12V - 350V 200mA converter for motorcycle CDI

[(*steve*)]

I didn't reply earlier, but these comments may help, or at least may reinforce what you and others have thought...

When you reported that the output voltage rose, the transformer started making a noise etc., this seemed to me as a case of failure in your feedback network.

Without feedback your power supply would go into an over-voltage situation - it is the feedback network that tells it to stop or slow down in order to maintain a regulated output.

The fact that rebuilding the device fixed the issue points to some sort of failure. That may have been as simple as a solder joint that failed, or more complex like a component failure.

Both of these issues need to be addressed in a motorcycle environment. Firstly it is subject to a lit of vibration, so you need a method of construction that is reliable and rugged. Secondly, any component failure issues caused by poor design need to be weeded out or you may find yourself stranded somewhere.

If I were you, I would go back and carefully examine the old board, testing to find out where the feedback signal stops. Alternatively, run the new board for a long time to "soak test" it.

I am concerned that you have replaced fast schottky diodes with far slower diodes for D3 and D4. What does LTSpice tell you will be the negative impact of using fast schottky diodes here?

 

[abuhafss]

I didn't reply earlier, but these comments may help, or at least may reinforce what you and others have thought...

Welcome back !

If I were you, I would go back and carefully examine the old board, testing to find out where the feedback signal stops. Alternatively, run the new board for a long time to "soak test" it.

I had spent 4-5 days on the old board, replaced almost all the components for the inverter (disabling the ignition section) but, could not get any clue of any dude component, loose solder, disconnected trace etc. So to get myself out of the frustration, I assembled step-by-step on a breadboard. The positive and improved results have refreshed me. So now, I shall construct a new PCB and then soak test it.

I am concerned that you have replaced fast schottky diodes with far slower diodes for D3 and D4. What does LTSpice tell you will be the negative impact of using fast schottky diodes here?

Actually, I started working on LTSPice with all components corresponding to the OEM CDI.

Here are the findings, working on LTSpice with zeners (100V + 100V):

D3, D4, Output, Time taken to reach maximum output
a) 1N4148, 1N4148, 190.66V, 54.7ms
b) 1N4148, 1N5819, 187.80V, 52ms
c) 1N4148, UF4007, 191.70V, 57ms

d) 1N5819, 1N4148, 191.20V, 49ms
e) 1N5819, 1N5819, 188.02V, 47ms
f) 1N5819, UF4007, 191.51V, 50.3ms

g) UF4007, 1N4148, 190.66V, 57ms
h) UF4007, 1N5819, 187.80V, 53.75ms
i) UF4007, UF4007, 191.24V, 57ms

Kris had recommended the combination (h). But, any of these combination doesn't show any significant difference in simulation or physically. Time difference is maximum 10ms and output difference is about 4V. So I stuck myself to the original components. Theoretically, the recommended combination might be the best.

 

[Rleo6965]

Pcb etch should be wider on high current circuit specially on emitter ground pcb. This will also help heat dissipate heat from power transistor, diodes and resistor.

 

[(*steve*)]

I had spent 4-5 days on the old board, replaced almost all the components for the inverter (disabling the ignition section) but, could not get any clue of any dude component, loose solder, disconnected trace etc.

It can be frustrating. If it happens again, be mindful that the problem could be related to some design issue. If so, it would serve you very well to track it down. If it does not happen again, maybe it was some random failure...

So to get myself out of the frustration, I assembled step-by-step on a breadboard. The positive and improved results have refreshed me. So now, I shall construct a new PCB and then soak test it.

Remember that breadboard (solderless breadboard) has relatively high capacitance. This may subtly change your circuit characteristics.

Here are the findings, working on LTSpice with zeners (100V + 100V):

You need to consider how well the diodes are characterised in LTSpice. I have no idea. The fact that they each give different results (albeit slightly) suggests that they are at least characterised differently

Kris' advice may have been that those particular diodes would be "best" for a reason other than "fastest to achieve output voltage".

 

[KrisBlueNZ]

Kris' advice may have been that those particular diodes would be "best" for a reason other than "fastest to achieve output voltage".

Say what now? I don't think I recommended any zener diodes. I've found post #54 where I recommended some fast recovery and Schottky diodes. Is that what you mean?

 

[(*steve*)]

Say what now? I don't think I recommended any zener diodes. I've found post #54 where I recommended some fast recovery and Schottky diodes. Is that what you mean?

Yeah, D3 and D4 on the circut I think he was referring to are a diode connected between emitter and base and another one. Neither were rectifying the generated voltage or a zener.

 

[abuhafss]

Yeah, D3 and D4 on the circut I think he was referring to are a diode connected between emitter and base and another one. Neither were rectifying the generated voltage or a zener.

Correct.

BTW, for the sake of knowledge, please tell me the function of C3, C4 and D4.

 

[abuhafss]

In the next phase, I assembled the UC3845 based third circuit with some modifications. Here is the final schematic:



Transformer is 0.4mm/15 turns for primary and 0.2mm/300 turns for secondary. I used IRFZ24N which was available with me. Though the LTspice simulation works perfectly with C1=1nF but, it physically it worked with C1=10nF. The output is constant 210V as expected and the mosfet is at ambient temperature. Please check and suggest any improvement/refinement, if required.

Next, I also tried to use ferrite bobbin as shown below, for making the transformer (though not recommended) with same configuration.



The result was same, no difference. Any comments???

 

[KrisBlueNZ]

The diagram looks good to me. I can't comment on the ferrite former without knowing its specifications.

It's easy to check for core saturation in that circuit - just put a scope across the source resistor.

 

[abuhafss]

The diagram looks good to me. I can't comment on the ferrite former without knowing its specifications.

It's easy to check for core saturation in that circuit - just put a scope across the source resistor.

Can you please guide how to check out the specs of the ferrite former. I just got those from a local electronics store, without specs. Most probably, they are from China.

 

[KrisBlueNZ]

OK, I'm dredging back quite a few years here. If anyone spots a mistake here, please say so!

Normally, the manufacturer gives you full specifications. These include µr (relative permeability, a dimensionless quantity), A_L ("inductance amplification" in nH/turn²), l_e (effective length), and B_sat (magnetic field density at which the core saturates). They will also tell you the core material so you can choose according to the operating frequency, hysteresis, and saturation characteristics (soft or sharp).

If you don't have that information, you can estimate the A_L value by winding a known number of turns and measuring the inductance. Convert the inductance to nH and divide by the square of the number of turns to get the A_L value.

Then you can use the A_L value to calculate the number of turns required to achieve any inductance you want, or vice versa.

The saturation current (for a given number of turns) can be found by testing. I suggest you make an adjustable-frequency oscillator, for example using a 555, to drive the MOSFET gate, and connect D1's cathode to the 0V rail, either directly or through a low value resistor such as 100Ω so the flux can discharge quickly. Then monitor the voltage across the source resistor (R7) with an oscilloscope. Start the oscillator frequency at, say, 50 khz and gradually reduce it. You will see the rising ramp getting longer as the frequency drops.

At a certain frequency, the ramp will start to rise rapidly at the end. This indicates the core is going into saturation; the current starts to increase rapidly. Find the voltage on the scope where the core starts to saturate, and divide it by R7's resistance (Ohm's Law), to calculate the core saturation current at the primary.

Then you need to ensure that the UC3845 will not try to saturate the core. You do this by ensuring that the voltage at the current sense pin will reach the maximum output voltage from the error amp (which is clamped at some specific voltage - see the data sheet) before the core reaches saturation. You do that by adjusting R7.

So, say saturation starts at 3.6A. You want a reasonable safety margin, so let's say you choose a maximum primary current of 3A. And assume the maximum error amp output is 1.2V (I forget the exact value). So you need 1.2V at the current sense pin at a primary current of 3A, so the ideal R7 would be 1.2 / 3 or 0.4Ω.

Any corrections please...?

 

[abuhafss]

Thanks for detailed explanation. I'll check and report.

BTW, in the present situation, when the output is as expected, mosfet is at ambient, transformer is cool & no noise; could there be saturation or doubt of saturation?

 

[abuhafss]

Ah yes, one important question.....
Can a UC3845 be soldered directly or should use IC base?

 

[KrisBlueNZ]

No, if the core was saturating on every switching cycle, the MOSFET would run warm at least. It might be saturating during the early part of each charge cycle. But you should make sure that it can't, by design. Which means you need to know the actual saturation current.

Have you been testing it at maximum RPM?

Have you looked at the output voltage (the voltage waveform on D1 cathode) at maximum RPM?

You can use a socket to make it easy to replace the IC if it gets damaged. That's not very likely unless you exceed the input voltge range.

 

[abuhafss]

Have you been testing it at maximum RPM?

Have you looked at the output voltage (the voltage waveform on D1 cathode) at maximum RPM?

I have been busy to build a bench tester to check the performance of the CDIs.
Although the output of the UC3845 based inverter is 210V, the timing components R2=47k & C1=10nF take too long (91ms) to charge C5 at about 1.84kHz (duty cycle about 50%). The bench tester shows very weak spark in a gap of 5mm at slow RPM and at about 7000RPM it would spark in only 2mm. As mentioned in #128, C1=1nF did not resulted the desired output. I also checked with increased output of about 300V but no improvement confirmed incomplete charging of C5.

I also installed the CDI in the bike, it failed to start the bike. Instead the UC3845 turned bad. I lost two ICs while testing the CDI in the bike.

 

[KrisBlueNZ]

What are your inductor characteristics?

Can you post a schematic?

 

[abuhafss]

What are your inductor characteristics?

Can you post a schematic?

Same as post # 128 !

 

[KrisBlueNZ]

OK, I won't be trying to suggest anything until you start using inductors with known characteristics instead of recycling bits from your junk box!

 

[KrisBlueNZ]

Obviously you can't provide details on those cores. This means you're taking the wrong approach. You need to get some cores that come with proper documentation.

I think NXP have a good range of ferrite components. At my previous job, we used a lot of NXP 3C80 gapped cores for switching supplies. These were only used for buck converters though. I remember having to learn about saturation details, calculating based on effective lenght, A_L value, and other parameters. Depending on the power requirements, an RM8, RM10 or maybe RM12 might be worth trying. Also, ETD cores.

Get a selection of formers and several reels of enamelled wire. Look for application notes from NXP to explain the practical considerations and calculations for custom-wound inductors and transformers. There is quite a lot to learn, but it is all valuable. Then you will be able to define the characteristics of your inductor - not just the inductance of each winding, but also saturation current (very important), DC resistance etc.

That's what I recommend. Get some cores, bobbins etc, and read up on the applications information so you know all of the important quantities and how to calculate them.

 

[duke37]

Many years ago, inverters were made to generate the sort of voltage and power that you desire. These used push-pull circuits which were self oscillating. When the transformer saturated, the output would switch. If you use a separate oscillator it must run fast enough to stop saturation.

I have used a ferrite core from a TV line output transformer with 1 turn/volt. The power does not pass through the core like the boost converter so so the permitted power was limited by wire size and switching capability. At 1t/V start at 10kHz.

The boost converter which you show loads the energy into the core and then dumps it into the output. Core characteristics then become critical.