• State Resolved
  • Locked Locked
  • Replies 14 replies
  • Subscribers 63 subscribers
  • Views 6739 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TL1451A Error Amp Input Voltage Rating

Bryce Salmi
Intellectual 310 points
Other Parts Discussed in Thread: TL1451, TL1451A, LM6144

The datasheet for the TL1451A leaves much to be desired in terms of specifications and ratings, I understand it's old. I have a design for a Maximum Power Point Tracker using the TL1451 as a PWM controller with external error amplifiers. I've seen a TERRIBLE yield rate where at least 1/6 are DOA when powered up for the first time. The PCB's were professionally pick & placed as well as reflowed by an experienced manufacturer for quick turn low volume manufacturing.

NOTE: This is an analog maximum power point tracker, the requirements were such that Perturb and observe with a microcontroller or ASIC was not an option. Please keep that in mind.


I may have found out why and would like some help to determine if this is in fact the issue.

TL1451A Rating

  • It appears that the TL1451 error amp input has an abs max of 20V input but it's not clear if this is common mode or differential
  • It's also not clear that the recommended operating range is 1.05V to 1.45V.
    • The recommended operating range is suspiciously +-200mV from 1.25V which is the internal reference voltage inside the TL1451A (see the block diagram). This makes me wonder if there are protection diodes or some sort of circuitry that will conduct/damage from overvoltage events much higher than the recommended operating range but lower than the 20V abs max.

    Maximum Power Point Tracker & TL1451A

Here is the design of the MPPT which utilizes the LM6144A as an external error amplifier. Now, the idea (and this works) is to tradeoff MPPT and output voltage regulation. The MPPT error amplifier obtains a reference voltage based on temperature that predicts the maximum power point voltage since the solar panel parameters are well understood. So:

  • When output power delivered to load is lower than input power from panel the MPPT does not perform power point tracking and the voltage regulation error amplifier has control.
  • When output power = input power (neglecting losses) the MPPT error amplifier obtains control (highest error voltage = lowest duty cycle).

The diodes (D8A and D8B) combine the signals and let the tradeoff happen. Verror is pumped directly into the TL1451A error amplifier setup as a unity gain buffer.

Possible Design Error

OK here is where I would absolutely LOVE some clarification. It appears that during startup I see the LM6144A MPPT error amplifier rail high. The LM6144A and the TL1451A are powered from the solar panel voltage so the op-amp voltage can never be higher than the TL1451A VCC which I thought was OK when designing this. However the LM6144A turns on around 1.8V and the TL1451A turns on around 2.7V which  at first was thought OK but this creates a race scenario.


The MPPT error amplifier detects the panel voltage is too high compared to the reference temperature based voltage (the measured temp signal will match the VCC voltage until it is high enough to properly operate). This causes the error amp to rail momentarily and it appears this dumps >10mA into the TL1451A non-inverting error amplifier pin 3 for several ms.

Below shows the D8A diode replaced with a 2.65K ohm resistor and D8B left unpopulated. This ensures that any current flow would be into the TL1451A from U3C above.

  • CH1 = Solar panel voltage (VCC)
  • CH2 = Pin 3 of TL1451 (Non-inverting error amp input of TL1451)
  • CH3 = Pin 8 of LM6144 (MPPT error amp output)

It's obvious there's a near 1.25V difference across the 2.65K resistor showing 471uA of current. Regulation is maintained around 1.8V after the initial spike.

Below is an attempt to show the op-amp inputs on the TL1451 and how they are getting pulled away from each other

  • Channel 2 below is Pins 4&5 of the TL1451A connected together (Feedback and inverting input)
  • Channel 3 below is Pin 3 of the TL1451 with 2K resistance
  • Math = CH2 - CH3

It's clear the TL1451A op-amp inputs are being pulled away from each other by 2V...

My fear is that this is stressing/breaking the TL1451A devices. I have replaced the broken ones and all seems fine but if this is what the issue is then I wonder if I have failures waiting to happen and i happened to find the ones that could resist this. Basically without current limiting resistors I'm hypothesising that I'm dumping several mA into pin 3 of the TL1451A using just diodes for feedback.

You can also see that the TL1451 error amp inputs are being pulled away from each other

Potential fix

Placement of 2.65K resistors in place of D8A and D8B seem to provide nominal operation and will limit the current to no more than 1.2mA during startup in a 6V system. below is a scope capture showing the MPPT error amp railing then the voltage regulation error amplifier taking over with two 2.65K resistors instead of diodes.

Big Questions I Would Love Answered/Help With

  • Could the stresses shown above with diodes D8A and D8B cause the TL1451A to intermittently fail from startup (unrecoverable) while some survive OK hundreds of startup cycles?
  • Would 2.49K resistors instead of diodes limiting current to about 1mA be sufficient to help ensure the TL1451A is NOT damaged?

Thank you

Bryce

  • 0
    TI__Guru**** 191445 points
    Yes you have picked a really old device. I will have to research to see if anyone can answer your questions TL1451 specific questions. You may also want to cross post in teh power, etc forum. That is where the MPPT topics are found. Maybe someone there has a similar application.
  • 0 in reply to JohnTucker
    Intellectual 310 points
    Thank you! I've cross-posted the topic in the etc..power forum.
  • 0
    Intellectual 310 points

    I've created a fault tree analysis which is attached below.

    Faults Identified

    • Ensuring operation of < 500KHz is good design but might not be causing the problem as reducing frequency of faulted channels doesn't seem to fix the problem.
      • If the timing resistor is ~4K it's almost like it's too low and a higher resistance causes less current. Increasing the timing capacitor decreases frequency but the triangle wave is still distorted slightly. Increasing resistance > 4K seems to clean up the signal while also decreasing frequency.
    • The LM6144 is highly suspect as I've recently identified that one of my freshly made PCB's had pin 13 of an LM6144 conducting to 700mV. Replacing it fixed the problem. I fail to see how it is being damaged as there is a ton of resistance limiting input current as shown in the op-amp error amplifier section & constant current RTD driver sections above.
      • If the TL1451 is sourcing too much current from the LM6144 during startup, maybe I'm breaching the 50mA supply current ABS MAX?

    Mitigations for Near Term Designs (Until Root Cause Found)

    • Replace error amplifier diodes with 2K resistors
      • Limits current to < 3.5mA in a 6.5V system
      • Allows proper operation of design as tested with PCB engineering board
      • Removes possible faults of LM6144 high current damage or TL1451 input current damage assuming < 5mA is OK rule of thumb.
    • Ensure operation below 500KHz
      • Less stress and good design. Tisk tisk I messed up. One less variable

    Doing this should at least prevent those issues (with the assumptions) from being possible causes on the next batch of PCB's. It will be interesting to see if there are more failures even if prior to first power on the current limiting resistors are placed.

    Help Still Needed

    Anyone knowledgeable about the internal design/requirements of the TL1451 or LM6144 weighing in on this would be amazing. I want to root cause this and stop having terrible yield (1/6 are dead on arrival from manufacturer).

    Fault Tree

    MPPT_Low_Yield_Rev_1_1_Fault_Tree.pdf

    Thank You,

    Bryce

  • 0 in reply to Bryce Salmi
    TI__Guru**** 191445 points
    I can make a couple comments here. They may not exactly answer your questions, but may help. Based on the specifications, it is likely that the error amplifier input is a PNP Darlington pair. For proper parametric operation, you want to maintain the pin 3 voltage to 1.05 to 1.40 V. That said, the abs max is 20 V before you could potentially break down the input structures which would be back biased with positive voltage on pin 3. That 20 V would probably be a short term stress. We believe that pin 3 should be able to handle up to around 5 V long term without damage. Input current should not exceed 0.5uA during normal operation. If you are able to source more than that into it, you have probably already damaged the device.
  • 0 in reply to JohnTucker
    Intellectual 310 points

    Thanks John!

    I'd like a clarification. I can limit current to about 1mA or so on turn on and the voltage on that pin during the turn-on transient is < 5V with 2K of resistance at all times. Can you clarify:

    • The 0.5uA into pin 3 requirement. Is that steady state or peak for "normal operation"?
    • If the voltage stays under 5V, even during turn-on when there appears to be about 1mA flowing, is that OK for long term damage?
    • Referring to "have probably already damaged the device" is that if I am constantly measuring >0.5uA into pin 3 even when voltage is nominal (about 1.05 to 1.45V)?

    Thank you a ton for the response.

  • 0 in reply to Bryce Salmi
    TI__Guru**** 191445 points
    I am really just speculating based on the specifications and the process capabilities at the time. As you probably know the error amplifier input should be high impedance. Ideally, the input current should be zero. Modern op amps have input current in the nano amp range or even lower. Since this is almost assuredly bipolar rather than CMOS or JFET input the input current can be higher. For most applications, the inverting input is connected to VREF / 2 so that is why the common mode spec is as stated (with feedback, the inverting input should be forced to track the non inverting input). let me see if I can look into this further. It may be a few days as I will be OOO for the rest of the week. I can't promise I will know the answer without physical failure analysis on a failing device (which may take several weeks time).
  • 0 in reply to JohnTucker
    Intellectual 310 points

    Thanks John,

    Here's an update on my end. I received some SOT23 2K resistors (Vishay MPM) and performed the following upgrades to an engineering board:

    * Replaced the diode array in the error feedback with 2K resistors, common pin 3 (SOT23)

    * TL1451 timing RC changed to 220pF and 4.99K = ~475KHz operation

    * Ensured 0.1uF capacitors across feedback resistors of LM6144 RTD driver + RTD scaling amplifier for stability.

    This board was the one I've used for a while. It works and I seem to have a nominal operations except for one MPPT (there are 6 per board). This is the interesting part. This particular TL1451 + LM6144 combo doesn't track MPPT or Vreg with an error voltage of 1.6V or so like all the others... it's actually about 2.5V. Other than that everything is nominal. I wonder if that's a sign of something damaged since these MPPT's on this board were used with the diodes that did not limit current.

    I have one more board that should be shipping to me soon which has never been turned on so switching out it's diodes for resistors should give a good idea if these are getting damaged with the diodes and unlimited current.

  • 0 in reply to Bryce Salmi
    TI__Guru**** 191445 points
    You could always replace the suspect ICs as well.
  • 0 in reply to Bryce Salmi
    Intellectual 310 points

    TL1451_Startup_Scope_Captures.pdf

    OK here's a testing update!

    I've added the 2K resistors in place of the two diodes in series with the LM6144 error amplifiers. 

    I've also reduced the switching frequency to 475KHz

    Attached is A PDF showing scope captures of the output of the error amplifiers U3B and U3C along with either pin 3 of the TL1451 or pins 4/5 of the TL1451 and the input voltage (solar simulator attached). I switched between having the 2K resistor pair connected normally into pin 3 of the TL1341 using 4/5 and a unity gain buffer and bypassing pin 3 and pumping the resistors directly into pins 4/5 of the TL1451 bypassing the unity buffer.

    Overall, when I test startup the configuration attached to pin 3 of the TL1451 seems to have issues with high load cases. This is when I pull enough current to pull the output to 3.3V (nominal low voltage) and the MPPT is tracking the input voltage (error amp is in control). As shown in the PDF, these cases tend to fail because the converter doesn't regulate and the output gets stuck to a <3.3V voltage and the input is some lower than MPP voltage. This is also tell-tailed by not having that signature "over voltage" during startup in those scope captures.

    The device seems to work fine when booting up into a high load when I bypass the TL1451 error amplifier and input directly onto pins 4/5 of the TL1451.

  • 0 in reply to Bryce Salmi
    Intellectual 310 points

    More updates

    • I've tested the load cases with respect to the switching node to show what the switching output is doing
    • I've scoped the LM6144 MPPT error amplifier inputs during a nominal start and a high load no-start
    • This is of interest because it appears it's the LM6144 might actually be the culprit of the no start cases when the error signal is pushed into the TL1451 Unity gain buffer

    TL1451_Switching_Startup_Scope_Captures.pdf

    TL1451_pin 3_LM6144_Inputs_MPPTERROR.pdf

    Switching nodes Captures:

    I only bothered to capture a few good cases and some of the bad startup cases. It's apparent that the TL1451 is switching the whole time and during a nominal startup the "overshoot" on the error signal actually does cause a 0% duty cycle as it should (higher Verror = lower duty cycle). During a no-startup case the input voltage never rises high enough to fully turn on (when Vout <3.3V and the input voltage is about 2V... it's NOT tracking MPPV as it should be

    I forced the input voltage to rise by bypassing the solar simulator I have and using a voltage regulated supply, this causes the circuit to turn on OK... however it never reaches MPPT because at this point there's too much input power to cause the input voltage to sag to the maximum power point so I stay in Voltage Regulation mode.

    So this could be caused by the solar simulator I have hooked up. HOwever, that doesn't explain why simply directly injecting the error signal into pins 4/5 of the TL1451 causes the circuit to startup fine in the high load case.


    Disregard images 9 and 10 which show the output of my Ideal diode. I suspected maybe it was oscillating or pulling current in the off nominal case. It actually is when the no-start happens but that's actually expected as the voltage on the output is below 3.3V and therefor not valid, the device will do that when that occurs so it is actually operating fine, there is no explanation for why the output voltage is pulled much lower.

    LM6144 Startup captures:

    This is interesting. As shown in the attached LM6144 PDF captures, during a nominal startup the inputs are slightly pulled away from each other but nominal Maximum power point tracking results in them becoming equal again. 

    however

    During a failed startup into a high load with the Tl1451 pin 3 being sent the error signal it's obvious that the LM6144 inputs pins 9 and 10 are pulled apart and never go back. This is strange. I measured the voltages and pin 9 (shown as low in the scope capture) is stuck around 750mV to 800mV. Maybe this is coincidence but that almost looks like a diode drop....

    Lastly I am using a home-built constant current source for my testing. It could conceivably cause an issue during startup. However, I'm not convinced of this because the effect is mitigated by simply inputing the error signal into the TL1451 pins 4/5 instead of pin 3 which leaves the constant current load in place and unchanged and therefore likely not causing the problem.

  • 0 in reply to Bryce Salmi
    Intellectual 310 points

    BIG UPDATE. I found out WHY THE LM6144 IS RAILING HIGH DURING TURN ON AND STOPPED IT. I'm not sure if this means I don't need to limit current with resistors into PIN 3 of the TL1451, we'll address that shortly.

    The LM6144 seems to be turning on it's output stage before the input stage is properly powered and biased. I suspect this causes the output to rail in the given startup conditions until the input stage biases itself.

    I switched out R125 with a 205K resistor and C72 with 0.2uF of capacitance. This created a RC filter of about 3.88Hz. This slows down the ability of the non-inverting input voltage to charge up faster than the biasing circuit of the op amp (internal). Essentially forcing the output to remain low during startup.

    LM6144 Error Amplifier Pin 10 No RC Filter/ Too high RC Constant Filter Turn ON

    Channel 1=1 Vin
    Channel 2=TL1451 pin3
    Channel 4= TL1451 pins 4 and 5

    This is circuit was setup for -60C panel temperature operation so voltages are lower but you get the point.

    As you can see, the LM6144 outputs a voltage equal to the panel voltage (VIN) during startup which drives the TL1451 pin 3 up to 4V apart from pins 4 and 5 of the TL1451.

    LM6144 Error Amplifier Pin 10 RC Filter Turn ON

    Adding a 3.88Hz RC Filter to Pin 10 of the LM6144 Error amplifier circuit

    Scope capture below:

    Channel 1=1 Vin
    Channel 2=TL1451 pin3
    Channel 3= LM6144 pin 8
    Channel 4 =  LM6144 pin 7

    This is my MPPT in configuration for a 0 Celsius panel temperature when starting up into a load sufficient to drop the output voltage to 3.3V (my lowest design output voltage). This is essentially charging a completely discharged battery on the satellite. As you can se there is no voltage spike and I've considerably slowed down the system (turn on now takes about 150ms.. not ideal but I'd rather that than dead circuits). It's hard to see but under channel 3 (LM6144 pin 8, the offending overvoltage output driver) channel 2 is directly matching channel 3. This voltage peaks around 1.76V... there is no turn on transient!!!!

    No Load Startup W/ RC Filter

    When no load is presented to the MPPT output (not really a design case but more of a possible case during testing) the overvoltage still occurs. Not sure if this will damage it but it's a start.

    Channel 1=1 Vin
    Channel 2=TL1451 pin3
    Channel 4= TL1451 pins 4 and 5

    Remaining Questions

    So givent that in all cases except no load turnon, it looks like I can use JUST diodes on the output of the LM6144 error amps to drive the TL1451. This is idea for operating modes. I've noticed with 3K resistance driving pin 3 of the TL1451 some heavy load startup conditions may not work correctly. I'll test some more but:

      Should the last image there, showing the no load case, damage the TL1451?

    This is significant progress to at least root causing the turn on transient!

  • 0 in reply to Bryce Salmi
    Intellectual 310 points

    Update!

    I performed many tests over the weekend to nail down what I wanted to do on the flight board, I had one shot to turn power on for the first time and hopefully mitigate any damage. Some notable tests and their results:

    • Directly inject error signal into TL1451 pin 5, cutting pins 3 and 4: Not Good
      • Likely because Tl1451 error amp inputs left floating, hoped I'd still overdrive pin 5
    • Connect TL1451 pins 3, 4, and 5 and just drive those, Not Good
      • Sort of a hail mary pass, maybe would have worked if not for 2K series resistance from real error amplifiers
    • No LM6144 non-inverting error amplifier (MPPT error ampZ) RC filter + 1K resistors instead of 2K:SAME
      • Hoping it would help with startup conditions being more stable but did not

    Flight Revision Design

    Let me start by saying in October I published this design in the AMSAT Symposium proceedings which gave it an ISBN number and technically made it public domain knowledge. Yay! Also this is for a cubesat which technically is no longer ITAR anyways, fixed from both ends... I'd call that redundant :P I'm very happy to show people the design methodology and thought process around this level of troubleshooting/analog design.

    I had to pick one design and go with it for flight (this will fly aboard a SpaceX Falcon 9 as a secondary cubesat :). So my thought process included:

    1. I really don't want to make more than a BOM change
    2. Flying components point-to-point soldered/cut traces/30 gauge wire routing fixes is not good
      1. High vibration and thermal stress environments make those changes extremely dangerous to a spacecraft
    3. I can't fully root cause but I can limit the potential causes by better design
      1. i.e. limiting potential inrush current cannot hurt and makes the fault-space smaller
    4. Time

    That said I decided to go with the following changes to the Maximum Power Point Tracker

    • Change Tl1451 switching frequency to be < 500KHz (bonehead mistake on my part.. live and learn)
    • Add feedback caps for RTD (thermal) sensor circuitry to reduce oscillations (op-amp GBW is too fast, gotta slow down!)
    • Use BOM replacement for SDM40E20LC diode array (MPMT4001AT1) providing 2K resistance from each LM6144 error amplifier
      • Limits maximum current ever (per op-amp) to 6.5V/2000 Ohms = 3.25mA which is < 5mA and intrinsically safer
    • Remove 10K pulldown resistors for error signal node since diodes are no longer used
    • To ensure startup high load operation is assumed to have a battery on the output
      • Some instability in startup into an active load (constant current circuit) was observed when full power was pulled from the MPPT and the circuit was started up from completely off into the load case.
        • Could be the active load causing this issue, not sure
        • Either way, the rest of the satellite is not designed for  no batter operation and it's safe to assume the battery is present for the main mission. Anything else is extra.

    Results

    Engineering Unit Update

    Initial tests on an engineering board with these modifications (not headed to space) proved very encouraging.

    • Only a BOM change, except for two stacked feedback capacitors in RTD circuitry
    • Inrush event mitigated during startup with RC filter
      • Used 1uF X5R capacitor (1uF 0402.. yup) so not sure over temperature that this is always the case but...
    • Even if inrush happens, only one small corner case actually allows BOTH LM6144 op-amps to drive HIGH voltage thus pulling TL1451 pin 3 to VIN
      • Panels must be at 60C or so AND the battery must be fully charged i.e. startup into voltage regulation mode instead of MPPT mode (low load)
      • Even then, Vin is 4.9V so less than 5V and abs max on the pin is 20V AND maximum current is 6.5mA (3.25mA X 2) so not super ideal but also better than before

    Below is a zip file of all my tests to validate the engineering board after the modifications were made:

    SN4_Post_FlightLike_Rework_Tests.zip

    one of the test cases should be highlighted. The zip file contains all six MPPT's on the PCB, here's the +X MPPT in Voltage Regulation mode and MPPT mode during startup:

    Channel 2: TL1451 Pin 3

    Channel 4: VIN

    Channel +X Voltage Regulation Mode Startup

    Notice how the TL1451 pin 3 shows a peak of 2.4V? That's because LM6144 pin 7 is railing low and pin 8 is railing high, the 2K resistors are performing a voltage division of 2 (see Vin is about 5V during that peak). There is only about 100uA flowing into the TL1451 pin 3 circuit (validated from other tests). yes this means pin 8 of the LM6144 likely railed high but it did so very briefly and at some point I've limited current anyways and should move on.

    Channel +X MPPT Mode Startup

    MPPT startup shows the RC filter slowing everything down a tad but the inrush events are mostly mitigated which helps. I'm happy to see that in both modes startup seems to not stress either the LM6144 or TL1451!

    As you can see the startup is reasonably fast.. around 100ms  give or take for the MPPT mode to fully get to the regulation voltage. Not super ideal but also not a deal breaker.

    Flight PCB First Turn On Results With Modifications

    Here's the main story. I modified the one board I had left which had never been turned on. This gives a pristine view of the board with the fixes made. All MPPT's booted up fine with 0 failures!!!. Now I can't say I 100% solved the problem, that would require many more PCB's to be made to gain confidence in the changes. Additionally theres the chance that I was just being stupid during bringup testing. Everyone makes mistakes. I became incredibly methodical when performing these last tests on the engineering board and especially the bringup of the flight board. Everything I did had a  OneNote with a checkbox and validation. I cannot stress how it was so easy to think something was wrong only to find out my test setup was slightly incorrect as it involves several pieces of hand-tuned equipment for each "state" test. Lessons learned from this entire experience:

    1. Be methodical, it is WORTHthe extra time to document tests and bringup notes very carefully.
      1. I bet I repeated the same tests several times in some cases because I didn't fully document it's results prior.. AKA "I see what it's doing, move on" is not a good answer.
    2. On these older chips (TL1451) it's best to limit any potential current into the op-amp inputs
      1. Unfamiliarity with using very old technology op-amps lead to this potential issue
    3. Schedules can be pushed in most cases
      1. One reason I used 700 KHz instead of the nominal 500KHz was that I saw it was "working" and had other issues to deal with so I moved on. That was nearly 5 months ago and I thought that the schedule required me to get hardware ASAP. As it turns out I'm here in November working on this and realizing that the project could have easily waited for me to be slow and methodical, unrealistic deadlines can lead to more delays if not treated as they are... unrealistic!

    I know this post ended up being a lot of me showing my tests and design choices during mitigation. Overall, it's probably a great post to see the troubleshooting and fault mitigation techniques used in these analog type circuits. There's even an aerospace flavor to this which is probably not common. If this way my day job I'd have found these problems months ago and spun a new board, this is however a project that I volunteer on (no pay!) and perform in my free time so some choices spin from that realm.

    I super appreciate the help of the several TI employees who did chime in and give me some info on the TL1451 and LM6144. Even outside of TI e2e some field reps at TI did some footwork to find people who knew about the IC's I used, big thanks for the help too!

  • 0 in reply to Bryce Salmi
    TI__Guru**** 191445 points

    I'm glad everything is working out.  Thanks for the detailed posting of your progress.

  • 0 in reply to JohnTucker
    TI__Guru* 92935 points

    Thanks for the update Bryce and thanks as always for helping out John!