UnintegratedCircuit
Part 2 - Circuit Design
​
Identifying Necessary Components
The first thing to do is identify what components will be needed in the circuit and fortunately, since the LT3420 is so application-specific and so highly integrated, the schematic provided in the datasheet can essentially just be copied with modified values to fit this application. Figure 1 shows the example schematic in the datasheet.
​
​
As can be seen, there are not many components at all. In fact, since I intend to tie the VCC and VBAT connections together, C2 can be omitted since only one supply needs bypassing. C3 will simply be the output smoothing capacitor for the supply and will be kept to a relatively low value to keep component size and cost down remembering that only a relatively small current will be drawn from the output.
​
Calculating component values
To start with, the two easiest components to establish values for were calculated, these were C1 and RREF. On page 6 of the datasheet, it says: "Reference Resistor Pin. Place a resistor (R2) from the RREF pin to GND. 2k is recommended.". Following this advice (a highly recommended course of action, especially when designing power circuits) generates the following 'equation', if you can even call it that.
Next, a value for the input bypass capacitor C1 was decided upon. Also on page 6, the datasheet states for the VCC pin: "Input Supply Pin. Must be locally bypassed with a 4.7µF or larger ceramic capacitor". Due to my stash of components left over from previous endeavours, a 10µF capacitor was chosen to bypass the input. This also helps accommodate for any bypassing that might be required for the VBAT pin (although I don't know if any is needed when tying the two pins together like this). As a result, the next equation is as follows:
The next component that needs to be calculated is the transformer (since every other remaining calculation relies upon these values). First off, the minimum turns ratio is calculated according to the equation on page 10 of the datasheet.
The next equation relating to the transformer is determining the minimum inductance of the primary winding. This is in order for the IC internals to be able to sense the magnitude of voltage on the transformer primary winding when the power switch turns off and the polarity of the voltage on the primary reverses. The minimum primary inductance takes the turns ratio as one of the variables; however, only a minimum turns ratio has currently been obtained, not the actual value since a transformer has not yet been decided upon. For an initial calculation (for ballpark figure purposes) the minimum turns ratio will be rounded up to the nearest integer value and plugged into the following equation
So, as a very rough figure, about 5-15µH of inductance on the primary winding, depending on the turns ratio, although this is a minimum value so going a few µH's above this value should be fine. Additionally, inductance appears to drop off in an inductor/transformer winding as it approaches saturation (don't believe me, look on the Coilcraft website, they specify the saturation current rating for basically every coil based upon a 10% drop in inductance).
​
With these figures in mind, it is time to choose a transformer. For this, I had a look on Coilcraft's website because they offer a lot of custom transformers with high turns ratios, which not many other companies do and the specification on their parts is second to none. They also offer samples on pretty much every single part which means less cost for me (ahh student life). I decided to choose the FA2469 from them due to its small size and high primary inductance (19µH dropping to 14.86µH at saturation) - as well as the overkill turns ratio of 1:16. An appropriate saturation current rating of 1.5A (versus the 1.4A peak of the LT3420), low primary winding DC resistance, high isolation voltage and low leakage inductance combine to make this an attractive option.
​
In hindsight, however, I would have more likely chosen its cousin the FA2470 with a slightly less overkill 1:8 turns ratio. As shown on page 12 of the datasheet the peak secondary current is shown to be the primary current limit (a constant) divided by the turns ratio, therefore, by using the FA2470 and halving the turns ratio, I could theoretically double the output current (probably not but it would probably increase the current capability somewhat). Even with this reduced turns ratio, it still satisfies the minimum requirement comfortably - for both turns ratio and inductance - and would also require a derated rectifying diode due to the equation below:
However, if this was changed to the lower turns ratio of 1:8, this would reduce the peak reverse voltage to:
This could allow the use of a more efficient (less lossy) Schottky diode as opposed to the dual switching diode that is currently implemented. This dual diode will likely have a slower reverse recovery time than a Schottky diode as well as double the forward voltage drop because, well, two diodes.
​
Talking of the dual switching diode, the MMBD3004S was chosen since it provides two 300V capable diodes in series (hence the trailing 'S') in a small SOT-23 package and is also recommended in table 3 on page 12 of the datasheet (and is therefore probably compatible with this IC).
​
Finally, resistor RFB can be calculated following the equation back on page 6 based upon the target output voltage. The calculation runs as follows although the forward voltage drop of the diodes here have been neglected since they will only form about 1% of the total output voltage.
Rounding this value to 22k because it is a very standard value gives the value for the feedback resistor. Again, this would have to be recalculated if the lower turns ratio transformer was used... But that is the other beauty with this IC. Analog Devices offer some similar ICs such as the LT3468 which has even fewer external components BUT, by sacrificing RFB it causes the output voltage to be entirely dependent on the turns ratio which limits the transformer selection even more. It also has a much lower input current limit which means even less output current which is also undesirable... Anyway, I digress.
​
The penultimate value remaining is the output capacitor, C3. This was chosen to be 1µF which might seem very low but again, only a small current is flowing and high-capacitance capacitors with a 250V voltage rating are often physically large and expensive. This gives:
The final component value to calculate is for the refresh capacitor, CT. The equation for calculating this is once again on page 6 of the datasheet and is as follows:
To get a short time period (in an attempt to eliminate audible noise) an initial value of 100pF was chosen which give a time period of 40µs and therefore a refresh frequency of the reciprocal of the time period which gives 25kHz. Therefore:
​
The Final Schematic
The final schematic - created using EasyEDA - is shown in Figure 2 below (including an annoyingly asymmetric symbol for the LT3420, a DIY symbol for the transformer and a ridiculously large symbol for the diode... Oh well).
​
​
The next instalment of this series will showcase the PCB layout, including some of the design 'tricks' that help optimise the operation of the end product.
​
​
​