1

My boost converter boosted voltage for a few seconds but it made a crackling sound, probably burnt. It doesn't switch the MOSFET now and TL494 is getting hot from the upper side which is where error amplifier pins are placed. The schematic is below:

enter image description here

I didn't have any input resistors for TL494 error amplifiers and I would like to know if they are relevant. Because in every application I see on the internet, even in the TL494 datasheet there is some resistor between the ref pin and error amplifier pin.

By the way, I had to invert the pin of the second error amplifier grounded beforehand which actually is a bad idea because even though the error must be zero in the ideal case, it was overwriting the feedback. Basically FB pin was always 4.8 V no matter what is applied form PIN1. By connecting PIN15 to ref pin as in the schematic, I managed to solve that issue until the crackling sound, now FB pin is fixed to 4.8 V again even though PIN1 is lower than PIN2.

I also made a test where I pulled FB pin to ground with an external wire to test MOSFET, which worked but this might have damaged the error amplifier. I want to replace the TL494 with a new one but I must be sure that input resistors aren't necessary. Otherwise, it will be extra work to create paths for resistors between REF-PIN2, REF-PIN15, PIN16-GND on an already printed PCB.

Thanks.

Edit:

Specs:

  • Vi: 7V-18V
  • Vo: 14V-100V
  • Max Ii: 10A
  • Max Io: 8A
  • Max Power: 120W

I added the resistors for OpAmp inputs as below: enter image description here

Edit 2:

It wooooorks, I replaced TL494 with a new one. I'll halve the inductance because they have 0.2ohm DC resistance and it creates loss at high current. Inductor current ripple will be %35 at most I think this is feasible.

CaveScientist
  • 375
  • 1
  • 10
  • What’s your input voltage? – winny Dec 08 '22 at 20:04
  • 7V-18V but mostly 12V – CaveScientist Dec 08 '22 at 22:45
  • That puts you on 0 % margin. This won’t work. – winny Dec 08 '22 at 23:24
  • 7V and 18V are absolute max. This will be used in a 12V car battery, so there is a margin. – CaveScientist Dec 08 '22 at 23:33
  • Car battery, or *car*? Automotive voltages can swing pretty high at times, even discounting jumpstart and load dump conditions. (Just a battery and charger should be fine though.) – Tim Williams Dec 09 '22 at 05:16
  • Oh, I didn't think of that. This is originally a school project but I wanted to use it in a car. What can I do against the voltage swings? Can I increase input capacitance, also could I add an input diode to stop current flow from input cap to battery when it experiences voltage drop? – CaveScientist Dec 09 '22 at 10:39
  • I must know how long does the battery voltage drop last to know how much capacitance I need, right? Is it whole duration of engine start? Because if it is, that's too long. – CaveScientist Dec 09 '22 at 12:58

3 Answers3

3

There are several potentially fatal oversights in this design:

  • Supply is unstated. UC4429 max is 18V. Any input or overshoot beyond this limit will destroy it.
  • Nominal output is unstated -- and can't even be determined from the feedback divider as it's composed of three potentiometers, one not even on the board(!!!).
  • There is no current sense, so there is nothing preventing the transistor from pulling nearly short-circuit input current, and likely destroying itself in the process.
  • There is no control compensation, so the TL494 seems to be used in a hysteretic mode instead. It will ping-pong between zero and full output (~97% duty?). There is soft start (DTC), so perhaps it's able to start up once, but almost any load seems likely to cause destruction.
  • R4 is quite large, giving a very slow risetime. Perhaps this makes the gate driver noise-sensitive as well, in which case it could oscillate at some ~MHz during the collector rising (gate falling) edge.
  • C10 and C12 also seem rather small, but with no part number given, nor nominal output voltage, this isn't necessarily a problem.
  • Likewise, it isn't clear whether the output filter helps; it's very easy, for example, to accidentally ground both sides (through external ground paths between source and load), shorting out the CMC and making it redundant.

These are approximately in order of severity.

Solutions include:

  • Specify and observe maximum input voltage. Likely the large (4700uF, electrolytic??) capacitor dampens overshoot, but this should be verified as well.

  • Limit maximum output voltage by designing a feedback divider using two fixed resistors, and one adjustable resistor in parallel with the top feedback resistor. This way, the output can not shoot up an unlimited amount if the adjustable resistor becomes disconnected (it is a common failure mode that the wiper terminal momentarily goes open-circuit).

  • Tl494 is traditionally a voltage-mode controller -- which is reason enough not to use it at all, and I would strongly recommend UC3843 or LM3481 in this application.

    TL494 can be used as peak or average current mode control, with some difficulty (see: SLVA001E, Designing Switching Voltage Regulators With the TL494 (Texas Instruments), especially section 4.4). Average current mode is the most reasonable, which requires some means of sensing inductor current: this can be a high-side current sense amplifier (e.g. INA180), Hall-effect sensor, or low-side current shunt. (The CMC implies the input and output shall not be common-ground, which would also be a perfect place to use a low-side current shunt.)

    Note that, in average current mode control, the internal error amp (only one is required, strap the other to VREF/GND to disable it as recommended) serves to control inductor current; a separate, external error amp is required to regulate output voltage. This can be a single op-amp (e.g. TLV2371), or a combined reference and error amp (like TL431).

    In both cases, compensation is required. Follow recommended error amp circuits; likely a type 2 compensator (resistor to -IN, R+C from OUT to -IN) will suffice for one or both (current and voltage, when applicable) error amps.

Tim Williams
  • 22,874
  • 1
  • 20
  • 71
  • I added the specs to the post. For output voltage limitation I say: Duty cycle must be %88 for 100V output, do boost converter beyond that duty cycle? I thought it is limited naturally. You are right with current limitation, I could use a current sense resistor, for this design let's ignore it. I'll do it when I improve myself. Can you explain compensation more because I really seen the operation. FB pin has 4.8 or 0.12V but nothing in between, it passes that phase really fast. In the datasheet it states error amplifiers have really large gain. – CaveScientist Dec 08 '22 at 23:20
  • I thought this is how controller is compensated internally and I didn't make it a trouble. Do you recommend me a way to lower the change in voltage at error amplifier inputs so that it won't jump between high and low? What resistor can I add to change the poles of the system. This is the only way I know TL494 works and I didn't know any way to manipulate the controller that's why asked a lot. I edited the question to contain my new circuit with added input protection resistors, no big change. – CaveScientist Dec 08 '22 at 23:22
  • TL494 does not contain internal compensation, you have to provide those elements yourself. If you want any hope of controlling input and/or output current to the limits shown, you MUST have a current sense -- perhaps two, even. – Tim Williams Dec 09 '22 at 05:14
  • Are you sure that TL494 error amplifiers have some sort of Type-2 or Type-3 compensators. Basically an internal OpAmp, few internal resistors and capacitors to determine the settling time, rising time, gain and phase of the FB signal? – CaveScientist Dec 09 '22 at 15:07
  • They are compensated in the sense that they are unity-gain stable. The intent is that this bandwidth is greater than what the control loop requires, and this is almost always the case. So, additional external compensation is required, and this is what I am referring to. In your last schematic, R6-C4 constitute a type 1 compensator. – Tim Williams Dec 09 '22 at 19:36
  • I also have a cap (C4) between Fb and pin2 which will add the system a pole or zero. Or for example C11 will decrease oscillations. Basically I've put a compensator without doing the math but rather by senses. So I must do a better math and select components between pin1,pin2,pin3. This will give me a better controller with required gains plot, for example it will attenuate high frequencies or it will have faster transients. I can learn to do the math, I just want to make sure I understand "what I don't know yet" correctly. – CaveScientist Dec 10 '22 at 01:06
  • C11 will increase oscillations, actually (reduced phase margin). I suggest concentrating on the dynamics of the switching loop itself, first, and understanding why I emphasize current mode control. Then move on to control theory, poles and zeros, and compensate it for good regulation and stability. – Tim Williams Dec 10 '22 at 04:43
  • I looked for current mode control, its enlightening thanks. I'll definetelly implement it on my next design. C11 adds a phase delay to feedback voltage thus probably increases oscillations. So should I just remove it? Does it have any useful purpose? – CaveScientist Dec 11 '22 at 21:45
  • C11 could be changed to an R+C, and moved to RV1, where it increases phase margin instead; combined with an R+C instead of C4, and with C12 having notable ESR, this is the usual way a voltage mode loop is stabilized. – Tim Williams Dec 12 '22 at 03:02
1

Problem Scope:

Supply Voltage ..................................................... +20V
Input Voltage ........................................................-5V to VDD + 0.3V
Input Current (Vin > Vdd)................................ 50 mA

Perhaps you may be correct that your test method damaged U1 with the reactive load and low ESR.

7 Specifications 7.1 Absolute Maximum Ratings

over operating free-air temperature range (unless otherwise noted)
                                 MAX  UNIT  
VCC Supply voltage               41   V  
VI Amplifier input voltage VCC + 0.3  V  
VO Collector output voltage      41   V  
IO Collector output current     250   mA

Your question is "do I need the 50 k resistor to protect the inputs for U2?". enter image description here

Inputs U2-2 and U2-15 control dead-time must not exceed Vcc +0.3V while the source is U2-14 REF = 5V internal LDO output and you have large caps on the input.

The 5V output cannot exceed the input Absolute Maximums. But with parasitic capacitance and switch step outputs and flyback nH voltages, ESD issues, who knows?

I suspect they wanted to measure if ESD diode inputs were drawing any current, so they added a testable input with a resistor. You must consider this feature as part of your Design for Testability (DFT). But I don't think that is the thermal fault. MY guess is the dual collectors are driving enough power to cause this due to U1 failure. C5 = 0.1uF seems like overkill.

You could put a Resistor from 5V to the capacitor instead to match power on conditions for the differential inputs as desired for DTC. (verify RC time constant)

enter image description here enter image description here

Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
  • Considering output source current of Feedback pin is 2mA stated in datasheet, I might have broken it yes. The capacitor placement you recommended seemed nice to me, I can't change the PCB right now but next time definitely will consider. I'll use brand new TL494 with the edited schematic on the original post. I don't know if electrostatic discharge will cause 12V exceed 42V max rating but it seemed unlikely to me. – CaveScientist Dec 08 '22 at 23:32
1

After looking at the design in more detail, I can see every reason why it fails. The Figure of Merit (FM) that makes SMPS servo's a good design is the phase margin in the loop and the Q ratio of stored energy/load ratio during flyback operation. Unless there are compensation methods such as dumping excess energy, phase-lead compensation in feedback, Mode switching from 100 KHz PWM to lower frequency PFM, crowbar protection, OVP,UVP,OCP,OTP the design will eventually crash and burn if not sooner.

I concur with most of Tim's observations. In addition, these are constructive critiques but harsh. SMPS design is not trivial. Don't take this a personal attack, rather, nice try.

. Pots imply a huge loop for coarse and fine control when you should be using a 10T pot. I suspect your layout was prone for crosstalk and ripple, just from judging the schematic.

  1. No design specs and an obscure reaction to look for microscopic error as in your title. Design specs should fill a page.
  2. No details of voltage on the schematic.
    • minor errors = No reason for 4 inductors. You get the same L/R ratio and a lower resonant frequency with 2 LC's in parallel with DCRs. SRF < 1MHz, fsw= 100kHz/1.1
    • major errors = L is far too big for 100uF with a resonant frequency too low (1.3kHz)
      • that means the stored energy cannot be shut off fast enough even with infinite feedback bandwidth and there is sufficent time to burn out your power FET from flyback over-voltage and damage the PWM IC.
      • Missing phase Lead-Lag RC feedback filter is critical to BODE plot margin and overshoot from a step load. enter image description here recommendations
  • Consider subscribing to on semi or TI.com webbench and get fully designed and tested working designs and read all theory and test methods with integrated SMPS from all the major suppliers who have extensive free design tools.
  • Consider the datasheet design before you DIY your design without specs.

There are significant differences and reasons why you have not included.

enter image description here

https://www.onsemi.com/pdf/datasheet/tl494-d.pdf

  • search for some of my answers and recently on DVT, DFM which ties into design specs.
  • Good luck next try.
Tony Stewart EE75
  • 1
  • 3
  • 54
  • 182
  • By "phase-lead compensation in feedback" do you advice to put a lowpass filter on feedback line? So that the controller won't see high frequency changes? Switching to PFM will require a different IC I guess. Crowbar protection is simple to apply, I'll use it next time. I didn't care about OVP, OCP, OTP because I just wanted to understand the theory of boost converter. There are 3 pots because I wanted to precisely set the limits of voltage range, I didn't know the 10T pots resistance exactly that's why I've put 2 adjustable pots around. How did you justify 1.3kHz is too low, what should it be? – CaveScientist Dec 09 '22 at 11:34
  • Phase Lead compensation for Bode Plot is the effect of a "treble boost RC filter" not a LPF, rather a HF boost. All the other concerns are addressed in every detailed DCDC converter Design. Since this IC has DBC with a 5V reference you choose the output Vdc and scale down to 5V to regulate but boost "treble" for better feedback on ripple error and loop stability like a predictor-correction output voltage from the rate of change. – Tony Stewart EE75 Dec 09 '22 at 18:18
  • The self-resonant frequency of any choke or loop filter must be much greater than the servo loop bandwidth in order not to introduce a disturbance. Otherwise addressed by Root Locus and/or Bode Plot Phase Margin analysis methods. This is due to the phase shift that makes negative feedback possibly into positive feedback and thus unstable, especially when unloaded with high Q = resonant gain. You want any resonance to be very high f where the feedback gain is <1 for stability criteria. – Tony Stewart EE75 Dec 09 '22 at 18:21
  • Oh I see, boosting treble has a similar effect to what they call "Kalman filter". Wow this is going so much detailed. How am I gonna decide if I am going to use a PID, phase-lead or phase lag-lead...? Okay I can read real design papers for that. Also, you are afraid that resonant frequency will be around crossover frequency which will cause instability and you want me to push that resonant frequency to very high frequencies where the gain is really small? Do I get it right? BTW the converter works as I wrote in the edit, I only tested it a resistive load though. Ofc I want to learn the better. – CaveScientist Dec 10 '22 at 01:47
  • never use phase lag compensation. It is a common RC//R filter as shown in datasheet. Nothing fancy. R loads are trivial. Try a speaker or motor step load. poof. Normally for very high power L goes towards 1~10 uH and low C = 10 uF with very low ESR 10 mOhm with f = 0.1 to 1.5 MHz, but don't take my word, try the TI Webbench – Tony Stewart EE75 Dec 10 '22 at 01:55
  • I tried it with an 18V DC drill which draws1.6A at max but since I am testing with an 12V 2A adapter, I couldn't press the button fully, the max current I drew was probably 0.7A. The drill worked perfectly, output voltage dropped from 18.4 to 18.2 when I run the drill which is not a lot. I couldn't measure any AC voltage with multimeter and measured 6.5kHz %99.3-%99.9 duty signal at output when the drill is working. This meant to me that output had high frequency ripple with a really low magnitude when the drill is connected. – CaveScientist Dec 12 '22 at 18:57