Problem with DC-to-DC converter MC34063

[patb]

Hi guys,

I am trying to add to my LM317 based power supply (this is actually a simple kit) an ICL7107 based voltmeter. Here is how it is built at the moment (w/o voltmeter):

1) input 230 VAC
2) transformer to 18 VAC
3) unregulated 24-26 VDC (4700uF capacitor)
4) LM317 based regulator
5) regulated output DC (up to 20V 1A)

Now, I tried to add a little DC-to-DC converter for powering ICL7107-based voltmeter. I had MC34063 in my junk box and I decided to try it. I assembled everything according to the datasheet and connected it in the same place where LM317 module is connected. Everything seemed to work fine, but when I tried to connect something to the regulated output (100mA load, no matter what VDC) I noticed that voltmeter showed higher voltage. After a short investigation I found the cause. DC-to-DC converter is adding a lot of short voltage spikes everywhere in the circuit (more spikes if you connect something to the lm317 output). I assumed that ICL7107 is sensitive to those spikes and shows higher voltage.

Now my question. How to solve this problem. Adding more caps before/after DC-DC converter doesn't help at all. This is the first time when I try to use 34063 chip so maybe I am doing something wrong. Or perhaps this chip is not suitable for this application. Any ideas?

Thanks in advance.

[Psi]

can you poste a scope image of there pulses
and maybe a pic of the dc/dc

[patb]

Sure.

Basically DC-DC is almost the same as in datasheet (I used 10k/30k instead of 1.2/3.6 R1/R2). I attached two images from scope - the one with more spikes has bigger load. Those show output of LM317.

[tecman]

What you are seeing can be the result of a number of things.  Measurement technique (grounding, etc) are a big possibility.  Next the wiring and grounding on the switched are important.  Current paths must be sufficient in critical areas, such as the inductor, freewheel diode and output cap.  The output cap should be large enough and a low ESR type for minimizing switching noise.

It is a bit hard to tell, but it looks like diode recovers might be part of your problem.  Are you using a Shottky or at minimum a fast recovery diode ?  Are the current paths from the diode directly from the inductor to the ground point of the output cap ?  Those connections should be short and direct.  When measuring, is the scope ground short and direct to the same ground junction ?

Paul

[patb]

Ok, let's start from the begining. I've made two almost exactly identical DC-DC converters - one on a breadboard and one soldered to a protoboard (a little messy in that case). Both have similar effect on the circuit. Now for the components. I do not have brand new low ESR capacitors and the ones I have are dodgy (taken from other device, I don't know their 'ESR condition'). Anyway, connecting these not-so-new Low ESR caps changes nothing. I use brand new Schottky diode model 1N5819. Output cap and the diode are very close to each other. Probe ground wire is connected to the common ground which is '-' of the main filtering cap (just after the transformer). But maybe something is in fact wrong with that 'ground rail' in the circuit as you suggested. I will investigate further.

And of course thank you for help, Paul!

Patryk.

[IanB]

2) transformer to 18 VAC
...
5) regulated output DC (up to 20V 1A)

A separate question here, but I often see this design where the DC output is intended to be higher than the AC input from the transformer secondary.

If we consider that the LM317 will have a 2.25 V dropout voltage at 1 A, then the transformer will be supplying power to the load only when the voltage is higher than 22.25 V. This means the transformer must supply the all the required current in pulses during about 32% of the AC cycle and essentially no current for the remaining 68%. This will put quite an unbalanced load on the windings and the resistance losses in the windings will be increased substantially compared to an ideal load. It will also require a bigger smoothing capacitor and make the regulator's job harder.

If you have an over sized transformer this might be OK, but I feel in theory that one should use a transformer with a higher voltage secondary than the regulated DC output voltage expected.

[patb]

Well, I only use up to 15V really, so there are no problems in that area. The transformer itself is small 24VA 230->18VAC, so I am aware that it can't be called "oversized" for more than those 15VDC. Anyway, this is really my first attempt to build a simply power supply and to be frank I am learning basic electronics building it. Anyway, thanks for the info regarding short bursts of energy from transformer. I didn't think about that.

As for my original problem I managed to solve it. tecman (Paul) showed me correct direction where to look and by accident I got another clue when using my scope. At one stage of my 'investigation' I used both channels of my scope. One connected to the DC-DC converter (ground there) and one to the LM317 (ground there). I don't know what is the internal construction of my scope, but I suspect that both probes share the same ground. The problem disappeared when both 'grounds' were shorted.

I believe that I made a classic mistake by not properly wiring ground. I disconnected the scope and measured the potential between those two 'grounds' - it was 0.15V. Shorting them solved the problem and now even Low ESR cap works very good. I don't see any spikes from DC-DC converter at the output of DC regulator. It seems that everything now works as expected.

Question is: did I fix that problem correctly and why the ground drifted so much? I know that these questions are probably trivial for most of you, but for newbie like me it is not so obvious. Again, thanks for your help guys. I really appreciate that.

[alm]

Is the voltmeter designed to measure a signal that shares a common ground with its own power supply? Many digital panel meters require the supply voltage to be floating (i.e. separate secondary winding) relative to the voltage under test, because they position the negative terminal at something like one diode drop above power supply ground.

[patb]

I believe so. I didn't see any warning or information that it doesn't support such configuration. In fact this kit is especially designed for power supplies.

UPDATE: I've found english data sheet for this kit. Attached.

[patb]

Ouch! I noticed something I didn't notice before. It seems that the negative (minus) pin of DC regulator is not connected to GND. This power supply kit uses another IC to measure current (look at the attached schematic). Shorting DC- with GND effectively disables current measurement which is what really happens (didn't notice it yesterday).

Do you see an easy way to use voltmeter kit with this power supply kit?

Patryk.

[alm]

Power it from a different power supply (eg. battery, small transformer or extra windings around the mains transformer if it's a toroidal one). It's quite common to have separate power for DPMs anyway because of the issue I mentioned earlier.

[patb]

Power it from a different power supply (eg. battery, small transformer or extra windings around the mains transformer if it's a toroidal one). It's quite common to have separate power for DPMs anyway because of the issue I mentioned earlier.

Yeah, I've already tried using batteries (4xAA in series @ 5.2V). Unfortunately, instead of 12V I got over 32V on the display. It looked like scaling was completely wrong.

[Zero999]

Using an SMPs to power something like a voltmeter is a bad idea because it's noisy. You should use an LM7805 instead.

Going from what you've said, you're exceeding the common more input range of the ICL7107 which will give inaccurate results if either of the inputs are taken below 1V.

Differential Input
The input can accept differential voltages anywhere within the
common mode range of the input amplifier, or specifically from
0.5V below the positive supply to 1V above the negative
supply.

http://www.intersil.com/data/fn/fn3082.pdf

To measure the output voltage correctly, the ICL7107 needs to be powered by an isolated power supply or a bipolar power supply..

Batteries will work but the negative battery terminal must not be connected to the power supply's 0V rail. If you're using a toroidal transformer, you could wind an additional low current 9VAC winding with thin magnet wire and power the ICL7107's 5V supply from that. If you're using a traditional E-core transferor, you'll need to use another little transformer (9V 100mA will do.fine) as it might be difficult to add another winding.

You could also try running the ICL7107 from -5V to +5V (relative to the power supply's 0V) which could be done with a voltage doubler, an LM7905 and an LM7805.

[patb]

The ICL7107 is powered by +5V from MC34063 and -5V from ICL7660 (+5V is inverted - this is a part of the voltmeter kit). I wanted to use LM7805 but it will probably be pretty hot as unregulated voltage from transformer/main cap is about 26VDC. The voltmeter (ICL7107 + display) draws a little less than 100mA, so we are talking about 2W to dissipate in 7805 alone. What a waste. I don't know if this will change anything, because output minus DC pin is not at the same level as GND (see diagram from my previous post). It seems to be a problem to the voltmeter (why?). If I short DC minus pin to the GND (and disable current sensing module effectively) it works w/o problems.

[Zero999]

Oh yes, I didn't notice the data sheet for the voltmeter module.

Your LM317 is already wasting loads of power anyway and the voltmeter won't draw 100mA continuously so I don't see the problem.

I'll have have another look tomorrow, I'm off out now.

[patb]

Your LM317 is already wasting loads of power anyway and the voltmeter won't draw 100mA continuously so I don't see the problem.

Good point.

I'll have have another look tomorrow, I'm off out now.

Thanks.

Ok. I had another look at the voltmeter diagram. I think I know what is the problem. ICL7107 has differential input (pins IN LO and IN HI), but it also has differential reference voltage pins (REF LO/HI). From what I understand this REF voltage is jumping around because it is tied to the IN LO pin and IN LO pin is not GND but some voltage above GND. I will try to reroute it to GND instead of IN LO - this should keep REF voltage stable.

I attached voltmeter diagram with marked routes to IN and REF.
Does it make sense?

[patb]

I did change voltmeter circuit as mentioned in my previous post and it seems to work properly now. Modifying already made PCB is so painful, especially when everything is assembled already. My changes are shown in the attachment (that dotted pinkish line is removed connection, blue is added one). I hope it will work correctly from now on.

Patryk.

[Zero999]

Good, glad you got it working and more imprtantly, figured it out for yourself.

[patb]

Yup, my first power supply is finally finished and I'm happy.

Thanks for help.