LT1505 Charger turns 1.35Mhz Oscillator

[SaabFAN]

Well, I'm out of ideas. Maybe you can help me with this...

The LT1505-based Lead Acid Battery-Charger I built switches at 1,35Mhz (and gets really hot) and is not charging the battery properly.

The symptoms I noticed is that the voltage at the Iprog-Output jumps up to 3,8V immediately upon startup both with and without battery connected. But the voltage across the Sense-Resistor Rs1 is 0mV.
Also Vc isn't going higher than 0,5V.
Once I manage to get it working, which is only possible by pulling down the Prog-Pin with the FullCHRG-Mosfet (which I put in there to have 2-Level charging in case a battery is deep-discharged or very cold), the chip starts switching, but not at 200kHz, like the datasheet says, it switches at 1,35Mhz +/- 100kHz (depending on Vcc). The required current to switch the 2 MOSFETs at that frequency heats the chip up to 125°C (at least that's what my infrared-thermometer shows - The temperature-sensor of my Multimeter is broken).

I wanted to equip the boat of my parents with this thing (together with an LTC1871 based Max Power-Point Tracker that is almost finished - I have posted about this in another thread here already) after the chargers based on the BQ24650 blew up once a Novitec "Megapulser" was connected. That's why I added the line-filter to the output of this charger. It almost completely eliminates the pulses from the megapulser (less than 50mV of the pulse remains, measured across the TVS-Diode).

I hope you can help me with this circuit. As I said, I'm completely out of ideas.
The circuit is almost exactly built like specified in the datasheet, but modified in terms of component-values. The 4093 isn't installed at the moment and is supposed to add low-level charging in case the battery is deep-discharged or too cold, and disable the entire charge in case the temperature is too high.

[Buriedcode]

I quick look at your schem, and the datasheet, you have a 4.7uH inductor which to me seems a bit low for 200kHz switching.  Have you tried powering it up with a light load attached instead of a battery, just to test the buck part?

[SaabFAN]

Yes. If I do that, it switches a few times at about 200kHz and then turns itself off. The voltage at the Prog-Pin rises to 3.8V and at the Vc-Pin it rises to bout 0.5V. The voltage at the output then slowly goes down to 0V and nothing ever happens, except when I pull the FullCHRG-Line high (which reduces the voltage at the Prog-Pin to 1.8V and at Vc-Pin to 0.18V). The current-consumption in this state is 9mA and when I pull the FullCHRG-Line high, it goes up to 70mA.

Because this charger is supposed to charge a large 160Ah-Battery, I have calculated all the values to make it deliver 20A, but the best I can get is 420mA at a supply-current of 350mA at 22V.
At these currents, the inductor should be sufficient.

What baffles me most is the fact that this chip actually switches at 1.35Mhz, despite both drivers being AND-Gated to a 200kHz oscillator.

[SaabFAN]

Before the chip released the magic smoke (smoke came out near the SENSE-Pin), I managed to capture the waveform on both the TGATE-Pin and BGATE-Pin side-by-side.

As you can see, the internal oscillator seems to be working, as well as the comparators and logic-components that switch BGATE off and stop TGATE from turning on before BGATE is off.
But TGATE then oscillates.
I suspect A2 to be responsible for that, as the BOOST-Pin shows the exact same waveform as SW, but with an offset of about 7V (according to the datasheet, this offset should be 9,1V btw.).
Also GBIAS, which should be sitting at 9,1V, goes down to about 6,7V and has almost no ripple on it.

Btw. I'm using BUK9Y65-100E N-Channel MOSFETs from NXP with a total Gate-Charge of 14nC at Vgs = 5V. I'm beginning to suspect that I might be overloading the drivers and the GBIAS Regulator^^

[SaabFAN]

Is there a possibility that the use of logic-level MOSFETs causes this problem? With 9.1V driver-Voltage from the GBIAS-Pin, I guess the chip was designed with older FETs in mind, which require a bit more voltage to turn on properly.

I also, on a hunch, increased the size of the bootstrap-capacitor with basically no effect on the waveform seen at the TGATE-Pin. But when I connected the battery after this modification and pulled the FullCHRG-Line HIGH, the chip released the "magic smoke".

There's also a rather crazy tendency of the circuit to start oscillating when I disconnect the Power-Supply but leave the battery attached: The Vin-Node suddenly showes a sawtooth waveform with an extremely fast rising edge (my PM3320A scope was only barely able to trigger on that edge and its trigger-circuits are good up to almost 600Mhz - the Rigol wasn't able to trigger at all on that).
This phenomenon also makes itself known by a fast ticking sound and went away when I set the Undervolt-Lockout to a higher voltage.

Edit: I attached the PCB-Layout. Maybe there's some error in there I haven't seen yet...

[Marco]

Btw. I'm using BUK9Y65-100E N-Channel MOSFETs from NXP with a total Gate-Charge of 14nC at Vgs = 5V. I'm beginning to suspect that I might be overloading the drivers and the GBIAS Regulator^^

I think it doesn't have low enough output impedance to keep the MOSFETs turned off, that would explain the high frequency (because of the shoot through prevention comparators going wild). Maybe adding a little extra gate-source capacitance could help?

[SaabFAN]

I think it doesn't have low enough output impedance to keep the MOSFETs turned off, that would explain the high frequency (because of the shoot through prevention comparators going wild). Maybe adding a little extra gate-source capacitance could help?

Interesting read.
I'm certainly going to try to add capacitance to the gates. According to that document, the BUK9Y65-100E is only just inside the green-zone of the ratio of GateSource-Capacitance and GateDrain-Capacitance. (Cgs/Cgd >= 15)
I'm also thinking of using an additional 0,1µF cap to GBIAS, to improve high frequency performance and charging of BOOST, which isn't a bunch of small traces, but instead a rather large piece of copper (could fit about 3 SOT-23 pacakges on it) sitting right beneath the plane that makes up the SW-Node.

[SaabFAN]

Well, unfortunately, adding capacitance to the gates didn't really help. It dampened the oscillations a bit and reduced the rise-times, but didn't do much else.
Reducing Vin to 19V did quite a bit more, but there's still oscillations when the chip switches.

So far it seems as if the problems are primarily related to low power-conditions, when the output-current is less than 1/20 programmed current (I currently have the circuit configured to charge with a maximum current of 10A). When I demand currents in excess of 1.5A, the circuit behaves more or less like it should. Below that, it starts oscillating everywhere, as the control-loops start to do stuff they aren't supposed to. At the same time, the chip also gets almost too hot to touch.

One other thing I noticed: Once I connect a battery, the voltage at the Vfb-Pin starts rising to about 2,7 to 3V, despite the fact that it is configured in a way that 14.40Vbat = 2.465V at the Vfb-Pin.
Somehow there's current flowing out of the Vfb-Pin if the chip isn't doing anything, which in turn prevents it from doing anything because the VA-Amplifier inside the chip effectively disables the chip.
The only source for current coming out of the Vfb-Pin seems to be Buffer-Amp A11, which is connected to the 3CELL-Pin and Vfb-Pin internally via resistors.
Btw. the datasheet explicitly states that the 3CELL-Pin should be left floating if not used.

Also interesting to note: If the load at the output is high enough but not too high to trigger any off-switch condition (Undervolt, Overcurrent, Overvolt at the output), the chip turns into a linear regulator, turning TGATE on for a prolonged period, only occasionally turning it off for a brief moment to recharge BOOST. Because BOOST can be held up for so long, my theory about A2 causing these oscillations is pretty much busted and I have to look at some other component...

[SaabFAN]

Well, I think I've managed to make it work with a few workarounds and see the limitations of this chip:
- Vin may not exceed 20V, otherwise the chip overheats. 18V seems to be best (Basically 19V Notebook-Supply - Diode-Drop. Not too surprising, considering this chip was designed with Notebook-Batteries in mind)
- The removal of the Gate-Resistor of the TOP-MOSFET and adding 220pF to the GateSource-Capacitance dampened the oscillations visibly on the scope, but they are still there.
- The chip doesn't like load-situations where the current is less than 1/10 of the programmed current. Once the current drops below that point, it seems that the control-loops are outside of their comfort-zone. Load-Currents below 1/20 of the max current make the chip really go crazy (current draw from the lab-supply wildly jumps from 50 to 300mA. Above that, the circuit becomes more and more stable.
Something I didn't expect, because so far I've only worked with chips that were easily able to regulate for 100% or fractions of the maximum current.
- The chip seems to have a current-path from the 3CELL-Pin to the Vfb-Pin. I don't know if this path is created because I may have used a bit too much heat (I had set the iron to 350°C, which every other chip I soldered so far is fine with), or if it is just there and the datasheet doesn't tell you about it. I have now simply connected 3CELL and the adjacent pin to GND and use the internal resistor from Vfb to 3Cell as the GND-Resistor for the voltage-divider.

There might also be a problem with either the board, or the chip is sensitive to flexing: If I put pressure on the board or the chip itself, the charger turns itself completely off. Could also be a problem related to overheating though.
I think I'll have to assemble another one and see how that one holds up.

[SaabFAN]

I am beginning to suspect a charge of bad chips.
I assembled three more boards, with completely different results:
- The first one started smoking as soon as I applied power
- The next one seemed to have an internal short between Vcc and BOOSTC, as well as problem internally
- The third one didn't supply 9,1V to the GBIAS pin (measured at the actual pin, not the pad) - Only when pressed in that area, GBIAS went up to 9.1V, but again, there was oscillations on the BOOSTC-Pin.
Both boards were soldered with the iron set to 270°C and I used 60/40-Solder. All the solder-joints looked good and measuring for shorts, I didn't find anything. Measuring the pins to the pads, I always got good contact.

Does anyone have any info on how to return chips to Digikey, or Linear Technology?
The chips have been fabricated in week 44 2015 and came in a plastic tube put inside a conductive bag.