Need Current Reduction in SMPS Circuit

[bobledoux]

Diver300:

I'm using the comparator, not the ADC for voltage control. Microchip also recommends a 10K input impedance limit for this function.

Looking at the original Walther schematic, that I posted earlier, the voltage divider consists of R8, which is 10 meg and R9, 300K.

As I recall, the battery current draw with charged capacitor was 28ma. My simulation suggests initial current draw with discharged cap can be 2.2 amps for 1millisecond. This suggests that keeping the pulser on may allow the power supply to function at a voltage, below which, it is able to start from zero charge.

I haven't experimented with cap front loading an ADC. I'll look into it.

 

[Blueteeth]

The sudden current spike at the start *might* indicate a saturated inductor. Perhaps the first on-time charged the inductor to say, 0.5A, and you started a new on-time before the inductors current dropped to 0. This is the trouble with constant on-time, it assumes that at the begining on the switch on, the inductors current is 0. Soft start usually starts off with a very low on time, and slowly increases it after the first few pulses.

 

[Blueteeth]

Sorry, just realised, at start up, your HV cap is at 0V. When you first ocnnect the power (with the mosfet off) the cap charges via the inductor and diode to battery voltage - diode voltage. This is the trouble with shorting the output in boost converters, there is no isolation, the output is directly connected to the input. Normally, an input capacitor will smooth this out. But for the first milisecond, I wouldn't worry. You could always put a resistor in series with the HV cap - it'll reduce efficiency, but means that when the battery is connected, the cap charges to the battery supply via the inductor (0.2ohms?), plus the diode (0.6V) and this resistor.

Often a coupled inductor is used (ergo, flybacks) to provide isolation, but thats a real hassle if others are to build this, as it means either winding ones own transformer, or finding a coupled inductor thats easy to get hold of

 

[bobledoux]

My simulation experiments suggest I can get the initial current below 200 ma by using a longer off time with a 330uh inductor. The period of high current appears to be about 1 millisecond. With the PIC I can set starting pulse parameters assuming a dead cap, and then modify parameters for restoring a partially discharged cap.

I'm getting ready to test the start method just mentioned. I'm also going to increase the feedback resistances by at least 10 times. I also want to try leaving the SMPS on so the cap retains its residual charge. These techniques are all intended to reduce peak current flow and power flow.

I need to monitor current flow on my scope. I have a nice Tek TDS1012, but only voltage probe. So I plan on running the circuit battery through a big 0.47 ohm resistor and measuring voltage drop across it leads. Any other ideas for current measurement?

 

[Blueteeth]

Sounds good man. Using a current sense resistor is handy jstu means you have to do some maths :/ For such a low current loses aren't a problem, a 0.1ohm would make calcualtions easier, but it'll only dorp a tiny voltage (good, because you want your circuit to see the original battery voltage), a 1ohm gives a bigger voltage, but thats more voltage drop :/ But 0.47 sounds great.

Just remember, when you 'fire', you're effectively shorting your power supply across your solenoid, diode, and inductor once the caps voltage drops below 9v-0.6V = 8.4V. Because caps voltage is greater than the batt voltage, the diode won't conduct. But... as the caps voltage drops below vbat-0.6...it conducts.

As for charging, I'm running a sim right now with a 330Uh inductor, to get a nice smooth constant current draw to charge the cap in 20 seconds. (you can top it up just by starting the charge cycle again, until it trips the desired voltage via your divider).

As for the 'idle' current draw. It all adds up, the voltage regulator for the PIC will draw some current (7805's draw around 5-7mA), the PIC will draw a few mA, as well as your MOSFET driver. I realise theres no point in really going for absolute minimum current draw because it becomes more and more complicated just to save a few mA. But minimising the current draw (and variation in current draw during a charge cycle should just be a case of inductor value and on/off times.

Edit:
For a 330uH inductor with a 0.2ohm resistance. With a 9V supply. In order to charge a 100uF cap, to 90V over 20 seconds. I make it: On time = 7uS. Off time = 220uS. Ipeak = 201mA. Current draw from battery (constant during charge) = 7.2mA.

The initial current draw is when the caps voltage is at 0, and is charge to ~8V via the diode and inductor ... thats 2A. But its a simulation, and in the real world, it won't spike that high, the battery's resistance prevents that.

 

[bobledoux]

My sims confirm your values. I want to see if the function meets the simulations. I'm curious whether I achieve full charge voltage with these long periods. My in circuit tests showed that an 8 volt battery would not achieve a full charge with less than 6usec pulse. And burying that within the longer periods just extended the charge times.

The PIC has a powerup delay of about 60msecs after it sees Vdd. During this period one cannot depend on the pins to be in desired states. So my R11 and R10 resistors are pull downs to ensure the FETS remain off. I consider them good practice rather than demonstrated need.

Thanks for your feedback.

 

[Blueteeth]

With the values shown, the ON time determines peak current (with only 1us resolution over 6us... you can't reduce it much, but can easily increase it). But the OFF time can be varied considerably, sinces its just the PR2 register + on time Reducing the PR2 value speeds up charging.

How about this. You start off charging with 6us on, 220us off (PR2 = 226). You could easily up the ON time, but lets just stick with this for now. Set your comparator to check for say, 30V. And with a timer/counter, time how long it takes to get there. If it takes too long the charging is too slow (low battery voltage?), so you can reduce the off time (PR2 dropped to say, I dunno 150). Then change your comparator reference to 50V. Again compare with a timer.

Whilst it sounds complicated, its really just changing a couple of registers, waiting for an interrupt (or just polling the comparator toggle bit), then making a comparison. No need to adjust the charging on/off times constantly, a few times just to 'steer' it towards your target charge time. This way is only uses up the current it needs to get to the target charge time - also keeps the inductor current at a minimum.

Good call on the pull downs. Also, in that 60ms start up the HV cap will be charged to batt voltage - 0.6V. So, once the PIC is up and running, there shouldn't be any large current spikes from the battery, only idle current.

 

[bobledoux]

Changing PR2 is simple. I understand how this would work. I could also do it by using a timer that reduces the period as voltage rises, though this would not be directly related to charge level. It would be selected based on the worst case scenario.

This is all predicated on how the area under the curve for current changes as cap voltage increases. Restating what you have said: Assuming a 7usec pulse, at the beginning of the charge period, the current reaches a peak of about 180ma as the pulse ends. It then takes about 24usec to fall back to zero. At about 300msec into charge, the capacitor voltage is about 27 volts. Here, the current fall back to zero takes about 3usec. So the current x time has fallen greatly. By reducing the PR2 we reduce the period for the beginning of the next pulse raising the average current x time between pulses. I'll play with this.

I've given further thought to charging time: If it exceeds 10 seconds there may be a tendency for the user to assume a low battery or malfunction. People won't wait much longer than that. The factory board charged on 5-7 seconds, so 10 is where I'm aiming. I also note the ten seconds doesn't have to start until the loading lever is closed so when the lever is open a "sustain current voltage in cap" routine makes sense. This should not be hard to do.

The decision making provided by the microcontroller really provides lots of options with a design of this nature

 

[bobledoux]

I’ve achieved good working parameters for my project. I made a 60 charge series using a well worn battery with a starting voltage of 8.46 volts under an average load of 24ma. The following set shows the data in the series:

Charge 1, Voltage 8.46, Charge time 5.6 seconds
Charge 21, Voltage 8.35, Charge time 5.8 seconds
Charge 31, Voltage 8.32, Charge time 6.7 seconds
Charge 41, Voltage 8.28, Charge time 7.0 seconds
Charge 51, Voltage 8.20, Charge time 7.3 seconds
Charge 61, Voltage 8.17, Charge time 7.8 seconds

A new battery voltage was 9.17 volts with a 4.4 second charge time.

The 0.29 voltage drop for 60 discharges suggests I don’t need heroic efforts to achieve a reasonable number of cycles per battery.

For this test I simply charged up to full charge level of 95 volts into the 100uf capacitor and then let the charge drain off. With the existing voltage divider (47K & 1K) in the circuit it took about 12 seconds for the capacitor to fully discharge. The battery power remained sufficient for continued use when the test terminated.

My final SMPS parameters are a duty width of 8usec and a period of 40usec. When I tried using a pulse of 7usec, the circuit would not achieve full charge with the battery of 8.46 volts. Increasing to 8usec lowered the functional battery voltage floor. But the peak current rose from 181 to 200 ma with the 1usec increase.

It appears that a longer pulse rate generates battery function to a lower voltage level. This comes at the cost of a higher current flow and shorter charging time when the battery is new.

I’ve achieved my aims in solving this problem. Thanks for the contributions.

 

[Blueteeth]

Good man!

Yeah its all a compromise isn't it? Charge time, battery life, peak current .. a 9v PP3 isn't ideal as a portable power source, because for size and weight it has less energy density than AA's, but AA's work at lower voltage etc..etc.. I often end up designing several versions of everything I build, purely because of my indecision in these matters.

Glad you acheieved your goals man. The lower peak current in the inductor makes it much easier for people to build the circuit themselves. Apologies if I seemed to take over, or saturate the thread with numbers and suggestions, I know sometimes it can be too easy to 'over engineer' some applications, but for this one? it can get pretty complicated given all the variables. Thanks for posting results and updates, I'm sureit will help others for any similar problems (charging caps, from battery, at any voltage).

Good luck!

 

[bobledoux]

Blueteeth: you gave me lots of ideas. I'd rather have a number of options to try than just one. Thanks for the input. Its nice to know that electron flow performs the same way in Great Britain

 

[bobledoux]

I'm back again with a final question or two. I've re-posted the earlier diagram.

My scope doesn't show that C7 does much; its 100uf. Is there a reason to leave it in?

Are there other suggestions for improving this circuit, especially the SMPS?

 

[Blueteeth]

Hi again. C7, the input cap as you probably know is to reduce transients on the input power. In this case with the input power being a battery, I doubt you'll need a special low-esr high capacitance cap. But its standard practice to add at least some capacitance there as it *can* supply a large current spike for a breif period of time should the circuit need it. In most SMPS's the input cap is important to prevent noise on the input power supply (which often powers other circuits that could be affected by high frequency ripple, as well as a sudden 'drop' in voltage for sudden current transients). That said, your input is only a battery. All batteries have internal resistance, so during your ON time, the sudden current rise in the inductor provided by the battery, will inevitably drag the batt voltage down during on time, changing charging time. Have you scoped the batt voltage withouth the C7? I'll be surprised if a 100uF has little effect. If the batt voltage has big ripples on it without C7, maybe just try a smaller value, 47u, 22u, ...

As for improvement. I would add a 10uF or greater value cap across the 5V power supply, because the 7805 regulator - whilst able to supress transients well with a 0.1uF cap - might need a bit more to keep the 5V rail stable. Also, as I don't see a voltage reference I'm assuming all the references for your PIC's comparator (and ADC) come fm the 5V rail.. all the more need for this to be 'clean'.

I'm quite surprised you're using a beefy MOSFET driver, although it IS ideal, as it draws very little power, provides very fast rise/fall times (important for efficiency) as well as level translation. I think it could be replaced by a few transistors but thats just more components. Easier to obtain for others wanting to build, but less ideal.

I can't really see any major improvements needed. Anything such as 'better mosfets' or 'low power techniques' for the PIC whilst be beneficial to battery life, wouldn't improve it very much considering the effort required .. then you get to the point of 'diminishing returns'. So, if it does the job you intended it to do, parts easily available and your scope tells you its working fine, go with it!

 

[bobledoux]

Would it be reasonable to put a 47uf capacitor in place of C1, and remove C7? What about a similar cap for C2?

I typically use the Vdd for reference voltage. Most of my applications don't use a voltage regulator so as Vdd varies the rest of the circuit varies as well. This keeps the comparator and AD inputs proportional to Vdd. In this application voltage reference will change if the battery voltage drops below the dropoff ceiling for the 78L05, or about 7 volts. This would change the high voltage output. However, my SMPS design parameters are such that full charge cannot be achieved before the battery falls to this voltage level.

I am considering addition of a timer is software. If it takes too long to achieve charge, like 10 seconds, the software would shut down until the battery is replaced. One consideration for this is safety. I have no experience as to how this circuit will act at low voltages. I can't be confident that at low voltage there might not be an unintended firing pulse. I have no reason to believe this is so, but in this application I can't afford for untested parameters to result in serious consequences.

Its great that software allows such additional measures and controls to be added after the hardware is complete.

 

[bobledoux]

I can't find much information on the input and output requirements for capacitors in the 78Lxx series voltage regulators. Most documents recommend .33uf on input side and 0.1uf on output. These are conditioned on distances to other components and need to prevent oscillation in the regulator.

My concern is about identifying components necessary for operation of the Voltage reg versus those contributing to down stream function.

 

[Blueteeth]

The voltage reg probably only really needs 0.1-0.47u on its output. As an NPN-pass linear regulator they are uncondtionally stable (unlike LDO's which can oscillate if you don't meet specific output cap requirements). So it'll do the job just fine with a low output capacitance as long as its not a long way from the PIC. Any additional capacitance is just gravy (you'll also notice many desginers use a 10uF electrolytic, and a 100nF/0.1u ceramic for the high frequency content).

Placing caps right at the intput of a voltage regulator, and close to the input of any switched-mode power supply is common practice - not always necessary, but usually prevents odd behavoir, and many designers just add them because they are easy and cheap, even though you end up with a few caps, or different values, essentially in paralllel. Any capcitance you add across the battery will do the job for the regulator as well, so I don't tihnk you need two caps there.

 

[bobledoux]

Capacitor Selection

I'm back with a question about the SMPS high voltage capacitor:

I need a 100uf, 160 volt rated capacitor for pulse discharge service. My current boards are using 100uf, 330 volt capacitors from single use camera strobes. These are working well but they are very small, suggesting they are designed for a minimal service life.

Pulse discharge capacitors are low ESR. Mouser doesn't list any low ESR caps with more than a 100 volt working voltage. Rubycon makes a series FW capacitors for strobe service, rated for 330 volts. But they are too long to fit my board. They do produce a 120uf, 330 volt cap of small size used in Kodak cameras. But its not in their catalog, nofr found at normal electronics parts houses.

Am I worrying too much about putting a "general use" 100uf, 160 volt capacitor into my circuit? What type of failure might I witness with these in the board?

 

[Blueteeth]

Hi,

Thats a very good point! I haven't taken apart any commercial high power 'strobes' which would pulse at up to 10Hz and probably have a life in terms of thousands of hours, but they would have to use a half decent cap. Where-as, as you pointed out, disposable camera flash circuits only need to work for 30 flashes.. Even though I would have thought manufacturers would use as cheap cap as possible, in my *limited* experience with playing with disposible camera circuits is, the xenon tube goes first, then the transformer - I haven't tested the HV cap to death. So I'm out of my depth here :/

That said, considering your coil has a fair ammount of inductance, current doesn't rise instantly anyway, and at the high voltage of 90-100V, roughly 8-15x the usual voltage of 'low ESR' applications would make the ESR less of an issue. And just because a cap isn't listed specifically in the 'low ESR' catagory, doesn't mean to say it doesn't have a very low ESR I would gather a lst of candidates from your supplier, with the right voltage/capacitance, and see if you can get the ESR for every one, and compare them with a cap specifically for the job (the rubycon FW ?). I'm pretty sure the higher the rated voltage of the cap, generally the lower the ESR for a given capacitance.

One more point, the cap you're using is designed for a 280-300V discharge (disposible camera's never charge to mre than 300V), but you're using it at 1/3rd of that voltage. My impression is, even if its a 'cheap cap' with limited life, using it at a much lower rating, combined with the fact its driving an inductive load (limited current rise time on discharge) means it should last a lot longer than its camera application. Sure, its probably not a good brand like rubycon, an may not be ideal for it, but seems like its not a bad idea - plus easily obtainable for those who wish to clone the circuit.

As for possible failures when a cap is at the end of its life in this app? I'm afraid i don't have the experience, just what I've been taught, and what I've read Large current spikes pushing the limitations of the caps 'rated riple current' (again, related to ESR) over heats the cap - however you're not discharging this 50 times a second.. such a low duty cycle as 'two shots a minute' should be very gentle on the cap. But I would up the rated voltage to something at least 20% more than the maximum charge voltage you intend to use. For 90V, 120V+. Over voltage can break down the dielectric, and if the cap survives, drop its rated voltage even futher.

Sorry I cannot be of more help, this response is pretty much just me 'brain storming' and thinking out loud, I have no had any experience of long-term use of caps in discharge circuits, but I'm *sure* someone in here has!

 

[bobledoux]

Late testing has shown a high voltage required of 150 volts, not 100 volts to get good primer ignition. This means I’ve changed the Q2 firing FET to a much higher current capacity. I’ve ordered the IRFR4620PbF for this purpose.