Slow fade-in circuit

• it turns out LED light bulbs can work on 170VDC, the n I have cobbled together a circuit I think might work . Please take a look at the link below and see if any of you can find any issues with this.

C supply is provided by an MP9488GS-P Buck regulator. U1 integrates the 0V differential signal provided by R1 - R4 with a time constant determined by R5 and C1. C2 provides a delay, keeping the output of U1 high un til C2 is charged to 6V by R4 (~2ms). Once C2 is char ged, the output of U1 slowly decreases from ~10V to ~2 V over a period of 5 seconds. U2 inverts the slope of the signal from U1 and produces a necessary offset. Its output changes from 4.5V to 12V over the same 5 se cond time period. The 555 timer X1 operates as a shor t pulse generator, creating pulses a few microseconds wide approximately every 8 milliseconds. Transistor Q

Thanks for sharing. I figure the exact PWM frequency does not matter so it can th en be simplified to a self oscillating comparator pwm. By limiting maximum light to just below 100% duty t hen the modulator supply voltage can be derived from across the fet. By using micro-power devices then s upply current can be kept in the tens of microamps and dissipation in the dropper resistor kept low.

Play with R1, R2, R3, R4 ratios to change the fade ramp curve from exponential thru linear to log. The values I show give a exp curve with a slow start then faster rise that I guessed is more eye frie ndly. Not shown in the schematic but I think R4 shou ld be shunted by a small

The three ICs are avail able in SOT-23 packs and whole circuit could be buil t very small.

piglet

ed together a circuit I think might work. Please take a look at the link be low and see if any of you can find any issues with this.

rates the 0V differential signal provided by R1 - R4 with a time constant d etermined by R5 and C1. C2 provides a delay, keeping the output of U1 high until C2 is charged to 6V by R4 (~2ms). Once C2 is charged, the output of U1 slowly decreases from ~10V to ~2V over a period of 5 seconds. U2 inver ts the slope of the signal from U1 and produces a necessary offset. Its ou tput changes from 4.5V to 12V over the same 5 second time period. The 555 timer X1 operates as a short pulse generator, creating pulses a few microse conds wide approximately every 8 milliseconds. Transistor Q1 prevents X1 f rom generating any pulses until C2 is charged to ~1.4V by R8, allowing time for the 12V buck converter to settle before X1 begins operation. X2 is co nfigured as a variable pulse width monostable timer triggered by Q2 from th e short pulses produced by X1. The output of X2 is held high by the intern al flip flop until the voltage across C6 rises to be greater than Control v oltage, which is the output of U2. Initially, the output of U2 is within 1 /2 volt of the Trigger voltage, so the voltage across C6 reaches this value very quickly. As time passes, the Control voltage increases, causing the output of X2 to remain high longer and longer, increasing the duty cycle to near 100%. The output of X2 drives Q3, causing M1 to turn on whenever X2 is high.

Nice. Thx. PWMing the fet keeps it cool.

George H.

Well..

If you look at the graph on page 3 of LED string current I think through a clever switching arrangement they've got good efficiency and power factor while keeping flicker that would be objectionable (noticeable to the human eye) to a minimum

Yes, but kind of depends ... the FET has to be chosen to enhance sufficiently with 6V gate drive. HV fets seem to have higher Vgs requirements in general. It is a design choice battle between Vgs and the max Vdd for the chips - from 5.5V to 7V depending how close to abs max one feels comfortable :(

piglet

OK, why not run it at higher voltage? Burn more power Scotty! Would a lnd150 current source feeding the zener make any sense?

George H.

That is *VERY* clever!

Why did you choose a mylar capacitor for C1? Why not ceramic?

I put the circuit into a simulator, and according to it, you are very corre ct about needing a cap around R4 (no big surprise, there). It rang like a bell. A 47pf cap quiets it nicely. The value of C1 you gave seems way too low, though. According to the sim, the driver reaches essentially 100% du ty cycle in considerably less than 2 seconds. Going with C1 = 2.2uF and R1 = 100M, the profile looks pretty nice. I'm not sure how far I can pus h R1. The datasheet on the MCP6231 series isn't all that thorough. Do yo u know what the input impedance of the MCP6231 op amps is? Are they JFET i nputs? Newark has the MCP6231UT-E/LT in stock for $0.20 each. Those shoul d fit the bill.

The power on reset chip, op amp and comparator I picked have 5.5V recommended and 7V absolute max rating. There must be higher voltage micro=power devices around but I did not spend time looking for them.

Since the lamps are low power a FET with very low Rds-on is not needed and so getting by with 5.5-6V gate drive seemed OK.

You are right that using a LND140 or BSS126 as dropper is better electrically - but a resistor is cheaper and probably good enough.

piglet

Aluminum and tantalum are way too leaky, I would normally avoid larger value ceramic in precision timing because of their voltage coefficient. However you are right - this case does not need high precision and the voltage coefficient of X7R and similar dielectrics actually work in favor of an exponential fade ramp, although at these low voltages the effect is not large. So please go ahead and use ceramic :)

piglet

Well done for simulating the circuit! Sorry I didn't optimize parts values and the values are mostly estimates. You will need to check and refine if actually building.

The MCP6231 is CMOS so input currents are pico-ampere level. Datasheet typ 1pA room temp, 20pA at 85degC. PCB leakage will likely dominate. The open drain on the MCP121 is likely to be a leakage limit too.

If you go to very large C1 like 10 or 22uF then I wonder if that is too much strain on the MCP121 and a resistor in series with output of MCP121 may be good, perhaps 330 ohm?

piglet

I had already bought the mylar cap. It's only $0.40. You are certainly correct Aluminum and Tantalum are way too leaky at these impedances.

orrect about needing a cap around R4 (no big surprise, there). It rang like a bell. A 47pf cap quiets it nicely. The value of C1 you gave seems way to o low, though. According to the sim, the driver reaches essentially 100% du ty cycle in considerably less than 2 seconds. Going with C1 = 2.2uF and R

Thanks!

No worries. I just thought perhaps you were expecting the values you used to yield 5 - 10 seconds.

That should be good. One pA at 0.1Gohm is less than a millivolt error. Th at should not be significant in this case. I am happy with +/- 20% for thi s application.

I used a 2N2222 to simulate the open drain of the MCP121 in the sim. It d idn't seem to cause much of an issue. (My sim doesn't have a model for the MCP121.)

I expect not. 2.2uF + 100M gets me very close in the sim. I don't expect to be an order of magnitude off, and with leakage on the order of a few pA, I should be able to increase R1 to nearly a Gigaohm.

I suppose if it becomes necessary. I doubt it will.

We haven't mentioned RFI yet but switching at a few hundred Hz into the building wiring could be a potent source of radio interference. I don't know what your site needs are but it may be prudent to allow for some inductance and filter caps needed somewhere.

piglet

How are you going to protect the PCB from moisture and accumulated conductive cruft? The leakage into the op-amp isn't likely to be your limit, not for very long anyhow.

I'd go with lower impedance and bigger capacitors.

CH

At less than a 1/4 ampere load, I'm not expecting much of a problem, especi ally not up around 2 - 5 GHz, which is where most of the electronics in the room operates. The impedance of the MOSFET I chose is 0.86 Ohms - perhaps I could choose one with a higher impedance and thus a lower slew rate, res ulting in lower frequency harmonics. The leads will be about 1/4 of a mete r, giving them an inductance of less than 500 microHenries - pretty insigni ficant at moderate RF frequencies, so a coil would not hurt. That said, a filter cap across the MOSFET of about 200 pF or so should quiet the noise w ell enough, I think, if it is even needed at all.

BTW, although the detector and the amp are rated at 7V, the comparator is l isted at 5.5V max, so I went with a 5.6V Zener, rather than 6V. It's still pretty tight, but it should give a little more margin, just in case. I do n't want to have to be replacing these every couple of years.

t to be an order of magnitude off, and with leakage on the order of a few p A, I should be able to increase R1 to nearly a Gigaohm.

That gets into other issues, like higher operating currents - a real proble m for this tiny power supply.

I intend to coat the high impedance section with lacquer. Corona Dope shou ld work. With a dielectric strength of around 20 KV / mm, I could use a la yer of Epoxy, too. I will keep the high impedance components well isolated and separated. Encase them in a block of epoxy, and the environment is no longer an issue. There are only three of them, after all. In any case, 1

In 1971 my pal who was also an EE student, designed and built a hand proximity dimmer using up/downcounters with hand wave duration step size in 3 steps or continuous with 256 steps in a 2 second. The counter value was then converted to phase angle in a non-linear fashion to linearize intensity for controlling the Triac. If you are driving only LED's and the dimmer fails due to leakage currents, the std fix was to add a 8W bulb which acts as a ~60W gradient load at low levels due to thermal resistance to shunt leakage currents. It would detect an rapid hand wave at 1ft away and like the Star Trek door effect. Or nowadays the Walmart door experience. This is all trivial Logic design and used a small hand-wound pulse transformer for isolation to the touchpad that acted as an antenna. The signals used for proximity were in the xx kHz range.

But rather than an inverter a phase controlled Triac can work well if the AC LED bulb is dimmable.

Tony.