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.

LM74722-Q1: More precisions about the PD discharge switch characteristics and timing

Part Number: LM74722-Q1
Other Parts Discussed in Thread: LM74720-Q1

Hi,

We have an application that need to isolate a circuit with motors from the power supply of a whole system consisting of many electronic boards. The power supply of the system is 36V.

The LM74722-Q1 seemed to be the proper device to act as an ideal diode and to be able to protect the power supply and the other circuit from the back electromotive force when the motors decelerate.

The output of the supply going to the motors has a large capacitance (2200uF) and the individual motor drive boards have themselves large capacitors.

We have a circuit to choke the motors back electromotive force re-injected in the supply.

We had a problem just powering the whole system and disabling the enable pin of the LM74722. It blew and punctured the device in the ground pin corner.

After reviewing the device datasheet, I saw that the revision history of this still pre-production part has been changed to update the "Load Disconnect Switch Control (PD)" section.

In this section it is mentioned that for system designs at 48V and above, a series resistor must be inserted between the output MosFET gate and the PD pin to protect the internal PD discharge switch. We did not had this resistor in our circuit, only a 15V zener to protect the gate/source of the load disconnect MosFET.

I suspect that systems with large output capacitance also can affect the internal discharge switch because it might not be discharged completely during the time the discharge switch is active.

I reviewed the LM74722 datasheet and the I(PD_SINK) condition are defined in the device characteristics table but only for overvoltage while the "Load Disconnect Switch Control (PD)" section clearly state that they are also when the enable is low (and erroneously when Vs drops below V(VS POR)). There is no VS pin on the LM74722, but there is one on the LM74720. It seems also that there are some current values inherited from the LM74720-Q1 datasheet because the figure 8-2 in the "Load Disconnect Switch Control (PD)" section use the typical LM74720 I(PD_SINK) currents.

I reviewed the LM74720-Q1 datasheet and it seems that there was an attempt to better describe the behavior of the PD discharge switch. The "tPD_Pk" parameter tries to explain the conditions and duration in which the discharge switch is active.

In my humble point of view, even the LM74720 datasheet is not clear enough about the behavior and timing for the PD pin. A timing diagram with ALL the conditions should be added.

The timing characteristic for the PD pin is still missing in the LM74722 datasheet as this device is in a pre-production state.

Is it possible to get more information about the behavior of the PD pin for the LM74722-Q1? Clear conditions and detailed timing would be great.

Also, pass the comment that the datasheets should be updated. If we encountered problems with this interesting and new device, others will too.

Any other information, like the SOA of the discharge switch could be useful.

For now, we will add the recommended series resistor with the PD pin but we are unsure if it is a definitive solution and if the large output capacitance of our circuit could still cause the discharge switch to blow.

Best regards,

Eric

  • Hi Eric,

    Welcome to e2e! 

    Yes, your suspension of the internal PD pull down switch getting damaged during FET turn OFF could be correct.  When the Enable is pulled low, the output capacitance will discharge through the gate-source zener diode and the internal PD pull down switch (if not discharged by the load). Higher output cap could put power dissipation stress on the internal FET and it can get damaged if the stress is beyond the Safe Operating Area.

    The behavior and timing for the PD pin will be similar for both LM74720-Q1 and LM74722-Q1. When the PD is pulled low, there would be a peak pull down current of  IPD_SINK (Peak pulldown current) for tPD_Pk duration. After tPD_Pk duration, the pulldown strength will be equal to IPD_SINK (DC pulldown current) as long as the PD pin is supposed to be held low. 

    I will try to get the information regarding SOA of internal pull down switch and will also pass on the information to update the datasheet to make it more clear on the behavior and timing for the PD pin. Meanwhile, Can you let us know what is the total capacitance at the output of LM74722-Q1 (including at the input of motor drive boards) ?

  • Hi Praveen,

    First, thanks for your fast response.

    The behavior of the PD pin is clearer now. The fact that the quiescent current is very low gave me the impression that the PD pull-down current of 12mA for the LM74722-Q1 was not continuous (if applicable). This 432mw (36V * 12mA) is probably not sufficient to burn the PD internal discharge switch but it still can contribute if it has been pushed to its limit before when using the 100mA pull-down current (3.6W for a pull-down current of 100mA).

    To answer your question about total capacitance, actually I see around 3200uF of total electrolytic capacitance at the output of the LM74722, including all small motors drive circuits. There are also ceramic capacitors but their total value should be derated because of the DC bias applied. I guess that it is about a supplemental 200uF of effective ceramic capacitance. I think that 3400uF should be a good guess.

    Knowing the behavior of the PD pin, I will be able to simulate it with a simple circuit and see what is the power dissipated by this switch. I will also be able to see the effect of the series resistor.

    Thanks.

    Best regards,

    Eric

  • Hi Eric,

    Do let us know your observations after adding a series resistor. Meanwhile, I will check if there is a problem in discharging 3.4mF cap through the PD pin. 

  • Hi Eric,

    We checked with design team and as per our understanding, discharging a 3.4mF capacitor charged at 36V through the PD internal discharge FET should not be a problem.

    Can you  share your schematics with us. If it is confidential to share in a public forum, you can use the private chat to share the schematics. 

  • Hi praveen,

    Information requested sent through private messaging. Supplemental tests and whole circuit informations added.

    Eric

  • Hi Praveen,

    The PD series resistor will be added on the next PCB. I was not sure at the beginning but now I consider this is the definitive fix to avoid burning the internal PD discharge switch.

    I however do not understand why the design team can clearly state that there should be no problem to discharge a 3.4mF capacitor charged at 36V (I assume without a PD series resistor) and that for a circuit at 48V (no output capacitor value specified in datasheet), a PD series resistor should be added. The 100mA discharge period (up to about 65us if LM74722 tPD_PK is similar to the LM74720 one) is creating about a 3.6W instantaneous power dissipation.

    Understand me well, I do not say that the design team answer is wrong. The design team may have done an analysis with the SOA of the discharge switch and it may be able to handle it for 36V and not for 48V. It probably also depends on the internal junction temperature at the moment the internal PD discharge switch is activated. From an external perspective without all the relevant informations, the difference between 36V and 48V cannot be justified instinctively.

    Examples of use of this device are automotive ones (low input/output voltage) but it seems that a couple of customers are using the device with higher voltage (device supports up to 65V). I think that it would definitively be a good idea to add supplemental mentions in the datasheet about the different cases that requires a PD series resistor, the zener diode, the effect in inrush current limiting, the effect of higher output voltage, the possible effect of large output capacitance, etc. You already passed some of this information to the clarified to the proper group. It would probably be good to update all devices datasheet in that series about that (LM7472x).

    There is probably no clear demarcation line between the output voltage that requires a series resistor and a one that don't. It probably depend on the output capacitance, the junction temperature inside the LM74722, the PCB heat dissipation and many other things I am not aware of. I hope that the PD series resistor use can be summarized or fully characterized in a future datasheet version.

    For people reading this thread, just changing the zener to protect the output MosFET from a one with a low power rating to a one with a high power rating changed the things and the LM74722 internal switch blew after this change (Zener forward biased equivalent resistance lower? Missing zener patch on the board with different zeners had an impact?).

    We currently have a couple of different PCBs to be done. It will take some time to get the final one with the PD pin series resistor.

    Just to complicate things, testing with the series resistor added on the current board is not simple because of the whole circuit needed to test the ideal diode circuit that blew the LM74722. Tests without the PD series resistor and with the "problematic" zener and with only the ideal diode circuit did not permitted to reproduce the case where the LM74722 blew. I will need a complex setup involving many boards, multiple software versions, etc to reproduce the problem. I grabbed oscilloscope waveforms but everything seemed normal. The simple and smallest hardware case unfortunately did not reproduced the problem.

    I will wait for further analysis of the design team on the requested schematics sent this morning but unless there is a clear evidence of a problem detected or a different point of view of the design team on my questions on the 36V to 48V PD series resistor requirement, I consider that there are too much things that can affect this requirement and that even if we not fit actually in the datasheet described case, adding the resistor will solve the problem.

    Regarding this, I leave the case opened but will close it when receiving advice from your side after the design team review of our schematics.

    Best regards,

    Eric

  • Hi Eric,

    Thanks for the detailed response. Can you please confirm if my understanding on the sequence of events is correct ?

    1. The Ideal Diode controller initially damaged because there is no zener across Gate-Source of the Mosfet.
      1. We will make sure the importance of zener is described properly in the datasheet.
    2. The Ideal Diode controller failed during initial power up (where the enable signal was high for few ms) with BZD27B15P-M Zener across FET Gate-Source
    3. The Ideal Diode controller passed the initial power up with MMBZ5245BLT1G zener across the FET Gate-Source but later failed *at a particular condition*  when plugged into the complete system (including the motors).
      1. The issue was not repeatable when testing with LM74722-Q1 circuit standalone ? (waveforms were captured in this case)
      2.  Can you explain the *particular condition*  in which the IC failed in more detail. Here I am trying to figure out if there is any other issue that we are missing. 

    Regarding your comments on the internal PD switch SOA dependence on the IC junction temperature (layout, RQJA, Ambient  temp., cooling, etc..) are completely true. When you test the entire system, what is the maximum junction temperature you estimate the IC to reach ? 

  • Hi Praveen,

    Here are the answers/clarifications to your questions/understandings :

    1) a)

    Not exactly.

    Found before receiving the PCB that a zener diode was missing to protect gate-source of output MosFET. During first ideal diode controller board bring up, tested at 12V without zener and it worked properly. Before testing at 36V, our technician patched the circuit to add a MMBZ5245BLT1G across gate-source to protect MosFET. We knew it would damage the MosFET gate if not added. He gave the board patched with the zener to the person developing the whole product that tested it on a bench power supply at 32V (30V/5A power supply + 2V over nominal voltage adjustment possible). Only 2200uF output capacitance was present. Ideal diode controller worked properly.

    2)

    Five supplemental ideal diode controller boards were patched with BZD27B15P-M zener diode instead of the MMBZ5245BLT1G because the BZD27B15P-M was already used somewhere else in the product.

    The person that developed the whole product assembled an ideal diode controller board with a BZD27B15P-M, small motor drives, a 600W peak 36V power supply in the whole product to test all the material together. This raised the total output capacitance at the output of the LM74722 from 2200uF to about 3400uF.

    He tested a first product and something burned as soon as the power supply was turned on. Since all the other boards had been tested before together on the 600W power supply without the ideal diode controller without problems and because a patch had been made on the ideal diode controller board to solder an important missing component (the zener), he tried a second product with all these boards and something also burned as soon as the power supply was turned on.

    You understood correctly that unfortunately, the LM74722 were enabled for few ms only after power up then disabled. This is due to a change in polarity of the LM74722 enable pin. That will be corrected when all the PCBs of the whole product will be received.

    The technician investigated and the 2 boards had the LM74722 blown and punctured in the device corner near the ground (can barely be seen by eye but visible easily at microscope).

    3)

    Not exactly.

    Under the request of the person that developed the product, investigations were done to repair the things burned on the 2 products. The LM74722 were replaced. The zener diode was changed for the MMBZ5245BLT1G initially used on the first board. On two control boards, microcontrollers that generated the enable signal for the LM74722 were replaced because it seemed they were at least partially damaged (heating more than normal). The two same whole products were re-assembled with the repaired ideal diode controller with the MMBZ5245BLT1G and were re-tested. There was no problem at power up and the product comprised of many PCBs has been put in test for many hours without failing.

    The 3 remaining patched ideal diode controller board were re-patched to use the MMBZ5245BLT1G instead of the BZD27B15P-M. For now, 6 "prototypes" of the whole product were tested and some of them survived a couple of tenth of hours in test (motor drives working) without any problem.

    a)

    Exactly. First test done with the MMBZ5245BLT1G on another standalone ideal diode controller board.

    Then, for second test BZD27B15P-M was re-used on on the ideal diode controller board to try to make the LM74722 to burn again as when it was used in the whole product

    After seeing the revision history of the LM74722 datasheet and re-reading the datasheet pointed, my assumption was the PD discharge switch burned because of the relatively high output voltage (36V), because a zener is present (must be for all output voltages over 18V to 20V, zener is forward biased during PD discharge applying almost Vout to PD discharge switch) and because we had a much larger output capacitance than the one in the examples shown in the datasheet and the LM74722 EVM. My assumption was that it was the disabling of the LM74722 that cause it to blow.

    Was not able to reproduce either with the BZD27B15P-M on a standalone ideal diode controller board.

    b)

    The "special conditions" appeared in step 2 in the whole product with a more powerful power supply, really at 36V instead of 32V, with a larger output capacitance. After changing zener back to MMBZ5245BLT1G and repairing board damaged in step 2, no LM74722 were burned again.

    Regarding your question of junction temperature :

    After the inrush current limiting, the output MosFET is always full on (Vgs limited to 15V by the zener so Rds ON is almost as the smallest possible). Also, the ideal diode MosFET at the input is also full on. This part of the circuit is not heating a lot. We are not so concerned by the power that would be lost in a Schottky diode if we did not use the ideal diode controller. Our concern is about the heat dissipation in a closed enclosure at high temperature (industrial application).

    Also remember that the 2 boards were only briefly enabled before being disabled due to the wrong polarity of the enable (un-synchronized different sub-assemblies PCB versions for the whole product). The ideal diode controller board MosFETs did not had time to heat as the system was not completely started (no motor drives had its motor spinning).

    The only heating I can think of is the one of the output MosFET during the inrush current limiting.

    I have for now really no idea of the junction temperature. Ambient temperature in the product will have to be measured and reported outside. For now, the base of our product is only warm at touch when the motors have been spinning for hours (whole product is at ambient temperature of about 20 degrees Celcius).

    Hope it is clearer now.

    I wish I could have reproduced the problem on the ideal diode controller only. It would have been simpler to investigate.

    As I said before, I do believe that we need a series resistor with the PD pin in our application. The schematic for the next PCB version of the ideal diode controller sent through private messaging is having it. Oscilloscope measurements will be made on this resistor when we will receive the next ideal diode controller PCB.

    If you need more information to clarify the exact conditions (as much as we know them) in which the LM74722 burned, let me know.

    Best regards Praveen,

    Eric

  • Hi Eric,

    Thanks for the detailed response. Let me get back to you within a couple of days.  

  • Hi Praveen,

    Ok. I will wait for the answers or the new questions of your design team, if any.

    In the meantime I wanted to reproduce the case where we burn the LM74722 with only the diode controller so I did more tests. I had interesting conclusions.

    I took the circuit alone still with the BZD27B15P-M (old circuit, no series resistor added) and tried it on a bench power supply at 36V with 3A of current (over-current limit set to 3.3A). Previous tests were done at 32V because I thought that my bench power supply was limited to 30V (+2V of over-voltage). When I did the test I was not able to turn on the circuit properly because the power led on the ideal diode controller did not turn on. Last time the same circuit seemed to work properly.

    I asked our technician to check if there was damage visible on the device under microscope and if not to change only the LM74722 on our ideal diode circuit. He did not saw anything so he changed only the LM74722.

    I then did tests again with the bench power supply at 36V but with a current of 100mA instead. By shorting the enable pin to the VIN (i.e. V anode, enable pin pulled-down by 100K) I was able to turn on the circuit many times properly. I used a small probing clip to short the input voltage to the enable pin with wire wrap probing wires soldered on these signals. I was grounded with a grounding strap and working on a grounding mat.

    I decided to change the current on the bench power supply to 3A instead of 100mA. I did this with the ideal diode circuit and the power supply output not enabled. I turned on the power supply then I clipped the enable pin on the VIN (36V input voltage) and the output power led of the ideal diode controller turned on briefly then the led stayed off. I was not able to turn it on again. There is no voltage at the output of the circuit. I think that I probably burned again the PD discharge switch with glitchs on the enable pin when doing my tests.

    It then seems that the 4V difference between my previous tests and this one (36V vs 32V) was sufficient to break something inside the LM74722. In this case, there is no visible damage on the LM74722. Note that the 600W peak 36V power supply in our product and the larger output capacitance are probably just making things worse than my bench power supply.

    It do not change our conclusion about the necessity of the PD series resistor in our circuit. I only wanted to make you know that we were able to reproduce a similar issue.

    We do not have the LM74720/LM74722 EVM but on your side this could be a good starting point by fitting different components to try to reproduce our problem. Work in this direction might already be started.

    Hope this help you.

    Best regards,

    Eric

  • Hi Eric,

    Good to know that you are able to reproduce the issue on standalone LM74722-Q1 circuit.

    From your description, I can understand that the device fails when the Enable pin is connected to input probably because of glitches on EN (ON and OFF the FETs),  36V input and the output cap. 

    • I can verify the same thing on EVM from my side. Were you using 3400uF while testing standalone circuit ?

    But I am unable to understand why the device does not damage with Input power supply current limit set to 100mA. 

    Does LM74722-Q1 survive with  TVS changed to MMBZ5245BLT1G in standalone testing ?

  • Hi Praveen,

    Yes, glitchs occurs when connecting the clip to set the enable pin to 36V. I captured oscilloscope waveforms when I did the test at 32V a while ago and I was able to see the glitchs. If the discharge time at 100mA for the LM74722 is about maximum 65us like for the LM74720, this is way below the glitchs time due to a human connecting the enable with a clip. In the real circuit, there is a microcontroller enabling the ideal diode circuit.

    There is only the 2200uF capacitor on the ideal diode controller board. The supplemental 1200uF to get to 3400uF are on the drives and the other boards in the whole product.

    I am unable to understand too how the current limiting of the bench power supply change the things. I can only see that the ideal diode controller board is then in a situation much similar as in the whole product where the current is not limited by the 600W power supply.

    There are other things that influence the fact that the PD discharge switch burn or not at 36V in our product.

    How can we explain that the MMBZ5245BLT1G do not burn the switch and that the BZD27B15P-M do? By the equivalent zener resistance being greater for the MMBZ5245BLT1G than for the BZD27B15P-M so it protect a bit the discharge switch (MMBZ5245BLT1G is a 225mW zener while the BZD27B15P-M is a 800mW one)?

    The patch to add a zener diode across the gate-source of the output MosFET on a PCB that had no footprint for it probably also influence a bit the behavior of the circuit when it is disabled.

    May be if continuous enable and disable tests were done on one of our ideal diode controller board with a MMBZ5245BLT1G zener, we would reproduce the PD discharge switch burned.

    The 7th ideal controller board on which I did my test has not been repaired yet. But there are 6 other ideal controller boards that are still working and they have all a patch with the MMBZ5245BLT1G zener. No one failed since we got back to this diode after the 2 boards had a burned and punctured LM74722 when tested in the whole product with all its boards.

    Our technician that assemble boards and do the patchs on PCBs is busy right now with our others products regular production. I will see if he has some other ideal diode controller boards patched with the MMBZ5245BLT1G zener or if he can repair the damage board and change the zener for an MMBZ5245BLT1G.

    I did simulation to see how react a 3400uF initially charged at 36V and discharged during 65us at 100mA and the voltage do not vary much. Even with the constant 12mA discharge after this 65us period (estimated from the LM74720 numbers), the simulation shows a very slow decay (probably up to seconds, I did not simulate it to this time).

    I changed the capacitor in the simulation to a 47uF like on some of the LM74722 circuit examples and it would take about 17ms to discharge it if the 100mA rate was constant and not only 65us. After 65us, there is almost still 36V on the capacitor. I am then not sure that the value of the output capacitor change something in the problem we discovered.

    Hope that tests on the LM74722 EVM will show something on your side.

    Best regards,

    Eric

     

  • Hi Eric,

    One possible condition which I think could be causing damage to the LM74722-Q1 is glitches on EN/UVLO. If the glitches are triggering the Gate pull down circuit multiple times, it can so happen that the 100mA pull down has been activated multiple times causing SOA stress on the internal pull-down FET, causing it to get damaged. 

  • Hi Praveen,

    I agree that my test connecting manually the enable pin to the input voltage can trigger the discharge many times. My tests to reproduce the issue may be invalid then.

    You got a good point that should be mentioned in the datasheet.

    Unfortunately, in our product the enable on the LM74722 was set at power up (due to un-synchronized versions of PCBs for all the different boards) then disabled few milliseconds after. There was a single disable and the LM74722 blew anyway.

    I will see if I eventually can manage to have any of our microcontroller board to use one of its output to control cleanly the enable for a new test with the ideal diode controller.

    Best regards,

    Eric

  • Hi Eric,

    Please update us with your test results after you are able to drive EN/UVLO with a clean signal. Meanwhile, I will try to see if a 3400uF cap at 36V can damage the device on EVM.