LT8304-1 double soft start?

Question

I've been trying the LT8304-1 for a high-voltage power supply for a prototype. I've laid out a test board based on the 1000V flyback application note at 
 www.analog.com/.../1000-v-output-no-opto-isolated-flyback-converter.html 
 I used the recommended transformer as well as the recommended snubber, and kept the critical loop nice and tight. (The pin-out of the IC is ideal for this.) 
 For the most part, it works really nicely, reaching full voltage quickly both when lightly loaded and with a heavier 5mA draw, and seems very stable as the load varies. Probing the switch node, the transients never exceed 90V, well below the 150V internal NMOS limit, so I guess the loop area can't be too bad and the snubber is doing its job well. 
 The only question I have is about the soft start, which often seems to re-trigger once or twice part way through start-up. I've attached some scope traces of output voltage (divided down) and switch node (pin 5) to show what I mean. 
 The yellow trace in each case is the pulse going into EN (pin 1). The behaviour is the same whether running off 8V or 18V, and whether running lightly or more heavily loaded. Once the voltage is over about half-way, the soft restarts don't seem to happen, and there are never drop outs or problems when running at full voltage, even when loaded up. I've zoomed right in at the time of the soft restart on the switch node to demonstrate there isn't anything obvious catastrophic going on around that time.

I checked both Vin and INTVcc are stable - I'd wondered if they might be dipping but they're fine - no evidence of any dip. 
 I also probed Rref (pin 7), which looks sensible, without lots of high-frequency noise. I tried increasing the 10p capacitor to 33p in case this was insufficiently damped, without any obvious effect, but it didn't look bad in any case, and the failure is happening when it's pretty low rather than when it's higher later on: 
 
 Assuming this double soft-start behaviour isn't normal, any suggestions of other things I should check? I'm wondering if these symptoms are an obvious smoking gun to someone who knows this IC better!

Chris81 · Answer

I submitted a support query to confirm, and had a very helpful response - thanks Hermo! I thought I would summarise the thread here in case someone else spots the same behaviour and searches this forum for answers. 
 Yes, the restart every 11ms until the output voltage exceeds 60% of the target is expected behaviour and is the reason for the hiccups I saw. 
 Reducing the output capacitance (so charging to 60% takes less than 11ms) would eliminate them, at the cost of a smaller charge reservoir/greater ripple. In my own application, I'll try using 470nF X7Ts instead of 1uF X7Rs, which should have the same capacitance at 300-400V DC bias, but lower capacitance earlier in the charge cycle. Unfortunately, 270nF C0Gs are probably too expensive and bulky at the required voltage, although they'd be ideal. 
 The hiccups aren't really a problem in my application. Even with them, full voltage is reached more than quickly enough. However, for curiosity's sake, I tested biasing the bottom of Rref to 0.5V and reducing Rfb to set the same output voltage at the 1.0V comparator threshold. 
 The converter does charge and stops at the correct voltage when you do this, but much slower than expected and there is an audible whine. I'm guessing the bias must interfere with the operation of the DCM comparator, causing the converter to drop right down to the 11kHz minimum switching frequency. In case anyone finding this thread is considering this workaround in their own application, it looks like it is a non-starter.

Partner Forums

Partner Libraries

Products Mentioned

LT8304-1 double soft start?