TOP266 start-up problem

Something fundamental is very wrong.   What is it?   Lets try to find it.

Please do the following to simplify the circuit. 

Disconnect one leg (or both legs or remove from the circuit) of R8,

Disconnect    R11

Disconnect R5,

Disconnect D4

Change C12 to 100nF. 

Connect the V pin of the TOPSwitch to SOURCE pin. 

Once you have done the above modifications apply input voltage.  Start from zero and gradually increase it. Tell me if you get regulation.  You must achieve output regulation at very low voltage.  If you do not achieve regulation at 85VAC then stop there.  Don’t go higher. 

What is the value you have for C9?

We managed to fix the problem. The wrong Zenner diode (D7) was purchased and mounted (2V4 instead of 24V). Power supply now starts normally. However, cross regulation between the outputs is very bad. We tried to introduce a certain fix load to both outputs (500R to 5V output and 2k to 24V output – we had to that since when outputs were unloaded ringing at the secondary, right at the beginning of the active part of the secondary cycle,  would charge the output capacitors way out of the nominal values). We also introduced another feedback resistor (from 5V output to U2 reference), but the problem remained. If both outputs are lightly loaded, or fully loaded, output voltages are within 5% of their nominal values. However, if one output is unloaded, and the other is fully loaded one of the outputs will go out of specifications. We tried many different feedback ratios between the outputs (5V100% - 24V0% to 5V0% - 24V100%) but without success. All components have been selected according to PI Expert 9 suggestions. We measured transformer ratios, primary inductance and primary leakage inductance and they are all O.K. Originally we specified 5% tolerance for 24V output and 2.5% tolerance for 5V output. In reality 5V output could go to within +-5% and 24V output could go to within +10%/-20%. Whatever we tried so far did not produce satisfactory results, and we are currently out of ideas what to do next. 

When you added another resistance from 5VOUT to reference,  did you increased the value of R12 for the current in R17 remains the same as with a single feedback resistance?  

You may change R17 to 4.99K.  Then change the feedback resistor/resistors accordingly to have the right VOUTS

Reducing the value of your feddback resistor/resistors to one half will increase the gain of your compensator circuit and better regulation can be achieved. 

Take a look to the circuit in this link

http://www.powerint.com/node/231 

We made the following changes: 1. Feedback is taken from 5V output instead from 24V. 2. 24V output is clamped to 25V by 20V Zenner diode placed between 5V and 24V outputs. 5V output is now well regulated, and 24V output is also within specifications, so it seams that the problem have been solved. I have, however, one additional question. We actually built two power supplies (I attached the schematics of that other module). The difference between power supply modules is in 24V output current rating (1A and 2A). There is one additional problem with that 24V/2A-5V/0.1A power supply. When module is fully loaded and when 220V AC supply is turned on, the module wont start. It will start normally if it is unloaded, or if AC voltage is lower (110V). Also, if module is loaded and if AC is suddenly turned off, and than quickly back on, the module will not start. We included soft finish circuit but that did not fix the problem completely (10k resistor, 10u capacitor and diode). We don't have similar problems with that first, less powerful power supply module (that one does not have soft finish circuit at all). I guess we could try to adjust values of the soft finish circuit components, but we are not sure how far we can go from the default values. Is there something else we could try?

The schematic you posted on october 28.  I think it is the old schematic, am I right?   Becasue you said you are now taking the feedback from your  5 VOUT.  

Or can you post the schematic of the power supply with the problem you meantioned?........ 

""""There is one additional problem with that 24V/2A-5V/0.1A power supply. When module is fully loaded and when 220V AC supply is turned on, the module wont start. It will start normally if it is unloaded, or if AC voltage is lower (110V). Also, if module is loaded and if AC is suddenly turned off, and than quickly back on, the module will not start. We included soft finish circuit but that did not fix the problem completely (10k resistor, 10u capacitor and diode)."""" 

Does your schematic follow the recomended components that PI expert listed?  

Thank you 

Attached PSI_M245_2A01A_1013_02 schematics shows the changes we made to 24V/2A-5V/0.1A power supply. With that power supply we are experiencing start-up problems I described. The problem was also present before we made any modifications. Modifications we made to the original 24V/1A power supply (first schematics I sent) are the same (except there is no soft finish circuit). Today we discovered high amount of ripple (+-200mV) on both power supplies, and on both outputs when outputs are unloaded. When load is present, the ripple is significantly decreased. And, yes, we followed all the component values, and transformer construction, suggested by PI expert.

First let try to fix the ripple issue.

It is possible the ripple is dues to:

1.      Bias voltage gets too low.  So check the voltage at C9 when the supply is unloaded.  You should have at least 8 Volts DC with no ripple. I am not sure what is the value of C9 you have, but for my personal design I never use less than 1 uF.  But it is nice to use 4.7uF or 10 uF if possible.

2.      If the ripple still there, then reduce R7 to 1k.   This reduction in R7 will boost your loop DC gain. 

Let me know if these suggestions fix your ripple problem.  Then we can talk about the other problem

Mean voltage at C9 is 13.4V (there is also a lot of). Capacitor's value is 22uF/50V (80mohm ESR). I reduced R7 to 1k but ripple voltage got bigger! (800mVpp comparing to 400mVpp with default R7 value). Ripple frequency is around 111Hz. By measuring voltage at D4 Anode I noticed that TOP268EG operates in "bursts". Period between bursts is 9ms, and each burst is 2.5ms long. One burst is consisted of TOP268EG 30kHz cycles (each cycle has small duty cycle).

There is a lot of ripple at C9...

what is the value you have for  C9?

Mean voltage at C9 is 13.4V. Capacitor's value (C9) is 22uF/50V (80mohm ESR). I increased C9 to 68uF but without result. I reduced R7 to 1k but ripple voltage got bigger! (800mVpp comparing to 400mVpp with default R7 value). Ripple frequency is around 111Hz. By measuring voltage at D4 Anode I noticed that TOP268EG operates in "bursts". Period between bursts is 9ms, and each burst is 2.5ms long. One burst is consisted of TOP268EG 30kHz cycles (each cycle has small duty cycle).

Mean voltage at C9 is 13.4V. Capacitor's value (C9) is 22uF/50V (80mohm ESR). I increased C9 to 68uF but without result. I reduced R7 to 1k8 but ripple voltage got bigger! (800mVpp comparing to 400mVpp with default R7 value). Ripple frequency is around 111Hz. By measuring voltage at D4 Anode I noticed that TOP268EG operates in "bursts". Period between bursts is 9ms, and each burst is 2.5ms long. One burst is consisted of TOP268EG 30kHz cycles (each cycle has small duty cycle).

In the past couple of days I was playing around with various values of feedback components but nothing worked.
The best result I got when I increased C12 to 220n (with R7 decreased to 1k8 and R20 and C14 modified in order
to get crossover frequency as close to 1k as possible with as high as possible phase margin). That produced
output ripple of about 200mVpp when output is lightly loaded. Finally, I removed C8, and left only small 10u
ceramic capacitor at the output (according to AN47, C8 should be around 68u in order to get filter resonant
frequency slightly above 10k - ). That step lowered ripple down to about few tens of mV. So I have a few additional questions:

1. From my understanding of how TOP268EG works, I guess that my problems were related to TOP268 operating in Multi-Cycle-Modulation mode. When I decreased output capacitor, I forced TOP268 to work in Low Frequency PWM mode so the ripple decreased. The condition when to switch between the modes is apparently related to peak drain current. It seams to me that if threshold to switch to Multi-Cycle-Modulation mode is set to high the output would experience high ripple. I can, of course, use a smaller output capacitance, but what if someone connects a device with a large input capacitors to 5V rail? How can I ensure that the supply will stay stable and will not generate high ripple at light loads?

2. I have some concerns regarding R7 value. PI expert suggested 6k49. However, since optocoupler's CTR may drop down significantly at higher temperatures, current trough optocoupler's LED will have to be increased accordingly in order to regulate the output (My worst case estimate is that it may go to about 12mA). On the other hand, with R7 being 6k49, the maximum current trough the optocoupler's LED would be much lower. I calculated worst case R7 to be round 1k3. However, with 1k3, and using PI expert, I was not able to find other feedback components to maintain crossover frequency close to 1k default, and, at the same time, to get sufficiently large phase margin. I could use some guidelines here.

3. I am not sure if analysis I presented here is correct, but I could use some advice from your side since I am not quite sure how to proceed.

Your very last message you sent,  where are you taking the feedback tot he optocoupler?   from the 5 VOUT or from the 24VOUT?

22 uF capacitance for C9 is enought.  You do not need more as long as your minimum ripple voltage is not below 8 Volts  

Optocoupler is connected to 24V rail, whereas TL431 reference input is connected to 5V rail (using resistor divider consisted of two 11k3 resistor). Originally, both optocoupler and TL431 were connected to 24V rail, but topology was than modified in order to get tighter regulated 5V rail and better crossregulation. The ripple problem was present even with TL431 connected to 24V rail.

 What you need is a phase boost.  It is a paper about hot to boost the phase in venable instruments webpage the paper  "  VENABLE TECHNICAL PAPER # 3"  can be downloadet.  it tell you how to calculate your componets in order to have an aptimum boost 

But I already have a phase boost (serial connection of 56R and 47n in parallel with R7). According to PI expert that (together with C12 set to 220n and R7 set to 1k3) produces phase margin of 52 degrees and crossover frequency of 3200Hz. I could get 60 degrees of phase margin by increasing boost capacitor from 47n to 68n, but that would also increase the crossover frequency to about 4k. Seams to me that phase margin is not so bad (assuming that PI expert calculations are correct). So, why would I need additional phase boost, and where exactly would I have to place additional components beside those I already have?

I am sorry for the misunderstood.  Sometimes is hard to follow the issue when I am away from you.

I think for my understanding, now you want to reduce the cross over frequency. Am I right?   If so, I suggest you to increase the value of R7.  This increment will shift your envelope gain and your power supply response with cross over earlier.   Let me ask you something, do you actually have a phase analyzer to measure the loop response of the supply?

Unfortunately we don't have phase analyzer, and we were hoping that calculated components would be more or less close to the optimum. All we can do is to change the component values and check how power supply reacts to various loads, load changes, and how it will react at startup. In our last iteration we moved feedback completely to 5V rail. R7 is decreased to 33R, C12 to 680n and we also added C14 and R15 (please see the attached schematics). This configuration produced the best results so far. When power supply is unloaded ripple is now reduced to 20mVpp which is acceptable. When both outputs are fully loaded ripple is decreased, and when only 24V is fully loaded ripple is increased to about 50mV. Power supply reacts well when load on one output goes from zero to maximum, and on the other from maximum to zero (and vice versa) - no significant overshoots or undershoots have been observed. Power supply now also starts when it is fully loaded. However, there is one "but". First, voltage rise during startup is not quite monotonous as we expected, and PS appears to be working in discontinuous mode (please see the attached images). According to your PIXls Designer 9, that should not be the case (supply should work in CCM, very close to DCM since Kp is little bi less that 1). Is there a way to improve startup voltage waveform, and switch to CCM?

Star Up

The waveform you have for your startup, are you using electronic load ? When testing Output star up, you should use a resistive (not electronic) load.

DCM.

You have KP about 1,  so it mean you are close to the  border between  DCM/CMD,   The number for KP that PI expert gives you is when you have minimum Vin and maximum POUT.   If you increase your VIN or decrease your load, then the power supply will work more in DCM.

The only way to make it work CCM is to redesign the transformer and use a lower KP number

We are using a set of 5W 4R7 resistors arranged in a way to get approximately 12 ohms.

why is R15 for?    does PI expert recomend it?

I would try to reduce C12 to about 100nF    

R15 provides discharge path for C14 (soft start function). I found that in your application note AN22 (Fig 10). I can't reduce C12 because if I do that the ripple becomes to big. I already tried reducing C12, but that did not work well. Regarding PI Expert, it does not seem to me that this tool is very reliable. As I mentioned before, we made two power supply modules (50W and 25W). For designing both we used PI Expert. We noticed that 25W power supply transformer runs quite hot when fully loaded (around 70 degrees C at room temperature). According to AN47, EF20 that PI Expert suggested is for power ratings from 10-20W, and we are dealing with 25W. When I did the calculation using PIXls Designer trying to get the same number of turns for EF20 core, I got bigger wire diameter for primary turns that the one suggested by PI Expert (Frequency is 132kHz). Also, for post filtering components PI Expert offers 100u capacitor and 3u3 inductor. On the other side, AN47 states that resonant frequency of post filtering components should be at or slightly beyond 10kHz (with suggested components it is below 10k - or I am missing something here). I seams that there is many contradictory information and I am not sure any more which one is correct.

I will try to answer all your questions as clear as possible.

Regarding R15  You are right , this is a discharge path for the soft finish cap C14 suggested in AN22.  Personally I do not recommend this arrange,  why?  Because the extra capacitance you are connected at the cathode of U2 is going to slow down the response of U2.  Is this a problem?  May be not but when I am adding  a soft finish cap directly connected to the cathode of U2 I slowing down the response to the TL431. Slower response may be mean more ripple. What I usually do is to isolate R15 and C14 by connecting a diode between the cathode of U2 and the node joining R15 and C14. The anode of the diode will be connected to the cathode of U2. This way the soft finish RC network (C14 and R15)   will be active only at start up.

Another suggestion for your circuit is to delete D11 from the circuit and increase the value of R12 to 97K

I think this change will make the loop more responsive and a phase boost may be added in parallel with R12.  What I mean phase boost is am RC circuit. We can talk about it later. 

Regarding your concern about PI expert.   I do not have an answer about it.  If you like you can contact PI expert engineer directly in this forum going to the PI expert section.

About the temperature of your transformer when running at 25 watts.  I am not sure if I understand what you said.

Do you mean when the power supply is running at 25 Watts output and 25C ambient, your transformer temperature is around 70 C?    if this is what you mean then you transformer is running with a temperature od about 50C over the ambient temperature ( 75 – 25 = 50C)   If this is the case then the transformer is having a very good temperature.