Why do I have an unstable output using the LTC4269-1? What you see in the scope is measured between +12VE and DGND. Refer to my schematic. Other than equivalent component replacements, it matches the DC1335B-C evaluation board. I've also attached my bill of materials.
What is interesting is if I apply heat the problem goes away. What I mean by that is I heat the area around the LTC4269-1 with hot air that would normally be used for soldering and the output spontaneously jumps to 12.1 V and is stable. I can power a 6.5 Ohm load with just 100 mV sag (down to 11.99 V) This is repeatable.
I am using a type 2 class for PSE for the supply. Phihong switching power supply. 802.3at TYPE 2 compliant. Model POE36U-1AT-R
I've solved the problem. C18 (C218) was undersized because I used a ceramic capacitor instead of a tantalum polymer capacitor. When I put a 22 uF ceramic on the evaluation board DC1335B-C, I was able to recreate the problem I observed on my design. When I add more capacitance to my design, the problem is fixed.
I can't use tantalum capacitors in my design. What is the most suitable substitute for the capacitance on the Vcc pin? A few options I'm considering:
Here are my measurements from testing some different configurations for Vcc capacitors.
First I replaced C18. It was this.CAP., TANT, 22uF, 16V, B2 NEO CAPACITOR, ESVB21C226MI changed it to this.CAP TANT POLY 22UF 25V 1411 KEMET T521B226M025ATE100
These two different poly caps have equivalent rise and fall time. KEMET T521B226M025ATE100 has 100 mOhm ESR.
Here's how start-up looks with a 22 uF ceramic capacitor.
The dip below the steady state voltage is not good, but it does provide some useful insight. It takes 1.6 ms for the LTC4269-1 to start-up. Therefore the minimum time we need the capacitors to supply energy before the transformer winding can take over is 1.6 ms. This is an assumption. Is my understanding correct?
Next I tested 2 x 22 uF ceramic. The rise time doubled.
The fall time increased. It's longer than 1.6 ms, but not by a generous margin. Keep in mind the 22 uF tantalum has a 5.0 ms fall time.
3 x 22 uF was my next test.
The rise time double again, now longer than the 150 ms for the 22 uF poly tant.
However the 2.7 ms fall time is still less than 5.0 ms.
For reference I measured the capacitance of these capacitors:
I used a TPI (Test Products Int) 192 II multimeter
What you're seeing here is the effect of DC bias on ceramic capacitors. Notice with ceramic capacitors, the VCC voltage ramp increases with voltage? That's because the capacitance of the ceramic is decreasing with voltage, so it charges faster with more DC bias. The tantalum's voltage ramp is relatively constant, because tantalum's have very stable capacitance across DC bias and temperature.
Looking at Murata's website, an equivalent 0805, X5R, 22uF, 25V capacitor may only be 5uF at 15V. As your data shows, you may need more than 3 ceramic capacitors to be equivalent to one tantalum. The ESR shouldn't come into play, this is strictly a capacitance issue.
You're correct, the VCC capacitor provides the bias current during startup. Once the output voltage is in regulation, the third winding will provide the VCC bias. This minimum time you refer to is probably closer to 2ms until the output is in regulation. Your design needs to have enough capacitance to power the IC during startup. Have you considered aluminium electrolytic or aluminium polymer instead of tantalum?
Yes, you're right. I'm surprised that this information isn't clearly illustrated in the datasheet for the ceramic capacitors. I found the information you referenced on Murata's website.
It looks like I can expect about 3 uF when 15 V is applied.
The tantalum polymer capacitor doesn't change capacitance (negligible amount) as the DC bias varies from 1 V to 25 V.
I made the false assumption that a ceramic 22 uF, 25 V capacitor would be close to 22 uF when used near 25 V. My mistake. Thanks for clarifying.