Author Topic: Half-Bridge 0-30V, 0-20A Feedback loop problem  (Read 10621 times)

0 Members and 2 Guests are viewing this topic.

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Half-Bridge 0-30V, 0-20A Feedback loop problem
« on: June 07, 2019, 06:50:17 am »
Hello, this is my first post,

I am writing this post because I hope that you will help me solve my problem.

Designs adjustable Half-Bridge converters with 0-30V and 0-20A parameters. The power block itself is able to give over 800W at maximum PWM and it is not a problem. The problem is the feedback loop, which works unstable. I thought it might be a problem with the low choke value, originally 47uH, but then I used the 250uH inductor and it's even harder to stabilize the converter.

In general, the whole works, it is possible to regulate the voltage and current, but there are considerable squeaks of the transformer at low voltages, ie 10V and quite large currents of 10A. There are also significant noises at the output as well as sinusoidal oscillations. Generally, these are noises with an amplitude up to 1V when measuring AC on an oscilloscope

The loop is created on the external LM324 operational amplifier, while the internal SG3525 amplifier works as a voltage follower. The switching frequency of the transformer is 40kHz. Using the book Switching Power Supplies fromA to Z - Sanjaya Maniktala I used and calculated coefficients for the type 3 compensator for the following data:

fESR = 6392 Hz
fLC = 424Hz for 47uH inductor and 184Hz for 250uH inductor
fz1 = fz2 = 424Hz
fp0 = 260Hz
fp1 = fESR = 6392Hz
fp2 = 1/2 fsw = 20kHz
frcoss = 1/10 * fsw = 4kHz
Vramp = 2.8V
Vin = 43V (output voltage after diodes)
fsw = 40kHz
Cout = 3 x 1000uF
Total ESR = 8.3mohm


I calculated the following element values ​​for the 47uH inductor

C1 = 30.6nF -> I used 33nF
C2 = 17.5nF -> I used 18nF
R2 = 12.3KOhm -> I used 12kohm
C3 = 678pF -> I used 680pF
R3 = 1.4kohm -> I used 1.5kohm

For inductor 250uH, the coefficients are:
C1 = 30.6nF -> I used 33nF
C2 = 42nF -> I used 42nF
R2 = 28,3kohm -> I used 27kohm
C3 = 297.6pF -> I used 300pF
R3 = 600ohm -> I used 560ohm

Despite many attempts, I have no idea where the problem is.
It also adds inverter schemes. The diagrams are two because the inverter consists of 2 PCBs
1. PCB is a power part
2. PCB is the control part

Thank you for your help
« Last Edit: June 07, 2019, 06:53:25 am by paladyn »
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #1 on: June 07, 2019, 09:46:35 am »
I can't help directly, I don't understand the theory well.
I mainly what to show how I did the fast current limiting via the soft start pin, which you are likely to need also.
It is temperature dependent but this can be useful.
For voltage regulation, I used the internal error amp in transconductance mode. The response is very good.
It doesn't sense the full PSU's output terminal voltage, it regulates to maintain 3V drop across the series pass MOSFET of the  proceeding linear post regulator.
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #2 on: June 07, 2019, 10:07:58 am »
It looks to be too much gain for the voltage loop. Try increasing R1 from 1.5K to 20K.
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #3 on: June 07, 2019, 10:37:15 am »
Did you notice the regulated output voltage is one of the parameters needed to calculate the compensator component values? As a result, if you want to make the output voltage adjustable over a wide range (NB - smoothly going down to 0V isn't an option, practically speaking) then you will have to grossly overcompensate the loop (that is, set a really low crossover frequency - perhaps a few 10s of Hz). Going with current mode control (CMC) will improve matters by eliminating the pole from the output LC filter, but CMC doesn't play well with the half-bridge unless measures are taken to ensure equal volt-seconds are applied each half cycle of operation (the full bridge is more amenable to CMC as long as the DC blocking capacitor that is usually in series with the primary is deleted).

The usual approach to making a wide output voltage range switcher is to cascade it with a linear regulator and design the switcher/regulator control loop such that the former maintains a few volts of headroom for the latter (that is, the switcher maintains a nominally constant difference between the input and output voltage of the linear regulator). The slower the transient response of the switcher (ie - the lower the crossover frequency), the more headroom is required for the linear regulator (along with a much larger filter capacitor on the output of the switcher; way more than is required to meet ripple specifications).

 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #4 on: June 07, 2019, 07:39:36 pm »
Thank you for the suggestions, I will try to change the R1 resistor to something bigger and I will try to change the fcross to a significant one, eg 500Hz.

From what I see, designing such a flexible converter will not be easy, but it will try.

I still have the question of calculating the coefficients, the transformer ratio is somehow included in the calculations?
 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #5 on: June 10, 2019, 06:21:19 am »
MagicSmoker

do you have an example of how this coupling is implemented in a real system?
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #6 on: June 10, 2019, 10:37:53 am »
MagicSmoker

do you have an example of how this coupling is implemented in a real system?

I'm not sure what coupling you are referring to, but based on the context it seems likely you are asking how to make the switching power supply pre-regulator maintain a constant headroom for a linear post-regulator, even when the latter is adjustable? If so, then the key is to derive the voltage feedback for the switcher via an op-amp wired to look at the voltage drop across the linear regulator (that is, a differential amplifier) but with an offset applied to the terminal sensing the input to the post-regulator (via, say, a zener diode, precision two-terminal reference, or several conventional diodes in series). Consequently, the output of the op-amp will be zero when the input voltage to the pre-regulator is higher than the output by the offset voltage; et voila, you are now regulating the headroom across the linear regulator automatically.

 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #7 on: June 10, 2019, 10:59:55 am »
MagicSmoker

do you have an example of how this coupling is implemented in a real system?

I'm not sure what coupling you are referring to, but based on the context it seems likely you are asking how to make the switching power supply pre-regulator maintain a constant headroom for a linear post-regulator, even when the latter is adjustable? If so, then the key is to derive the voltage feedback for the switcher via an op-amp wired to look at the voltage drop across the linear regulator (that is, a differential amplifier) but with an offset applied to the terminal sensing the input to the post-regulator (via, say, a zener diode, precision two-terminal reference, or several conventional diodes in series). Consequently, the output of the op-amp will be zero when the input voltage to the pre-regulator is higher than the output by the offset voltage; et voila, you are now regulating the headroom across the linear regulator automatically.
That's how my bench supply works.
I used the Harrison topology linear regulator witch has everything referenced to the positive output terminal being ground.
So the SG3525 control block is also ground referenced to the positive terminal.
The 3V Sense simply connects to the Drain of the post regulator's series pass MOSFET which also connects to the positive output of the SMPS. 
« Last Edit: June 10, 2019, 09:17:55 pm by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #8 on: June 11, 2019, 06:12:22 am »
Colleagues, the problem is a little different. I understand that your constructions work in two stages.

AC main -> SMPS converter -> Linear power supply -> output

I have a problem matching the feedback of the SMPS converter itself. At low power (100W) the transformer starts to beep loudly. Perhaps this is because the transformer has not been stuck together and the windings fall into vibration?

I also carried out tests at a lower frequency of fcross 500Hz but I had the impression that the coupling works even worse.

I did not plan to make some great power supply with 20mV noise, I just wanted to make a similar construction as commercial low power supplies that are controlled by TL494 such as Gophert-cps-3205https://www.kirich.blog/obzory/128-laboratornyy-blok-pitaniya-gophert-cps-3205-ili-tot-redkiy-sluchay.html

I also noticed that the current feedback loop works worse than the voltage loop. The only thing that is different in these loops is an additional operational amplifier INA213, which strengthens the shunt voltage. Do you use any additional amps to measure the current?

I'm still thinking about changing the operational amplifier from LM324 to MC33074. Is the sense or rather not the problem here?
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #9 on: June 11, 2019, 06:58:19 am »
Can you see the instability in the loop? Like a few KHz sine wave?
I think that you have too much gain.
Because I have used the SG3525's error amp in transconductance mode, it is difficult to see what the gain is from the schematic because there is no local feedback. I estimate the gain to be 5 with the 10K load.
I recommend the SG3525 mainly because it has multiple pulse suppression and it is easy to implement pulse by pulse current limiting.
Your primary snubber looks like it will have high dissipation. Mine is 120Ω and 470pF.
Watch for ringing at the secondary. It adds a lot to the diodes' PIV and can kill them.

Correction: The gain of the error amp is 15 with the 10K load.
« Last Edit: June 12, 2019, 01:26:57 am by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #10 on: June 11, 2019, 07:50:16 am »
Removing R1 made no difference?
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #11 on: June 11, 2019, 08:53:17 am »
xavier60

I tested different feedback configurations and sometimes I could see high-frequency sinusoidal waveforms, but this is only when testing on the values ​​of elements that deviate a lot from those calculated in accordance with the book. On the other hand, for the calculated values ​​theoretically good, I have not observed such things. At higher power (600W and more) a sawtooth or trapezoidal waveform appeared, I do not remember too well now - I would have to check it out.

Maybe I can do some discharges from the oscilloscope, I will post it.

Snubbery actually lose some power but 2W resistors can handle it.

On the side of the second snubbery, you only use the diddy or transformer. I used it only at the transformer and this change has improved a lot with respect to the noise at the output.

I have not yet tried to change the R1 resistor, but today I will try to do something with it.

Have you never had problems with squeezing the transformer at low power?

Multiple pulse suppression occurs only with the shutdown function (PIN 10 SG3525)? Is it also at the fault amplifier?

Using the pin 10, it is possible to implement a smooth control of the output current?

Edit:

Do you have a proven material, article or book from which I could calculate the value of the compensator for transconductance mode?
« Last Edit: June 11, 2019, 09:06:47 am by paladyn »
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #12 on: June 11, 2019, 09:28:19 am »
xavier60

I tested different feedback configurations and sometimes I could see high-frequency sinusoidal waveforms, but this is only when testing on the values ​​of elements that deviate a lot from those calculated in accordance with the book. On the other hand, for the calculated values ​​theoretically good, I have not observed such things. At higher power (600W and more) a sawtooth or trapezoidal waveform appeared, I do not remember too well now - I would have to check it out.

Maybe I can do some discharges from the oscilloscope, I will post it.

Snubbery actually lose some power but 2W resistors can handle it.

On the side of the second snubbery, you only use the diddy or transformer. I used it only at the transformer and this change has improved a lot with respect to the noise at the output.

I have not yet tried to change the R1 resistor, but today I will try to do something with it.

Have you never had problems with squeezing the transformer at low power?

Multiple pulse suppression occurs only with the shutdown function (PIN 10 SG3525)? Is it also at the fault amplifier?

Using the pin 10, it is possible to implement a smooth control of the output current?
I used pin 8, the Soft Start pin for fast current limiting. It gives smooth current regulation but it's not accurate. It is important for the CT to have lots of inductance so that its waveform has a rising slope. If the slope is falling, the fast current limiting wont be smooth.
 A slow and more accurate CC loop can then take over. It can be made to reduce the CV loop's reference voltage.

"LATCHING PWM TO PREVENT MULTIPLE PULSES" prevents the steering flip-flop toggling multiple times  if switching noise gets into the error amplifier, and ensures no more than one drive pulse per phase per half cycle. https://www.st.com/en/power-management/sg3525.html
If too much switching noise gets into the error amp, it still can upset the loop causing rattling or squealing sounds.

Mine makes a small noise that can be heard when the cover is removed when the load is very small at low voltage. This is only because the SG3525 cant go below a certain duty cycle causing the drive pulses to become intermittent.

I did use a DIY transformer with bifilar  center tapped secondary. I use a 39Ω and 2.2nF snubber across the outer legs to reduce 60V peak ringing down to a 20V spike.

Edit: maybe the Shutdown pin can be used, it's able to discharge the soft start capacitor also. But it might need a good logic level pulse to work properly.
« Last Edit: June 15, 2019, 12:42:58 am by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #13 on: June 11, 2019, 09:42:21 am »
Did you measure how much noise you have on the output?
From what is known, current control requires much less elements than voltage control.

Please write in which ranges you can regulate the output voltage?

Do you have current regulations, but more accurate?
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #14 on: June 11, 2019, 10:00:00 am »
Did you measure how much noise you have on the output?
From what is known, current control requires much less elements than voltage control.

Please write in which ranges you can regulate the output voltage?

Do you have current regulations, but more accurate?
It regulates from 3V to 33V
I dont actually have a slow regulation loop for the SMPS because the linear post regulator mainly does the current limiting.
I cant remember what the output noise is, It wasn't much. There are two types of noise. The 80KHz ripple which is easy to filter.
And the high frequency hash, bursts of damped oscillations caused by the fast switching of the MOSFETs and output diodes. This hash gets into everything and can be difficult to deal with. Good PCB layout helps a lot.
« Last Edit: June 11, 2019, 10:18:37 am by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #15 on: June 11, 2019, 10:30:56 am »
Do you have a proven material, article or book from which I could calculate the value of the compensator for transconductance mode?
This might mean something to someone, http://www.ti.com/lit/an/slva662/slva662.pdf
From experimenting with mine I found that the Type III compensator was not necessary for good stability, it gives faster response than the Type II.
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #16 on: June 11, 2019, 10:33:33 am »
...
I have a problem matching the feedback of the SMPS converter itself. At low power (100W) the transformer starts to beep loudly. Perhaps this is because the transformer has not been stuck together and the windings fall into vibration?
...
I also noticed that the current feedback loop works worse than the voltage loop. The only thing that is different in these loops is an additional operational amplifier INA213, which strengthens the shunt voltage. ...

All this suggests that noise is coupling into the feedback pathway, causing instability. The most likely source of such noise is switch and/or diode junction capacitance resonating with stray/leakage inductance at device turn-off. Besides snubbing the likely culprits, good board layout is the ideal cure.
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #17 on: June 15, 2019, 11:00:00 pm »
The op's DSO photos can be found here, https://www.edaboard.com/showthread.php?384844-Half-Bridge-0-30V-0-20A-Feedback-loop-problem&p=1653014#post1653014
There appears to be 100Hz ripple rather than loop instability.
Although it doesn't explain the transformer squealing sounds.
As with my experiences, all that hash makes  SMPS  debugging messy.
« Last Edit: June 16, 2019, 01:25:33 am by xavier60 »
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #18 on: June 17, 2019, 06:25:24 am »
xavier60

Yes, this is a parallel topic to get a broader spectrum of information. At edaboard, I added some graphs of how the output and input waveforms look like.

The most confusing thing for me is that the theory was described in the book

Switching Power Supplies fromA to Z - Sanjaya Maniktala and article https://www.biricha.com/uploads/8/9/8/0/89803127/foundations__part2.b__2.pdf


is quite agreeable with each other. Despite the calculation of coefficients according to the patterns, the output curves are quite poor. I do not know if the problem lies in calculations or maybe somewhere in hardware.

edaboard and eevblog are the last places I ask for help. Despite the size of the internet, people just do not do such things. Today, he goes to the store and buys a regulated regulated power supply, and does not sit at home and combines how to do it.
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #19 on: June 17, 2019, 08:32:11 am »
Although I got my project to work properly, I did the  compensation by trail and error. I found that the values were not too critical.
I don't understand the formulas and calculations.
Mine runs at 40Khz also. The inductor is about 80µH and 6000µF of output capacitance.
While I was debugging mine, I directly connected the live side ground to the output side ground, I was powering through an isolation transformer of course. This greatly reduced to amount of hash. Later, I mounted the PCB on a metal plate and made the grounding connections to the plate. I can't remember how much this helped in reducing the hash.
 I'm still curious to know the result of removing R1.
The 100Hz ripple seems to indicate too little gain near 100Hz.
Check for missing pulses by doing single shot captures.


 
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #20 on: June 17, 2019, 09:21:21 am »
xavier60

If I did not write about R1, I'm sorry, I had to forget.

I did a few tests, I tried without R1 resistor and with a few hundred ohms and with very large> 20KOhm but each time the effect was much worse than the calculated resistor

I think I know why the transformer is squeaking but I do not know why this is happening. I checked the symmetry of the pulses on the primary side of the transformer using an oscilloscope and I came to the conclusion that the transformer lost pulse symmeters. Sometimes, 260us are missing the upper half of the pulses while there are impulses at the bottom. Of course, my Rigol 1074Z oscilloscope does not allow very accurate measurements.

It seems to me that when there is a lack of impulses, then it may come to sub-mapping the core. The only question is why SG3525 allows such behavior?

« Last Edit: June 17, 2019, 10:01:25 am by paladyn »
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #21 on: June 17, 2019, 09:48:05 am »
I tested mine. The SG3525 can't output pulses shorter than 600ns.
When a very low duty cycle is needed, it skips pulses. Sometimes randomly, making a hissing sound. Other times there it is a regular pattern causing a small squealing sound. The drive stays mostly symmetrical.
Actually I prefer that the IC does this instead of outputting needle like pulses.
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 

Offline paladynTopic starter

  • Contributor
  • Posts: 42
  • Country: pl
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #22 on: June 17, 2019, 10:37:44 am »
Add simulations in the LTSpice SG3525 in the enclosure along with a feedback loop. According to the simulation, everything is ok, but in reality it is not.
 

Offline MagicSmoker

  • Super Contributor
  • ***
  • Posts: 1408
  • Country: us
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #23 on: June 17, 2019, 10:41:45 am »
Add simulations in the LTSpice SG3525 in the enclosure along with a feedback loop. According to the simulation, everything is ok, but in reality it is not.

Sigh... See my post #16. When SPICE says your circuit works fine but the real world shows it most certainly does not then the difference is in the parasitics (stray C and L) and layout.
 

Offline xavier60

  • Super Contributor
  • ***
  • Posts: 3034
  • Country: au
Re: Half-Bridge 0-30V, 0-20A Feedback loop problem
« Reply #24 on: June 17, 2019, 11:05:25 am »
Output ripple should  only contain 80Khz. A small amount of this getting to the PWM compactor shouldn't cause a big problem.
If 40Khz gets in, then it will cause a big problem.
It can be difficult getting output voltage from the output capacitors to the input of the error amp without picking up interference.
I suggest getting the voltage from the output capacitors to the input of the error amp through a temporary common mode choke.
Try a 10 turn or more bifilar winding on a core from a mains hash choke.
HP 54645A dso, Fluke 87V dmm,  Agilent U8002A psu,  FY6600 function gen,  Brymen BM857S, HAKKO FM-204, New! HAKKO FX-971.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf