Your first two waveforms look strange. Is the gate drive inverted? I'd expect the source waveform to look very similar to the one you give for the fet current.
It does not look like 'classic' subharmonic oscillation. For that you would expect the peak current to be terminated at the same value but the pulse widths would alternate with inductor current being continuous.
In this case it looks like it is the oscillator which is terminating the drive on one switching cycle before the current limit comparator has been tripped. After the, short, deadtime another switching cycle is initiated almost immediately and the thing comes up with continuous inductor current.
80% duty cycle sounds 'wrong' if this is a single switch forward converter. Unless you are allowing 4xVIN plus VIN or 5VIN during reset. If this is an offline converter with say a 155VDC bus then you are in the realms of 800V plus rated devices......
Subharmonic oscillation supposedly only occurs above 50% duty cycle with continuous inductor current and 80% certainly satisfies one of those conditions.... If you are intending to operate with continuous inductor current then you will need the slope compensation.
It might be possible to end up with the behaviour you have if the output of the VEA can rise to too high a level such that it 'thinks' it in regulation but it isn't. Try clamping its output to a level that is consistant with your expected operating conditions.
You might also try increasing (doubling) the value of your filter inductor to gaurantee continuous inductor current even if it tries to operate in the way you are seeing and then it won't be able to.
There is meant to be some issue about the internal current loop bandwidth changeing as the operating mode shifts from continuous to discontinuous operation but I would not be so sure that that is such a major problem.
No, you're just seeing one heck of a spike on the FET's drain at turn-off due to the reset winding being 1 turn and the primary being 4 turns (needed to that I can get up to ~80% duty cycle -- this was driven by trying to keep the primary currents at a halfway sane level given that I'm using 3V input.) -- hence you get (1+4)*3V=15V during reset.
Thanks, that's quite helpful to know.
I posted the schematic and more details on ABSE; it's "nominally" 3.6V in ->14.7V out, with an input range of 3V-4.2V -- hence all the voltages are quite "tame" relative to off-line switchers.
Ah, good idea -- I should have thought to try that.
Oh Right, I've just seen your circuit diagram. SRC is the drain. I thought SRC referred to the source.....
I've had a little play with your LTspice model. One thing that struck me as odd is the value of filtering components used on the Isense pin.
47R/10n seem exceptionally small/large but that's what they show on the application note. Depending on other things it may be possible that the slope compensation signal is just dissapearing down the 10n capacitor....
Try making R6 100K in your model (sort of gives the minimum the IC will give) and doing a transient analysis from 500u-600u and look at Sense. Look at the waveform at the Isense pin of the IC after you have filtered it. It's almost but not quite classic subharmonic oscillation.......
Now change R12 to 1K and C4 to 220p, make it's ESR zero. Leave R6 as
100K. Run the analysis again. Now you get stable operation. Also you can see, on the ICs Isense pin the effect of the injected slope, the bit where the mosfet turns off. The sense voltage does not fall to zero and there is an additional ramp bit on it.
The dirty sheet block diagram does show an internal resistor that might isolate such an effect and the fact it works here does depend on how good the behavioural model supplied with LTspice is but it does look a bit suspicious....
Beyond that it looks like you might have over/under cooked the compensation. The following is a quick slapped together guess as to a method for doing it. I'd verify it proper like but....
The dirty sheet says the active range of the voltage error amplifier is 1V to 2.5V or a range of 1.5V. That gets converted throught the internal current source thing to 0-100mV. An attenuation of 0.0667. With your 8mR current sense resistor the primary side sensitivity is
8=2E33A/V (0.667/0.008). Your transformer has a turns ratio of 6.75 so that appears on the secondary side as 1.23A/V.
The ESR zero of your output capacitor occurs at about 10KHz. Above that frequency with an ESR of 15mR then the gain is flat at 1.23x15m or 0.01845. With a 15K feedback resistor to the error amplifier and a
1M5 feedback resistor around the error amplifier your gain is 100 so the loop gain at and above that point is 1.845........ and ideally, the loop is not going to cross over.
It will because something else will run out of steam but it's a bit 'wet finger in the air'. Oh... that will be it. The gain bandwidth of the error amplifier is quoted as being 3MHz and you are asking for a gain of 100 from it so it runs out of steam at 30KHz.
Going back to it supposedly, with appropriate slope compensation, the internal current loop in the circuit crosses over at Fs/2piD where Fs is the switching frequency and D is the operating duty cycle. In your circuit that will be about 20KHz. You will want to cross over the voltage loop below this frequency to avoid additional phase shifts so pick 10KHz.
At the moment you are being forced to use large value resistors in your error amp feedback network so parallel your 15K resistor with a
1K5/10n series network. That has a breakpoint at the ESR zero and will make the loop response flat from 1KHz to 10KHz..... then it will continue being flat due to the ESR zero.
To get back to an overall first order response you can use a simple capacitor around your error amplifier. To cross over at 10KHz the gain needs to be 54. Rin is approximately 1K5 so the capacitor needs to have an impedance of 81K at 10KHz so its value is about 220p.
With that overall first order response your start up is going to be nice and lazy and you won't have problems with overshoot and other nasties, fingers crossed.
You've mentioned concern about your filter capacitor lifetime and quoted a 7000Hr/105C figure.
That will be the manufacturers quoted headline figure at maximum RMS ripple current. I can't quite remember but 1000uF low ESR devices are probably rated at something like 2A ripple current. If you operate at lower values then lifetime multipliers come into effect.
In your converter the ripple current is something like 300mA peak to peak and triangular. The RMS value is something like Ipk/root3.....
150mA/root3 or 86mA. As such, when you apply the appropriate multiplier you will find out that your capacitor lifetime becomes something silly like 100,000Hrs..... You'll have to check my sums and check with the manufacturer, that was a quick stab.
I've had a fiddle with your circuit, and messed it up a bit, sorry. The output capacitor has been changed to 100u with 150m ESR. That meant I had to make the feedback capacitor 10 times bigger. The required gain from the VEA drops to 5.4 so it's going to be happier, that gain-bandwidth thing.
Don't rely on that 100u cap, I got bored with slow start up as well and spice is nice. I've had to fudge a couple of other bits to make it go, as opposed to stall. Mainly coupling coefficients and series resistance for the inductors. The IC gets its bootstrap up in about
200uS and the supply comes into regulation at about 2mS, with no overshoot. At 3mS it starts getting hit with a 500mA-1A transient.
Output ripple, due to inductor ripple current, is 60mV peak to peak...... wait a while. 500mA to 1A load step is, well, have a look yourself. The excursion is about 100mV and the loop responds in about
60uS without a hint of ringing ultimately dragging its arse back up to the set value in about 500uS. Going from 1A down to 500mA does, more or less, the same.
It's a dude. (but I'm sure it could be broken elsewhere.... Damn, I'm not mean to say things like that)
DNA
Version 4 SHEET 1 2104 772 WIRE -704 -224 -832 -224 WIRE -240 -224 -704 -224 WIRE 400 -224 -240 -224 WIRE 608 -224 400 -224 WIRE 736 -224 608 -224 WIRE 864 -224 736 -224 WIRE 992 -224 944 -224 WIRE 1104 -224 1056 -224 WIRE 1152 -224 1104 -224 WIRE 1280 -224 1232 -224 WIRE 1360 -224 1280 -224 WIRE 1488 -224 1360 -224 WIRE 1616 -224 1488 -224 WIRE 1648 -224 1616 -224 WIRE 400 -192 400 -224 WIRE 608 -192 608 -224 WIRE 736 -192 736 -224 WIRE 864 -192 864 -224 WIRE 944 -192 944 -224 WIRE 1104 -176 1104 -224 WIRE 1280 -160 1280 -224 WIRE 1360 -160 1360 -224 WIRE 608 -80 608 -112 WIRE -240 -48 -240 -224 WIRE -208 -48 -240 -48 WIRE -112 -48 -128 -48 WIRE -80 -48 -112 -48 WIRE 16 -48 -16 -48 WIRE 48 -48 16 -48 WIRE 128 -48 112 -48 WIRE 1360 -32 1360 -80 WIRE -704 16 -704 -224 WIRE 608 32 608 -16 WIRE 608 32 576 32 WIRE 736 32 736 -112 WIRE 736 32 608 32 WIRE -240 48 -240 -48 WIRE -112 48 -112 -48 WIRE 16 48 16 -48 WIRE 1280 64 1280 -80 WIRE 1360 64 1360 32 WIRE 1360 64 1280 64 WIRE 736 112 736 32 WIRE -704 128 -704 96 WIRE -304 128 -704 128 WIRE 400 128 400 -112 WIRE 400 128 64 128 WIRE 608 128 576 128 WIRE -512 192 -608 192 WIRE -304 192 -512 192 WIRE 160 192 64 192 WIRE 608 192 608 128 WIRE 608 192 240 192 WIRE 688 192 608 192 WIRE -512 256 -512 192 WIRE -464 256 -512 256 WIRE -304 256 -384 256 WIRE 240 256 64 256 WIRE 272 256 240 256 WIRE 736 256 736 208 WIRE 736 256 352 256 WIRE -304 320 -512 320 WIRE 112 320 64 320 WIRE 208 320 176 320 WIRE 736 320 736 256 WIRE 736 320 576 320 WIRE 1488 336 1488 -224 WIRE -304 384 -400 384 WIRE 208 384 208 320 WIRE 208 384 64 384 WIRE 1360 384 1360 64 WIRE 1360 384 208 384 WIRE -832 480 -832 -224 WIRE -704 480 -704 128 WIRE -608 480 -608 192 WIRE -512 480 -512 320 WIRE -400 480 -400 384 WIRE 400 480 400 128 WIRE 736 480 736 320 WIRE 1360 480 1360 384 WIRE 1488 480 1488 400 WIRE 240 496 240 256 WIRE 864 496 864 -112 WIRE 1616 496 1616 -224 WIRE -832 592 -832 560 WIRE -704 592 -704 560 WIRE -704 592 -832 592 WIRE -608 592 -608 544 WIRE -608 592 -704 592 WIRE -512 592 -512 560 WIRE -512 592 -608 592 WIRE -400 592 -400 560 WIRE -400 592 -512 592 WIRE -192 592 -192 448 WIRE -192 592 -400 592 WIRE -48 592 -48 448 WIRE -48 592 -192 592 WIRE 240 592 240 560 WIRE 240 592 -48 592 WIRE 400 592 400 560 WIRE 400 592 240 592 WIRE 736 592 736 560 WIRE 736 592 400 592 WIRE 864 592 864 560 WIRE 864 592 736 592 WIRE 944 592 944 -112 WIRE 944 592 864 592 WIRE 1104 592 1104 -112 WIRE 1104 592 944 592 WIRE 1360 592 1360 560 WIRE 1360 592 1104 592 WIRE 1488 592 1488 560 WIRE 1488 592 1360 592 WIRE 1616 592 1616 560 WIRE 1616 592 1488 592 WIRE -832 624 -832 592 FLAG 128 -48 0 FLAG 1648 -224 OUT IOPIN 1648 -224 Out FLAG 576 320 ISNS IOPIN 576 320 Out FLAG -832 624 0 FLAG 576 32 DRAIN IOPIN 576 32 Out FLAG 576 128 GATE IOPIN 576 128 Out SYMBOL PowerProducts\\\\LT1950 -112 256 R0 SYMATTR InstName U1 SYMBOL res 1344 464 R0 WINDOW 0 38 50 Left 0 WINDOW 3 36 70 Left 0 SYMATTR InstName R1 SYMATTR Value 1.37k SYMBOL cap 176 304 R90 WINDOW 0 0 32 VBottom 0 WINDOW 3 32 32 VTop 0 SYMATTR InstName C1 SYMATTR Value 2n2 SYMBOL nmos 688 112 R0 SYMATTR InstName Q1 SYMATTR Value IRF6635 SYMBOL res 720 464 R0 SYMATTR InstName R3 SYMATTR Value 8m SYMBOL ind2 720 -208 R0 WINDOW 0 -30 45 Right 0 WINDOW 3 -23 72 Right 0 SYMATTR InstName LPRI SYMATTR Value 94.4=B5 SYMATTR Type ind SYMATTR SpiceLine Rser=3D1m SYMBOL ind2 960 -208 M0 WINDOW 0 -2 46 Right 0 WINDOW 3 -3 73 Right 0 SYMATTR InstName LSEC SYMATTR Value 4.3m SYMATTR Type ind SYMATTR SpiceLine Rser=3D1m SYMBOL voltage -832 464 R0 WINDOW 0 68 46 Bottom 0 WINDOW 3 61 48 Top 0 WINDOW 123 32 56 VTop 0 WINDOW 39 32 56 VTop 0 SYMATTR InstName VDD SYMATTR Value 3V SYMBOL ind 1136 -208 R270 WINDOW 0 32 56 VTop 0 WINDOW 3 5 56 VBottom 0 SYMATTR InstName LFILT SYMATTR Value 100=B5 SYMATTR SpiceLine Rser=3D10m SYMBOL cap 1472 336 R0 WINDOW 0 39 18 Left 0 WINDOW 3 43 42 Left 0 SYMATTR InstName CFILT SYMATTR Value 100=B5 SYMBOL res 1264 -176 R0 SYMATTR InstName R4 SYMATTR Value 15k SYMBOL res -528 464 R0 SYMATTR InstName R5 SYMATTR Value 510k SYMBOL ind2 -224 -32 R270 WINDOW 0 32 56 VTop 0 WINDOW 3 5 56 VBottom 0 SYMATTR InstName L1 SYMATTR Value 4=B57 SYMATTR SpiceLine Rser=3D1R SYMATTR Type ind SYMBOL cap 48 -32 R270 WINDOW 0 32 32 VTop 0 WINDOW 3 0 32 VBottom 0 SYMATTR InstName C3 SYMATTR Value 1=B5 SYMBOL cap -624 480 R0 WINDOW 0 43 25 Left 0 WINDOW 3 42 52 Left 0 SYMATTR InstName C5 SYMATTR Value 100n SYMATTR SpiceLine Rser=3D.2 SYMBOL res -368 240 R90 WINDOW 0 -3 55 VBottom 0 WINDOW 3 36 54 VTop 0 SYMATTR InstName R6 SYMATTR Value 100K SYMBOL res 384 -208 R0 WINDOW 0 37 52 Left 0 SYMATTR InstName R7 SYMATTR Value 1.69k SYMBOL res 384 464 R0 WINDOW 0 39 47 Left 0 SYMATTR InstName R8 SYMATTR Value 1.10k SYMBOL res -720 0 R0 WINDOW 0 44 48 Left 0 WINDOW 3 49 71 Left 0 SYMATTR InstName R9 SYMATTR Value 1.40k SYMBOL res -720 464 R0 SYMATTR InstName R10 SYMATTR Value 1.10k SYMBOL res -416 464 R0 SYMATTR InstName R11 SYMATTR Value 75k SYMBOL ind2 880 -96 R180 WINDOW 0 34 67 Left 0 WINDOW 3 66 38 Left 0 SYMATTR InstName LCLMP SYMATTR Value 5.9=B5 SYMATTR Type ind SYMATTR SpiceLine Rser=3D1m SYMBOL res 256 272 R270 WINDOW 0 37 57 VTop 0 WINDOW 3 0 56 VBottom 0 SYMATTR InstName R12 SYMATTR Value 1K SYMBOL cap 224 496 R0 WINDOW 0 49 17 Left 0 WINDOW 3 51 38 Left 0 SYMATTR InstName C4 SYMATTR Value 220p SYMBOL res 1344 -176 R0 SYMATTR InstName R2 SYMATTR Value 1K5 SYMBOL cap 1344 -32 R0 SYMATTR InstName C6 SYMATTR Value 10n SYMBOL diode 880 560 R180 WINDOW 0 -34 40 Left 0 WINDOW 3 -41 16 Left 0 SYMATTR InstName D2 SYMATTR Value DID SYMBOL diode 1120 -112 R180 WINDOW 0 -37 47 Left 0 WINDOW 3 -44 22 Left 0 SYMATTR InstName D4 SYMATTR Value DID SYMBOL diode 992 -208 R270 WINDOW 0 32 32 VTop 0 WINDOW 3 0 32 VBottom 0 SYMATTR InstName D3 SYMATTR Value DID SYMBOL diode -80 -32 R270 WINDOW 0 32 32 VTop 0 WINDOW 3 0 32 VBottom 0 SYMATTR InstName D1 SYMATTR Value DID SYMBOL res 256 176 R90 WINDOW 0 0 56 VBottom 0 WINDOW 3 32 56 VTop 0 SYMATTR InstName R13 SYMATTR Value 2R2 SYMBOL res 1472 464 R0 SYMATTR InstName RESR SYMATTR Value 150m SYMBOL load 1600 496 R0 WINDOW 0 40 31 Left 0 WINDOW 3 48 56 Invisible 0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 SYMATTR InstName I1 SYMATTR Value PULSE(500mA 1A 3m 10n 10n 1m 2m) SYMBOL res 592 -208 R0 WINDOW 0 -46 52 Left 0 WINDOW 3 -46 75 Left 0 SYMATTR InstName R14 SYMATTR Value 6R8 SYMBOL cap 624 -16 R180 WINDOW 0 50 48 Left 0 WINDOW 3 42 20 Left 0 SYMATTR InstName C2 SYMATTR Value 4n7 TEXT -776 656 Left 0 !.tran 0 10m 1u uic TEXT 664 -256 Left 0 !K1 LPRI LCLMP LSEC 0.999 TEXT -776 632 Left 0 !.MODEL DID D(RON=3D10m) LINE Normal 464 -480 464 -480 LINE Normal 464 -480 464 -480 LINE Normal 464 -480 464 -480 LINE Normal 464 -480 464 -480 LINE Normal 464 -480 464 -480
Ha... yeah, you would think that. So would I. :-) That was a mistake in labeling.
"I've had a little play with your LTspice model. One thing that struck me as odd is the value of filtering components used on the Isense pin.
47R/10n seem exceptionally small/large but that's what they show on the application note."
Agreed, I used those values because that's what Linear did and I hadn't gotten around to playing with them yet. My guess was that they were trying to keep R12 as small as possible (to avoid noise influencing the Isense pin) while C4 remained a "reasonable" size. Prior to go to 47R/10n I had actually used an
0805 10uF cap right across R3 (8m) and it worked as well, although even I recognize that as horribly crude.
"Try making R6 100K in your model (sort of gives the minimum the IC will give) and doing a transient analysis from 500u-600u and look at Sense. Look at the waveform at the Isense pin of the IC after you have filtered it. It's almost but not quite classic subharmonic oscillation......."
Gotcha.
"Now change R12 to 1K and C4 to 220p, make it's ESR zero. Leave R6 as
100K. Run the analysis again. Now you get stable operation."
Given that 1k/220p actually has a shorter time constant than 47R/10n, what you're saying is that *too much* filtering at the Isense pin can actually cause sub-harmonic oscillation -- correct?
"Also you can see, on the ICs Isense pin the effect of the injected slope, the bit where the mosfet turns off. The sense voltage does not fall to zero and there is an additional ramp bit on it."
Neat, thanks for pointing that out.
"The dirty sheet says the active range of the voltage error amplifier is 1V to 2.5V or a range of 1.5V. That gets converted throught the internal current source thing to 0-100mV. An attenuation of 0.0667. With your 8mR current sense resistor the primary side sensitivity is
8.33A/V (0.667/0.008). Your transformer has a turns ratio of 6.75 so that appears on the secondary side as 1.23A/V."
Yep, that's -- roughly -- how I calculated it as well. Sine I believe you're a lot more experienced with switchers than I am, I'm glad you're doing it the same way.
"The ESR zero of your output capacitor occurs at about 10KHz. Above that frequency with an ESR of 15mR then the gain is flat at 1.23x15m or 0.01845. With a 15K feedback resistor to the error amplifier and a
1M5 feedback resistor around the error amplifier your gain is 100 so the loop gain at and above that point is 1.845........ and ideally, the loop is not going to cross over."
I did some SPICE simulations while playing with the feedback network. I initially wanted to crossover at 20kHz, but the GBW of the error amp doesn't have nearly enough gain to make that happen so I ended up moving the crossover to 1kHz instead.
"At the moment you are being forced to use large value resistors in your error amp feedback network so parallel your 15K resistor with a
1K5/10n series network. That has a breakpoint at the ESR zero and will make the loop response flat from 1KHz to 10KHz..... then it will continue being flat due to the ESR zero."
Ah, now I understand why they do that in the example circuits in the data sheet. Great, that helps a lot.
To get back to an overall first order response you can use a simple capacitor around your error amplifier. To cross over at 10KHz the gain needs to be 54. Rin is approximately 1K5 so the capacitor needs to have an impedance of 81K at 10KHz so its value is about 220p.
With that overall first order response your start up is going to be nice and lazy and you won't have problems with overshoot and other nasties, fingers crossed.
"You've mentioned concern about your filter capacitor lifetime and quoted a 7000Hr/105C figure.
That will be the manufacturers quoted headline figure at maximum RMS ripple current. I can't quite remember but 1000uF low ESR devices are probably rated at something like 2A ripple current. If you operate at lower values then lifetime multipliers come into effect."
The cap I'm using (Panasonic FM series) is actually rated a 3.19A ripple current. The data sheet for my cap --
formatting link
-- doesn't mention life time variations. Do you know of a better paper than, say, this one -->
formatting link
...for a discussion of the subject?
Thanks a lot for your help, Genome -- this will definitely let me improve the design's performance.
I figured out why they used such a low-value RC filter on the Isense pin... according to the data sheet, that pin sinks ~ 170uA, and with a 1k resistor that gets you 170mV -- well above the 125mV point where the IC will just sit there and do absolutely nothing!
The PCB I have draws more like 155uA. Annoyingly, the model in LTspice only draws ~20uA -- hence the reason the simulation still worked for you.
Why is it that the supply does 2V ripple without it? Ok, my last one generates only 20mA at 80V but it lived with a whopping 0.1uF. It also has a 1uF for noise feedback reasons but for stability tests I took it out, not much ripple increase. Somehow I don't understand why 1A needs
1000uF. I have never seen anything that high on a switcher, even really on the big ones. Is something in there running out of breath?
In retrospect I think it's going unstable when I just drop the capacitance from 1mF to 100uF without upping the ESR. Sorry about that; I didn't look at it closely enough when I wrote the 2V number yesterday.
In a 100kHz, 60% DC sticher with 100uF, if you're pulling 1A out of the cap for the 4us that the switch is off (and ignoring the output inductor), you still get dV = dt * I/C = 10us*0.4 * 1/100uF = 40mV of ripple. I suppose that's acceptable?
It doesn't need it for 1A/1000uF, it was driven by the 7000 hour life span (and 15mOhm ESR, although 40mV is far more than the ripple caused by the ESR, so that would be moot too). As Genome has pointed out to me, though, since it's being used as far less than its rated ripple current of 3.19A, the actual life span should be much, much longer anyway... and hence dropping to a shorter life / lower capacitance part should still be perfectly acceptable.
It does look a little odd to have physically small capacitors surrounded by physically larger inductors and trasnformers, though (a 3019 pot core is ~1" in diameter...). :-)
That instability would concern me. Might be time to re-visit the loop behavior.
On mine 40mV would have always been well within spec. Sensitive stuff down the road gets a "scrubber", for example an emitter follower with its collector hung onto the rail and the base connected to the rail via a resistor, plus a 1uF-10uF cap from base to ground. I think some people call that capacitor multiplier because to the connected circuit it looks like a huge cap was hanging on the rail. If you do that make sure there are back discharge diodes in case someone or soemthing shorts out the rail.
The life span is usually much longer. Heck, I've got fully functional electrolytics in tube radios that are older than I am. Since you mentioned high temps, electrolytics suffer quite a bit under those. They can dry out over time. There are the notable exceptions: In our Hammond organ they placed it right next to the freaking hot rectifier tube and it's still fine since 1961, no hum.
Yeah, I could probably get away with a 2616 (~26mm diameter) and machine-wound perhaps a 2213 (~22mm), but I like having a little extra room while prototyping; perhaps I'll try the 2616 once it's fully working.
Do you know of a source for enameled wire "tape" (i.e., "flat wire") in smallish (hundreds of feet) quantities? I'm thinking it'd be a lot easier to wind than the two 16 ga. wires in parallel for the primary that I'm using now
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.