since the transitions are fast enough to keep the losses at minimum, but I thought it wouldn't hurt to keep them.

 

The main reason for two mosfets instead of one was to distribute the heat dissipation losses and consequently utilize smaller heatsinks. The first version of the circuit used a driver made from discrete components which (at least in simulation) weren't able to provide rise and fall times short enough for mosfets to minimize the periods spent in linear region. With driver IC there's probably no need to use two mosfets, since the transitions are fast enough to keep the losses at minimum, but I thought it wouldn't hurt to keep them.

There are conduction losses which decrease as you put more MOSFETs in parallel, and there are switching losses which increase, due to the increased capacitive load on the driver. Somewhere there is an optimum, but putting ever more MOSFETs in parallel doesn't keep reducing the losses.

 

OK, based on yesterday's comments (for which I am truly thankful) about the IR2110 being unable to work in non-synchronous buck topology , I've managed to obtain several IR2117 driver ICs. The IC datasheet states that it utilizes a floating channel designed for floating operation, being capable to drive a n-channel MOSFET in high or low configuration. Furthermore, the International Rectifier application note AN-978 (https://www.infineon.com/dgdl/Infin...N.pdf?fileId=5546d4626c1f3dc3016c47de609d140a), in Chapter 10 on page 22 shows an example of a buck converter with the high-side drive function performed by the IR2117.
So I reworked the PCB again ("frankensteinized" it with scalpel and jumper wires, to be more exact) to adjust for pinout mismatch between two driver ICs and changed the bootstrap capacitor supply voltage source to a higher one, as per schematics:



However, still no luck. I get 5.9V on output instead 12V-ish, and it doesn't react to voltage regulation by means of the RV102 pot. I'm not much smarter after scoping the relevant signals:

IR2117 IN pin:



IR2117 HO pin:



MOSFET V_GS:




IR2117 VB pin:




MOSFET V_DS:



MOSFET V_S:



The detail that immediately catches the eye is the large amount of ringing, which I attribute to my rats' nest of jumper wires which I soldered to rework the PCB according to the new schematics. So, I presume that it will be gone once I order the final PCB.

The other thing is V_GS insufficient to fully turn on the MOSFETs, rising only to 3.8V. Furthermore, the width of the V_GS signal doesn't match that of the driver output (if it did, the output voltage would probably be correct). Could it be that the bootstrap capacitor doesn't charge to a sufficient voltage? Maybe it's wrong to connect it directly to the rectifier output (which is also the VS of the mosfets). I'm pretty much clueless at this point.

 

There are conduction losses which decrease as you put more MOSFETs in parallel, and there are switching losses which increase, due to the increased capacitive load on the driver. Somewhere there is an optimum, but putting ever more MOSFETs in parallel doesn't keep reducing the losses.

I see what you mean. So it's basically trial and error, or a true in-depth insight of the components utilized, together with a capable simulator software to determine the optimum...

 

The ringing is normal (and harmless)

 

..
Furthermore, the width of the V_GS signal doesn't match that of the driver output (if it did, the output voltage would probably be correct).
...

I meant "input"...

 

Your post #16 scope shots indicate that the bootstrap isn't working. Missing connection to the bootstrap cap?
If the bootstrap were working, HO should be giving about 48V peak (according to my LTspice simulation). However, VBS would then be about 27V, violating the 25V Absolute Maximum spec for the 2117 (or the 2110).

Edit:
The decaying ringing in your scope shots suggest that diode 105 could be leaky. It is being used very close to its rated reverse voltage limit (30V).