Inverting Buck Boost PSU for PI TX

[Cazavor]

While thinking about different ways to go about making a PI detector power supply, I came up with this and thought it looked promising. I am no expert in designing electronics, so I figured I'd post it here in hope of some feedback on why this would work or would not.

Also as inspiration I saw some users boosting the whole battery pack up to +15V with a switcher and then having a conventional PSU after that. Didn't find anything about using an inverting buck boost though.

I mean something like this:


The +5V analog supply and the +3.3V digital supply can taken from the battery via linear regulators U4 and U3.
The -12V rail that powers the TX coil is created by U2, which is one of those 5-pin "simple switcher" DC/DC converters, configured in an inverting buck-boost topology. It can be shut down by the micro during sampling. Charge is stored in capacitor C3, to provide -12V when the switcher is off.
The -5V analog rail is derived from the negative -12V rail with a linear regulator U5, decoupled with a resistor and a diode.

TX part:


Coil would be connected between battery ground and -12V. Signal from the micro has to be level-shifted to drive the FET. With an optocoupler-MOSFET-driver combined chip, for example. The RX signal would be referenced to the ground rail.

My opinion:
Pros:
- Variable TX voltage
- TX voltage is independent of battery voltage
- Can operate from batteries as low as 6V
- Analog rails aren't limited to charge pump currents
- RX signal is referenced to ground
Cons:
- Uses a switcher, potentially noisy, harder to design
- Needs an inductor and a large capacitor for the switcher
- Uses an optocoupler to drive the TX FET
- Not as efficient as driving coil from the battery
- PPS is limited to the switcher's ability to charge the reservoir capacitor, especially if switched off during sampling
- -5V analog rail is derived from the noisy switcher output while it is operating
Unknowns:
- Will the switcher be able to shutdown and restart fast enough to switch off during sampling?

Thoughts?

[waltr]

start with your last question- No, I do not think the switcher will start fast enough to turn off during
RX period then back on. Prototyping is required to test this.

In general the schem looks ok.

Possibly use a different switcher chip that can be sync'ed to the TX pulse cycle. This would remove the need to turn it off during RX.
This is a practice behind many of the PI design on this forum.

[Carl-NC]

Suggestions:

You may want to split C3 into multiple parallel caps. Better ESR & probably smaller/cheaper.

Instead of disabling the switcher, might can alter the feedback to fool it into thinking the output is too high. I don't know if that will temporarily disable the switching or if the PWM just drops to a low duty cycle, you'd have to try it.

[Cazavor]

Thank you for the feedback. I'll try 3x 3300 uF caps in the final version. About a clockable switcher chip, I have seen clockable charge pump ICs used so far on the forum. I'm sure there are DC/DC converter ones as well though. I'll look for them if this 5-pin switcher I already have doesn't work out.

So far I have built the switcher part of this circuit with a 1000 uF input and 470 uF output cap for testing. It does work. Well, with a 12V sealed lead-acid battery at least. Draws < 20 mA of current while practically idling.

What worries me though is that it does not operate from high resistance sources, such as a 9V alkaline battery or a 12V battery with a 27 ohm resistor in series. In that case, the switcher draws 0.5A of current while the output voltage is stuck at -2.3V. The 9V battery voltage drops to 2V. If I slowly crank the output voltage up from -1.25V to -12V, it works fine. Can't say for sure what is happening, especially as my scope is broken. I am guessing that the switcher increases duty cycle too agressively. Perhaps even to 100%, which would be fine for a normal buck converter, but not for an inverting one.

[Altra]

Hi

Inverting switch mode regulators have a large inrush current which can cause all kinds of problems. Your regulator looks a lot like a LM2596, which I started doing an identical circuit as yours. I made a test circuit that worked well, but I never tested the enable part in a actual pi circuit. In the circuit you posted above I'm not sure the enable will work fast enough or with the proper logic levels? The inverting configuration logic high is B- and low is V-. See link for data sheet figure 20 for level shifter that uses a common base transistor circuit set up to clamp the high logic to B-. There are also soft start circuits examples to ease the inrush current problem. Keep us posted on your progress

http://www.ti.com/lit/ds/symlink/lm2596.pdf

[Cazavor]

Hi

Regarding the enable circuit, the logic high doesn't necessarily have to be B-. 2V above V- turns the regulator off. Up to 25V is allowed in case of the LM2596 and up to 40V in the case of AP1501. Both B- and +3.3V rails sit well in this range, if one is not pushing the voltage specs of the regulator.

I went ahead with the enable circuit I posted, because I already had the traces laid out and I couldn't think of a clean way to have the regulator off by default with the common-base transistor. I lowered the values of the resistors to speed things up, and changed the pull-up resistor R1 to pull-down instead, to have the regulator off by default.


Inrush current seems to be the problem yes. While searching on the net, I found an article saying "Because of the negative resistance characteristic associated with the inputs of many switchers, insufficient soft-start time will result in high input currents, causing the input voltage source to be pulled down. With a lower input voltage, the switching regulator requires even more input current to correctly start up. This, in turn, will pull the input voltage lower still, resulting in a latch condition. When in a latch condition, the input current is relatively high, the input voltage is pulled low, and the switching regulator?s output never reaches its regulated value." [1]. This looks to be exactly what I experienced.

Since I'll have to turn the switcher off during sampling period with the micro anyway, I wanted to also do a microcontroller based soft-start circuit. For this I first made a circuit to monitor the switcher output voltage with the micro's built-in ADC. . Also threw in battery monitoring, since the LM258 op-amp has 2 channels anyway. I chose the LM258 because its input and output ranges include ground, meaning it can output in the 0-1V range for the ADC without needing the -5V rail.

Now for the micro I am using an ESP32 at the moment. Not wanting to mess with the power from the USB cable and from the detector's power supply creating problems, I moved to wireless programming over Wi-Fi. Quite neat. This meant setting up a HTTP web server on the micro, which I also used to create a sort of development or debug interface for the detector. . Here is the micro uploading the measurements to my computer over websockets. Even at just sampling every 50 ms the dip in the battery voltage can be seen when the switcher is turned on instantly.

So to solve the problem, first I just tried waiting a few seconds for the input capacitors to charge etc. before turning on the switcher, but that proved to not to be enough (I believe this is also how the soft-start circuit in the datasheet works). Next I applied a PWM signal to the enable pin of the switcher with a frequency of 1000 Hz and a duty cycle of 0% at first, which I increased by 0.5% every 50 ms. This did work, and allowed the circuit to start from a rather empty 9V battery for example. From the micro's monitoring graph it can be seen that the TX voltage also builds up gradually: . Also I made it so that once the switcher voltage reaches close to its nominal value, the enable pin duty cycle is set to 100% to allow the switcher to do its thing.

[1] https://www.electronicdesign.com/pow...-startup-times

[waltr]

Nice progress and report.

[Altra]

Looking good, one more point. In the datasheet for the LM2596 see figure 18, D3 is important to include and is explained in the above paragraph same page.


I have been playing with Annex Wifi RDS, what tools are you using for the esp32?


https://sites.google.com/site/annexwifi/home

[Cazavor]

Good find, but given my circuit as-is, the D1 and D3 won't be necessary, as I have the input capacitor connected across VIN and ground, rather than VIN and VOUT as in the LM2596 datasheet. The literature I read when designing the circuit mentioned that a capacitor from VIN to VOUT is not necessary, but can reduce output voltage ripple and improve transient response. So I decided to omit it to avoid coupling the VIN and VOUT rails during a TX pulse. But on a second thought, the diode D1 would solve the coupling issue, as shown in the LM2596 datasheet. Had I known about this excellent datasheet TI wrote, I would've likely built it like that.

I'm using Notepad++ to write code, Arduino IDE to compile it and export as a binary file, and upload it via a web browser to the ESP32, which hosts a web server that I set up previously. I can use the Arduino libraries for doing quick simple stuff, and the official ESP-IDF features for more specific interaction with the ESP32 hardware. The graphs I posted are made with data streamed over websockets and displayed with Chart.js. I prefer the C/C++ approach as with scripting libraries at one point you might discover it doesn't have the feature you need.

[Altra]

My thoughts on D3 are- your feedback resistors are connected across the output capacitor. D3 limits the reverse voltage across the capacitor to under a volt while the inrush settles. In turn the regulator feedback sees less variation and is able to stabilize faster.

Thanks for the info on your esp32 tools. Annex Wifi has been easy for me so far. It relies on html for the web page parts, which I am still learning. But as you say at some point you may need a feature which can't be used with curtain tools. Some where on the forum a guy posted an esp32 metal detector project. He was using bluetooth instead of wifi.