• State TI Thinks Resolved
  • Locked Locked
  • Replies 2 replies
  • Answers 1 answer
  • Subscribers 62 subscribers
  • Views 269 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.

WEBENCH® Tools/UC3825A: uc3825A

rich walters
Prodigy 90 points
Part Number: UC3825A
Other Parts Discussed in Thread: UC3710, UC3845, UC3825

Tool/software: WEBENCH® Design Tools

Got a tricky one here. I am using the UC3825A/B in a high powered Buck 2kW running at 80Khz @50% td with Vin 400-450VDC. The inductor is a large KoolMu Core, L =400uH deliberately with a low I ripple % exhibiting a small inductance swing under 13.5A load 150V with an ample power margin. The efficiency is around 98-99% at this power level so there are no instability nor deep switching related losses. However anyone using a high side driver with a gate transformer with high voltages with power will be too aware:- Read on.

The problem is start up with already 400V appearing on the input bus. A current sense transformer in conventional mode in the drain path switching mos. This technique isn´t ideal for start up for sensing narrow pulse widths. Perhaps a dual circuit compromising a flyback current sense with different core low perm material would be better at increased circuit complexity. The start up with a full Bmax is a very nasty mode and not surprisingly lost some mosfets rated well above circuit currents. The dv/dt plus needle Miller effect overstrains the switching mos plus driver UC3710 into auto shutdown making this chip a bad choice for operational reliability. Another type has been sourced.

Question:- As I understand from the application note U-128 and the data sheet SLUS334E, the LEB programming would blank out the minimum pulse width on start up to the deadband width created by the LEB capacitor values. If I used the conventional ramp with capacitor filter i.e on pin 7 ramp would be more soft start feasible and detectable than the soft start being blanked out to the minimum duty cycle pulse dictated then set by the LEB. Using the LEB would imply a snap-soft start at a pulse width that would be on start up destructive. Am I seeing this right ?? I see a conflict between actual soft start performance, the LEB and peak I lim. A very soft start in my application is the No1 priority hence the choice in a small delay ic. The Lim could make the start pulse situation by hiccuping much worse.

Your views would be welcome.

richy

  • 0
    TI__Mastermind 24715 points

    Rich,

    For whatever reason you selected this bipolar PWM and bipolar driver, there are better choices for high-voltage, high power, high-speed power conversion. Some of the parameters causing you problems are:

    1. the high ISENSE limit threshold of 1 V, means you need a CS transformer for high power applications
    2. The very narrow pulse width (PW) you require and long soft-start time combined with long LEB time, means your initial minimum duty pulses will not appear until the pulse width exceeds the LEB time. This behavior means you will not see monotonic soft start (a step in the soft start as you described)
    3. Current sense transformer in MOSFET drain could saturate during start up (volt-second balance not satisfied)
    4. Total prop delay through the PWM+ILIM delay+LEM+driver+gate xfmr can be a significant portion of a switching period with regards to starting up "blind"

    Maybe you could accept some leading edge spike at start-up by reducing the LEB time and increasing the RC filter, allowing start-up in true cycle-by-cycle current limit - a balance between blanking and filtering the current sense?

    Regards,

    Steve M

  • 0 in reply to Steven Mappus1
    Prodigy 90 points

    Steve,thanks for reply;

    The reason for using the UC3825A/B is I have quite alot of them. Very short propagation delays should deal with the issues, but not so. Something isn´t happy. Maybe, you are spot on re bullet No4, the so-called blind-start is exactly what I am finding below but even so with fast propagation speeds claimed, so I reverted to my simple volt second configuration of the UC3845 (which works). There is alot to think about here, volt second control with limited duty cycle it appears seems more stable without overshoots is the way to go.

    I attempted to configure the circuit using one half output (45%)  with the claimed benefits of the soft start and fault circuitry but this doesn´t  function properly resulting in the Buck pass mosfet latch-up into short circuit destruction. I am seeing an immediate pulse on the controller output the moment Bmax is applied and BEFORE the actual softstart starts. Yes, the component values equating to thresholds are correct and the gate waveform is sound. The worst case is high bus voltage applied before ic control voltage appears and with the circuit arranged for the worst case peak circuit. Using the circuit configured as per application notes, it appears I limit isn´t resetting, given the softstart cap is a fairly high value. The waveforms when it is running are correct. No LEB is being used.

    Both UC3825/45 circuits are built on a ground plane with very short interconnects and the current transformer is well dimensioned. It is often a mystery why well rated mosfets in both dv/dt of the recovery diode, rated current and  max Bmax mysteriously fail for no reason, so I reverted to my bench simple UC3845 and this works every time without mosfet destruction.

    Worst case application is either start control power up before high voltage appears, or vice-versa with hot plugin permutations. The fault circuitry and other should be able to deal with this.

    Gate driver is the FAN 3122.

    Your previous mail mentioned there are better control ic´s for my application; can you suggest one or several ?

    regards,

    rich