Lesson on autodidactism

I have probably talked about this earlier, or at least talked about how I learned electronics design on my own.

But I wanted to revisit the topic. This time in the form of automobile diagnosis.

About three years ago I noticed that my turbo isn’t boosting as it should. So I queried around to get an offer to have it fixed. But the only offers I received were “we can look into it”. And because no one offered me any concrete paths, I knew it would cost a lot of money to have it checked.

So what I ended up doing was I began to seek knowledge what could it be. Of course this was extremely difficult because I had no knowledge even of the words! The words that describe parts and systems. So how do you began? Well I began trying to find the solution for the problem. As in: why turbo isn’t boosting. Literally, Google: “why turbo isn’t boosting.”

That led me somewhere, but couldn’t answer me because I couldn’t understand the role of each of the parts in the system. None of it made any sense to me. I couldn’t use reasoning because I couldn’t understand the parts and the system as a whole.

But slowly, slowly the research went into direction of simply wanting to understand the turbo system. What parts is it made of.

And then, after understanding each part in the system, and comparing that to my vehicle; what it had; what it did not have; I could began to parse together a diagnosis. And that is another important thing: what my vehicle has, and what it does not have. Because not only is “turbo system” an abstract concept of forced induction, but there are variations within.

So in my mind I now have understanding of what happens at each point when the pedal is pressed.

And I was able to reason, that the problem could be the wastegate control solenoid. And I asked mechanics to check it. And it was leaky indeed.

But the problem remained. So I learned there is something called symposer in the system, that may also become leaky. So I bypassed that system. But the problem still remained.

So now I thought it could only be the bypass valve or if not, then it would have to be the turbo itself, which I didn’t believe it would be as it boosted yes, but not the way it was supposed to; so the turbo unit itself did produce forced induction, but only at very high revs (airflow).

I went to a specialist whom agreed to change the component. My usual mechanics wouldn’t touch it. And I understand why: the placement is “tricky”.

The specialist had his doubts because he had never seen that part fail. But after more than 2 hours of fiddling, the car flew down the road exactly like it did some 60 000km ago.

So. It is possible to do diagnosis on your own. Is it smart? In a direct sense of the word perhaps yes but, did it come any cheaper? That is debatable. I ended up changing one unnecessary part. But one of two parts which were faulty was replaced with aftermarket upgrade. And the one part which wasn’t faulty, too, was replaced with aftermarket upgraded replacement.

So in doing the diagnosis myself, I was able to improve the vehicle performance; had it been diagnosed not by me, I would have had to have paid for 1) the diagnosis and 2) the replacement of the parts, while the car would have either been unusable waiting for the parts, or I would have had to pay more, as it would have required more visits to the shop, and, it may have not gotten the upgraded parts.

And now I understand the turbo system. And many more automotive systems. I have learned the ins and outs of my own power system. And I can do more diagnostics. And keep the vehicle in top performing condition.

So yes, autodictatism is perfectly valid method. And can be applied to automotive diagnosis with enough patients.

Cost optimized light profile MK1

The first panel (1-5) won’t be exactly like this, because row 3 is violet (from old light profile) and row 1 is Super Cool White. But this is cost optimized version of the earlier one, where violet and ultraviolet have been replaced with Super Cool White, and more blue has been added.

I got one trick in my sleeve to top this one up, but it will take some time to get it made, since it isn’t the first priority on the list. Although it should increase the overall efficiency of the plant. Take that as a hint. Nothing more will be reveleaded until it is in production.

Also this requires only two voltages so it is a simpler setup.

New light profile for MK1 (revisited)

Here is the new light profile which was created because the use of Super Cool White is clear mistake: it burns too much power into green which will be wasted.

Here the Super Cool White has been replaced with Bright Blue and Royal Blue, and half of the Violet have been replaced with Ultraviolet, and the different colours have been re-arranged so that the blue colours are more on the outer edges, and the red colours at the center of the panel. This is because there is a need to have concentrated powerful red spectrum but blue spectrum can be more scattered.

Also, it isn’t the LED cooling that requires attention, but more the cooling of power electronics. At least with this 12V setup. At 24V the efficiency should be better and losses smaller. Input power is 148W and output power 109W, so only 73% efficiency and 39W dissipated.

Revisiting the changes

It seems the issue of green spectrum isn’t as simple as I had thought:


And it would require extensive calculations on effencies to come up with the most effective spectrum for the light, considering both the chlorophyll a and b, and also the carotenoids and other aspects of plant physiology. It would be fun to do all that but I don’t think I have enough stamina to do all that work. So perhaps for now I might go with the original profile and keep the wider band of spectrum to fill the caps that would be left should I try to early optimize the spectrum too far based on what seems at this point to be guess work.

But if the document is too long to read, then this summarizes it pretty well:

The literature and our present examinations indicate that the intra-leaf light absorption profi le is in most cases steeper than the photosynthetic capacity profi le. In strong white light, therefore, the quantum yield of photosynthesis would be lower in the upper chloroplasts, located near the illuminated surface, than that in the lower chloroplasts. Because green light can penetrate further into the leaf than red or blue light, in strong white light, any additional green light absorbed by the lower chloroplasts would increase leaf photosynthesis to a greater extent than would additional red or blue light.


Partly assembled light panel and cooling tests

Doing more tests with the new 20 LED batch; now with new aluminium profiles and 30 LEDs total:

On top is Super Cool White, in the middle Full Spectrum Par, and at the bottom Violet, which in this photo looks purple, because it requires that much more voltage, and hence the FSP is dominating the color with Super Cool White.

A low-power 120mm computer fan is enough to keep this setup at barely warm temperature. Super Cool White and FSP were running at their full rated power, but the Violet was lagging behind, and two whole rows are missing, but still it should stay cool, and if not, then 140mm and a bit higher powered one will be enough. Or two 120mm, which would keep the panel not much above the ambient.

Pretty easy to tell the lack of screen, because once you spend a minute with the light and return, everything looks very green. Which is good because we don’t want to use power, money or footprint to produce green light.

Calculating component values

Current sense

RADJ = 0.25V / 0.75A = 0.35714285714285714285 ≅ 0.36Ω

P(RADJ) = (0.75^2 * 0.36) = 200mW

Duty cycle

Dnom = (40 – 0.9 * 26 + 0.6) / (40 + 0.6) = 0.42

Dmax = (40 – 0.9 * 26 + 0.6) / (40 + 0.6) = 0.42

Dmin = (40 – 0.9 * 26 + 0.6) / (40 + 0.6) = 0.42


IIN_RMS_max = (40 * 0.75) / (0.9 * 26) = 1.28

IIN_RMS_nom = (40 * 0.75) / (0.9 * 26) = 1.28

IIN_RMS_min = (40 * 0.75) / (0.9 * 26) = 1.28


Iin_pp_nom = 0.4 * 1.28 = 0.5

L = (26 * 0.42 * 2.5) / 0.5 ≅ 54.6µH

(T = 1 / f = 1 / 400000)

Using ≥56µH

Iin_PP = (26 * 0.42 * 2.5) / 56 = 0.49

IIN_AVE_max = sqrt(1.282 – ((0.492) / 12)) = sqrt(1.64 – 0.02) = 1.27

IL_PK_max = 1.27 + 0.5 * 0.49 = 1.52

PINDUCTOR = 1.282 * 0.0965 = 0.158 = 158mW

For SDR2207-560KL

In addition to copper loss, an iron-core coil (inductor) has two iron losses. These are called HYSTERESIS LOSS and EDDY-CURRENT LOSS. Hysteresis loss is due to power that is consumed in reversing the magnetic field of the inductor core each time the direction of current in the inductor changes.


Current limit and slope compensation

IL_pklimit = 1.2 * 1.52 = 1.824


I don’t know how they get 179mΩ because I get 193mΩ no matter what I do. So I have no choice but to trust my calculation.

RCS = 0.45/((((40-26)*(0.5)) / ((56*10^-6) * 400000)) + 1.9) = 203mΩ

Using 200mΩ

RSLC = ((40-26) * (150*10^-3)) / ((47*10^-6) * (250*10^-6) * 400000) = 446Ω

Using 453Ω

Iin_actuallimit = (0.45 – (250*10^-6) * 446 * 0.42) / 0.2 = 2.02

Make sure the inductor will not sature at the actual current limit [2.02A]

VIS_PIN = (250*10^-6)*0.49*446+2.02*(0.2) = 0.45863500000000000000

I(RCS_RMS) = sqrt(0.42*((1.27^2)+((0.49^2) / 12))) = 0.83A

P(RCS) = 0.832 * 0.2 = 0.14W

Using ¼W

Output capacitor

ILED = 0.1 * 0.75 = 75mA

LED equivalent ac resistance isn’t available so using 100mΩ

Cout = (0.75 * 0.42 * (2.5*10^-6)) / (0.075 * (0.36 + 1)) = 7.72µF

Using 2 x 10µF ceramic in parallel

Input capacitor

CIN = 0.5 / (8 * 0.05 * 400000) = 3.125µF

Using 2 x 10µF ceramic in parallel

MOSFET selection

Using FQP50N06 [for rough estimates]

IFET_RMS = sqrt(0.42*((1.27^2)+((0.49^2)/12))) = 0.8281434054558425

RDSon(25C) = 0.02Ω

RDSon(125C) = 0.02 * (1.007^(125-25)) = 0.04Ω

PFET_COND = (0.828^2)*(40*10^-3) = 27mW

ttransition = 32nC / 2A = 34ns

PFET_SWITCH = 0.828 * 40 * (34*10^-9) * 400000 = 0.45W

PFET = 27mW + 0.45W = 0.48W

FETtemp(AMBtemp(30C)) = 30C + 62.5C/W * 0.48W = 60C

Rectifier diode

Pdiode = 0.6 * 0.75 = 0.45W

OVP-over voltage protection

For 40V:

R9 = ((100*10^3) * 1.245) / (45 – 1.245) = 2.845kΩ

Using 2.8k

For 32.5V:

R9 = ((100*10^3) * 1.245) / (35 – 1.245) = 3.688kΩ

Using 3.7k

Heatsink changes

$30 for two so is cheaper than 10 of the 25mm wide ones, and will be less hassle. Has less fins but is twice as high so should also have about the same thermal mass. 140mm will fit one large fan but might go with two for redundancy for Mk1.