MC34063 boost drooping

I bought 10x MC34063 for £1 to experiment with, built a little board for one since since they are soic packaged (hence the cheap cost) to use it as a boost converter with 5v in, 28v out, running about 100KHz. Worked lovely with only my dmm connected, but as soon as a light load (1.5k) was connected it dropped down to little over 10v. What's going on? Anyone know?

Best guesses so far, it's flakey because of the amount of boost used for the stage and is getting unstable, or I'm dinging a bad t-on / t-off ratio.

TIA

Edit. "flakey". Tsk, darn, I meant unstable, didn't I... Long day, need sleep zzzzzzz

for what it's worth; I had titanium caps and I put one in backwards that fixed my one problem. also one started to droop when overheated untill I put a better heatsink on it. other then that the only other problem was when my psu was set to a lower amperage then what I was using which caused the voltage to drop way low, when the fans kicked in then I knew I was overloading the supply.

also that was with Linear LTC3245 convertor which I've been using for a little while,(I like them due to no inductor is needed, great for tight places) I've got some LM46002 which have a much higher output then that of the MC34063, but needed to save some space.

Titanium caps? You mean tantalum I hope! I guess I'll just have to play with it some more.

One thing I did do - I used a high speed rectifier instead of a schottky diode. Could this be the culprit?

tantalum is correct, I'm a machinist by trade.....

Hi,

A high speed rectifier should be ok, but if still in doubt simply try a Schottky and see if it helps.

There is a limit however on the boost ratio input to output which depends on a number of factors most notably the circuit resistances.

First there is the switch resistance, which must be low for high boost ratios.
Second and more under control is the inductor resistance. If the inductor resistance is above a certain value then it wont be able to boost to a given output voltage no matter how good the circuit is.

So the first thing to check is the inductor resistance. To the output the resistance of the inductor looks like it is multiplied. If you can test the resistance using a decent ohm meter we can figure out if that is causing the decrease in output with load. Many times that's all that it is, and that means an inductor with lower series resistance is required.

The max inductor series resistance can be estimated from this:
RL=(Rout*Vin^2)/(4*Vout^2)

where
RL is the maximum allowable inductor resistance,
Rout is the load resistance,
Vin is the input voltage,
Vout is the desired output voltage with the desired Rout load.

With Vin=5 and Vout=28 and a load of 28 ohms we get:
RL=0.223 ohms max.

With those same parameters but Rout=1500 ohms we get:
RL=11.96 ohms max.

Check the inductor, and also the input source resistance which also plays a big part in the max output voltage with load.

Also note that these figures are for the inductor alone. When there is switch resistance as well then the inductor max resistance must be lower than this estimate too.

I'm using pretty chunky inductors, I think I can ignore the resistance for this. I will try it with an external switching transistor however, I might get a lower switch resistance. From what I've read, possibly I need to run the oscillator at a lower frequency too.

Thanks

Hi,

Well if you are sure the inductor is ok then you do have to check the other things. 100kHz doesnt sound too high, but then 50kHz would work better if it was that.
You might want to post the circuit so we can take a better look.

I will look over the next few days - and post the circuit, though it's just the one off the data sheet with my own values.

Hi again,

Ok, maybe if we take a look we can spot something that might be out of place or something overlooded. This happens sometimes and it turns out to be very simple.
The chip itself should work ok i guess, but then we have to see what you are using for the main switch too. If it is the chip itself maybe the chip is bad or a bad connection.
Not too long ago i had what is probably my first cold solder joint problem ever. I had soldered a 7 segment display chip socket (an older off the shelf part with tarnished pins) and did not use the right soldering iron and tip so it did not get hot enough for a good solder joint in 2 or three joints. Needless to say it did not work right, but again this was a simple problem to fix once found. The cold solder joint can produce a joint that is either totally not connected at all or just higher than normal resistance which can really mess up some circuits where even smaller currents are involved.

Ok here is schematic and PCB. It's the same as the example boost converter given in the data sheet. Values are a little bogus, not sure exactly what I used. I wanted to be able to adjust the output, hence preset inserted.

Tested it with the 'scope today. Changed the cap from a 330pF ceramic to 680pF polystyrene, however it made no difference. I discovered that as I adjust the preset, it goes from discontinuous mode at lower output voltages, to continuous mode at higher output voltages. The point at which it changes from discontinuous to continuous mode is the same point above which the output will droop when a load is applied. Used a heavier load - 150 ohm, output dropped even further.

I did find this web-site, which I think gives the best insight into what may be going wrong. https://www.changpuak.ch/electronics/High_Voltage_Power_Supply_MC34063.php

Hi,

I'll take a look at that. In the mean time, what is the duty cycle when your load is 1k ?

One thing you can also check is what the max current through the inductor should be given your switching frequency.

Another thing to think about is that in a boost converter if the duty cycle goes above a certain limit, the output voltage actually decreases. So a quick test would be to start the circuit with no load, then apply a 1k load while it is still running. See if the output stays higher or goes lower again.

Typical boost converters are a little strange compared to other topologies. With the circuit operating normally when a load is applied the output voltage drops down a little and this in turn tells the control circuit to increase the duty cycle. The duty cycle then increases, and the output rises back up again to the right value. Once at the right value the duty cycle stays more or less constant.
If the load is again increased however, the same thing happens, but at some point when the load is increased the control circuit again increases the duty cycle but instead the output voltage DROPS FARTHER DOWN, and this in turn tells the control circuit to increase the duty cycle even more, and that makes the output voltage drop again. This continues and it may NEVER snap back into normal operation even if we remove some of the load. The feedback mechanism changes from the normal negative feedback to positive feedback, so the output stays low.
This occurs if the duty cycle exceeds that magic limit due to the resistances in the circuit.

A problem that could result from this maybe is that during turn on, the output voltage is low so maybe the control circuit increases the duty cycle more and more as the voltage rises, but before it gets to the normal operating point it passes that magic duty cycle limit and so it never gets up to normal operating voltage.
In this case we would have to add a slow start mechanism somehow, which only allows the duty cycle to change so fast during start up so that the output gets a chance to respond before the control circuit takes it past the point of no return.

Ah, you just explained the slight hysteresis I observed. Not had chance to play with it and examine duty cycle since last post.

Thanks for showing the interest, Mr Al.

Ok just had another tinker. I tried a couple of other rectifiers, first a big fat high speed rectifier, which improved matters, then I found a shottky on an old board, so put that in amd things improved much more. I also tried a few different inductors. With a 12v input (well, actually just over 10v, it's supposed to 12v), I can get 26v out into 100R load, so I'm happy with that. With a 5v input, I get 15v out with the same load. Looking at the duty cycle I'm getting a lot of overshoot and ringing, which I suppose is causing inefficiency. M/S ratio seems to go no higher than 5:1, which I think is a duty cycle of 83%, close to the limit allowed.

So I think I'm going to get a beefy shottky I saw on eBay, fit an external transistor, and wind a tapped inductor, and see how that goes.

I'm wondering about parallel rectifiers. I would need to put in current balancing resistors, but I wonder if these would slow things down noticeably due to junction capacitance? And if it does, if I put a small fast signal diode in reverse parallel with each resistor, would it overcome the effect?

I think you are suffering from the old tolerance build up problem. The biggest is probably the on to off ratio. The range extender in the page you posted would really help that. Then you loose some across the sense resistor, the inductor, the Darlington and finally the diode.

Yes, I finally discovered that the losses you mention make a big difference at low input voltages. I just hadn't realised how much. Given the PSU I'm using is also a little low, it's no wonder I'm having problems.

Well, I intended a learning exercise here, it's certainly educational!

Now I am re-designing it for more current, I can include the range-extender at the same time.

The data sheets for the 34063 are really useful.
The key component is the INDUCTOR. You need to get the inductance value and start to DESIGN the operating frequency.
Using the design formulae will tell you what is happening.
These things are commonly used in 12 volt car/phone chargers. I use them all the time and ALWAYS you need to change the operating frequency. Sometimes a boost transistor is necessary. For low voltages output a Shottky diode is necessary, but at 28 volt output its not so significant. The boost output implies a high inductor current at switch-off so you ned to do the sums as per the datasheet.
hope this helps.

I hate to admit it but I played with it a long time in LTSpice before I went thru the data sheet close enough to see the problems.