Full Bridge SMPS lost me my career in electronics...how solve?

[Flyback]

Hello,

I have difficulty getting work in electronics now because at a previous company where I worked as an Electronic Engineer in 2012, I revealed a design mistake in the company’s 250W Full Bridge converter. The designers felt that the mistake was merely a “tolerance issue”, and that I should have just kept quiet about it.
After I revealed the design error, the Engineering Manager was extremely angry with my boss (the Chief Engineer). The Chief Engineer was then very unhappy with me because he felt that the problem was simply a “tolerance issue” that would never have ‘reared its ugly head’. He felt that I should not have gone forward and revealed the error.
This Chief Engineer speaks regularly with recruitment consultants and of course provides my reference for work at this company (I don’t work there any more). As you’d expect ,my reference from this Chief Engineer is always absolutely terrible. I thus now have difficulty getting work in Electronics.

What should I do?

The Full Bridge converter in question is shown in two pdf’s below. I had to make it in two parts because otherwise you would have been able to see the company name in the schematic title window.

Spec of Full Bridge Converter:
P(out) = 250W
V(in) = 48V
V(out) = 26V
F(sw) = 300KHz.
Isolated secondary
Isolated feedback by optocoupler/TL431
Current mode.

Details of design Fault.

The design fault concerned the feedback loop.
The converter operated fine in the lab into an electronic load when the optocoupler , “ISO1” was CNY17-X by either Fairchild or Vishay. However, when this optocoupler was the Fairchild one, then the converter became unstable when operated with its actual intended load (the intended load had quite a large amount of input capacitance, obviously unlike the electronic load)
The instability precipitated itself by making the power supply output voltage so full of oscillation, that the load could not operate at all properly, and the whole equipment that was supplied by this SMPS mal-functioned.
(by the way, I may have got Vishay and Fairchild the wrong way round in the above explanation, and I cant remember which specific version of “CNY17-X” the optocoupler was).

Anyway, before the fault was revealed, I was asked to do tests on this Full Bridge Converter. (In the lab, using the electronic load)
I noticed on the schematic that the feedback loop appeared to be very unusual (you can see this yourself on the schematic).

Unusual-ness of the feedback loop:
In the secondary, following after the split secondary diodes, there is 40uF of ceramic capacitance (C47, C48, C49 & C50)…as you can see, the node with this capacitance is called “MAIN_PSU”.
Going downstream, there is then a 680nH inductor (L4).
Then there is a 68uF electrolytic capacitor, C54. This electrolytic capacitor has a 220 milliOhm resistor (R37) in series with it.
The feedback node is the “MAIN_PSU” node, and this can be seen feeding into the “REF” pin of the TL431 (U8)

…I am sure you would agree with me that this feedback loop architecture is very strange. Taking the TL431 feedback from the “MAIN_PSU” node is very strange indeed. Also, the resistor in series with the main output electrolytic capacitor, C54 is very unusual and necessary.

I would say that L4 should be massively less than 680nH, and preferably not used at all. I would also say that the 220milliOhm resistor, R37, in series with the output capacitor , C54, is totally unnecessary, and serves only to increase the possibility of instability.
The TL431 feedback should certainly not be taken from upstream of L4. If anything, the “MAIN_PSU” node should only be used as a kind of “fast lane” feedback via R47 & C63.

Anyway, despite the feedback loop’s unusual-ness, the converter was seen to be stable when running into an electronic load in the lab.
However, as mentioned, when the optocoupler “ISO1” was changed for a same-part-number optocoupler by Fairchild instead of Vishay, and the load was changed to be the actual load, then the Full Bridge converter became unstable. The actual load had a significant amount of input capacitance, which obviously the electronic load did not have.

Chief Engineers response
Initially, I approached the Chief Engineer with my concerns about the feedback loop. He told me that the converter was seen to be stable in the lab when run into the electronic load, and that I should therefore just keep quiet about the feedback loop, because there was no instability seen at that stage.

In these early days, we did not have access to the actual load which this Full Bridge SMPS would be powering, since the actual load was itself being developed, and not only that, it was a top secret device (due to it incorporating the company’s IP) , and so the company were keeping it under lock and key. Thus we just used the lab’s electronic loads.

Anyway, I was still not comfortable with the feedback loop. However, my feedback loop calculation for that particular loop configuration reported a phase margin of 45 degrees, so I could not formally prove to the Chief Engineer that the Phase margin was inadequate. Of course, I had neglected to include the load capacitance as I had no access to the load....so I couldn't have known that the L4 / Load capacitance resonant frequency of 8700Hz was just under the crossover frequency of 13,300Hz.


Anyway, I decided to send the schematic (in confidence) to a fellow who I presumed to be a friend. This friend was an Applications Engineer at a Semiconductor IC company. I Requested if this Applications Engineer could assist myself with repeating a feedback loop calculation for this Full Bridge SMPS.
Anyway, this so called friend did not give me help with the feedback loop calculation, but instead, he immediately contacted my Engineering Manager where I worked, and said that he had seen the schematic, (also said that I had sent it to him) and told the Engineering Manager that the feedback loop was a great cause of concern to him.
This applications engineer proposed to my Engineering Manager that the Full Bridge SMPS be operated into the actual intended load (instead of an electronic load). He also proposed that the optocoupler “ISO1” be changed for a same-part-number optocoupler but made by Fairchild. (our PCB assembly house actually had the option of using either the Vishay or the Fairchild optocoupler part, but were just by chance using the Vishay part until the Applications Engineer came into the picture)
My Engineering Manager did this, (changed the optocoupler “ISO1” for the Fairchild version of it) and the Full Bridge SMPS went unstable.

Several hundred units (of the overall equipment that the SMPS powered) had been delivered to initial product test sites, and indeed, some of these were showing the instability and not working. (the ones that weren’t working had the Fairchild version of the optocoupler in them)

However, the units had not yet been delivered to actual customers, so despite hundreds of thousands of pounds being wasted, it was not the disaster that it could have been if the fault had not been detected till later on.

Anyway, the result of this was that the feedback loop was completely changed to make it more standard, and it then worked fine.

However, my Chief Engineer got severely reprimanded by the Engineering Manager for having let an unstable SMPS go into initial production and site testing.

The Chief Engineer came to find out that it was me that had contacted the Applications Engineer about this issue.
This Chief Engineer was furious with me for “going behind his back” and discussing my concerns with others. The Chief Engineer felt that the Fairchild version of the optocoupler was only introduced by the Engineering Manager due to my having spoken about the feedback loop. This Chief Engineer thought that if this optocoupler had remained as the Vishay version, then the instability would never have precipitated, and everything would have been fine.
The Chief Engineer was very, very unhappy with me indeed. This Chief Engineer had lots of “buddies” in the company, and encouraged all of them to express disgust at me whenever possible. –I found that his ”buddies” would swear at me under their breath as I walked past, etc etc, and I often found that equipment such as soldering irons etc, would mysteriously disappear off my bench when I returned from coffee break etc.
Also, after I left the company, this Chief Engineer (who is regularly in contact with recruitment consultants) has made sure that my reference is as bad as possible.

What should I do about this? This Chief Engineer’s opinion is widely respected by recruitment consultants and hiring Managers in other electronics companies.

 

[picbits]

What country are you in ? In the UK, companies who have given unjust references have been taken to court and sued by the applicants.

 

[Nigel Goodwin]

Thread moved to more suitable forum - moderator.

 

[Flyback]

At the end of the day, its very difficult to stop someone having a "private conversation" with a hiring manager or a recruitment consultant.

 

[picbits]

It doesn't matter - if you get a company you're friendly with to apply for a reference and get a bad one back (unjustly) then you could employ a solicitor to sue for appropriate lost earnings.

You still haven't said what country you're in.

 

[Joe G]

I've already seen good ex-employees who got "shunned" and made to quite, get bad references and the management just laughed about it, also crappy ones get excellent references just to get rid of them? and yes, the management also just laughed and made jokes about the same, It's a screwed up system, tools missing, sabotage of programs, setups and tools changed, it does happen alot which is a shame.

 

[Flyback]

Incidentally, has anyone ever seen a feedback loop like the attached in the first post?
Sorry I cant say which country.

 

[samovar]

Incidentally, has anyone ever seen a feedback loop like the attached in the first post?.

You can look at it this way: The SMPS ends at your MAIN PSU point. Then there's a low pass filter after the SMPS, which is not related to the SMPS.

If some isolators work and others don't, it looks like a bad selection of the isolator. IMHO, the instability is not because reading is taken before the choke.

But I don't have a work neither. Don't listen to me

 

[Flyback]

Some of the readers here will instantly recognise what equipment that SMPS is going to be driving. It costs millions. Why in such an expensive system didn't they just put in a cheap cascode circuit into the optocoupler feedback?.....it completely releases you from the 'terrorism' and havoc that the opto pole tolerance can inflict on the feedback loop, and it costs just cents.


Cascoded optocoupler feedback:
https://www.electro-tech-online.com...coupler-feedback-has-any-disadvantage.134059/

 

[rogs]

It doesn't matter - if you get a company you're friendly with to apply for a reference and get a bad one back (unjustly) then you could employ a solicitor to sue for appropriate lost earnings.

You still haven't said what country you're in.

From the first post..... " However, the units had not yet been delivered to actual customers, so despite hundreds of thousands of pounds being wasted, it was not the disaster that it could have been if the fault had not been detected till later on."

Not many countries have 'pounds' as their currency. Put your bet on the UK....

 

[KeepItSimpleStupid]

Sounds like you wer paid too much or wee not well liked.

There is no way anyone would sign off on anything that wasn;t operating with the load intended. I designed an I-V converter for in-house use that worked fine for everything but the calibration devices. Too much capacitance. I had to then isolate the capacitance of three similar stages. This didnt show up unti actual testing.

We had systems that HAD to use ELH lamps from GE. Sylvania would not work because of the differenc in spectrum.

It happens.

The only real thing you could have done is specified what brand optocoupler had to be used becasue of lab testing.

The ONLY way you can help yourself is to say, that I believe will get an unfavorable reference from my past employer and if you would listen, I'd like you to hear my side of the story. In simple terms, another namufacturer's part rendered the design unstable and "my reference" sent the unsable similar par art into pre-production and that's why I'm not working there anymore.

It's not about the design.

Now USUALLY, you should contact the person being named as a reference first, before using them.

Find another reference and get permission from that reference.

Application Engineers are supposed to help you and that is a resource which I think your entitled to use.
I have no idea way the funky people feedback loop.

Taking them to court almost seems like an option: An employee deliberatey substituting a part that was shown not to work.

 

[Flyback]

I cant tell you which country...but of course, I haven't exactly given away their IP here....after all, the above schematic is a faulty one. The fault being one that ultimately cost my career. And in any case, the Full Bridge stage is a basic , known stage in electronics and not anyone's IP. You could get that kind of schematic from loads of application notes. I cannot understand why no cascode was used......after all, the overall equipment costs millions, and a cascode stage is just a few cents.

Optocouplers are lousy, and cannot be trusted in feedback loops without cascoding, unless you want a very slow feedback loop bandwidth.

The only real thing you could have done is specified what brand optocoupler had to be used becasue of lab testing.

...the opto was just an opto, and awful like they all are...........the feedback loop should have been designed to be ok with either Vishay or Fairchild opto's

 

[ericgibbs]

hi F,

You dont say what position you held in the product development team.?

E

 

[Flyback]

You can look at it this way: The SMPS ends at your MAIN PSU point. Then there's a low pass filter after the SMPS, which is not related to the SMPS.

that's interesting, because I actually went through a feedback loop calculation like that...and came up with a crossover frequency of 13351 Hz and a phase margin of 45 degrees, which is tight but not completely unacceptable.. of course, my problem was that I was not considering the extra 420uF of load capacitance because I had no access to the load as it was a secret. (by the way, this was with an opto pole frequency of 14000Hz, and as you know, the tolerance on this figure is impossible to know, as theres no way of calculating the tolerance of optocoupler parts as the datasheets don't allure to it.....you can calculate the opto pole frequency for the one part that you have in the lab, but whether that's the same as in a different batch is totally unknown.....and this again harps back to the poorness of opto's in feedback loops, and why we should use cascodeing of the opto.




With the extra 420uF capacitance of the added load, this resonates with L4 at a frequency of 8740Hz. I believe this is not far enough above the crossover frequency, and resulted in the oscillation. The inductor could actually be heard singing.

Hi E...
My position in the company....I wasn't there long enough to really ever have a position......it was a huge company and I sort of paled into the fabric.

 

[blueroomelectronics]

The OP is a mystery wrapped in an enigma.

 

[Flyback]

What would you say is the minimum resonant frequency of the output filter here, which wouldn't give oscilations when the 420uF load capacitance was added.?

What value of optocoupler pole frequency did you base that on and why?

 

[tcmtech]

The OP is a mystery wrapped in an enigma.

And then some.

I have difficulty getting work in electronics now because at a previous company where I worked as an Electronic Engineer in 2012

Never would have guessed him to be an EE either. In fact I would have put that as being about as plausible as me being a space shuttle pilot.

Plus after experiencing the isotera fixation thread I have other far more plausible suspicions as to why he really got blacklisted.


As far as the schematic goes and its related feedback circuit to me it looks like the standard SMPS system for a common computer power supply but without the extra voltage outputs.

Basic variation of the standard common industry design for SMPS feedback loops as far as I can tell for a quick look.

 

[jpanhalt]

@Flyback

The Hard View:

You made a mistake and should have been fired. If your immediate supervisor/manager had not fired you, I would have considered firing that supervisor. First, you made your concerns known. That was OK. Then you didn't take "no" for an answer and persisted, despite your very junior position. Finally, you shared the company's intellectual property with an outsider. That was not your property to share. Learn from that outsider. He did the correct thing and contacted the company.

Redemption:

You are young and made a mistake. No reasonable person will hold that against you for long, if you show you learned from it.

1) Write a sincere letter of apology to the supervisor who fired you. Be humble and admit your mistakes. Carbon copy the manager.

2) Never again share that whole story. It does not make you look good. Ask yourself, why would a potential employer want to know what is wrong with your previous employer? The potential employer is only interested in why he should hire you.

3) If the issue of your previous job comes up (which may happen in the context of explaining a gap in your background), just say you were young, and it was not a good fit. You can add that you have learned from that experience, but be prepared for the obvious follow up question to that. Try not to mention the disclosure of IP, but if you do, then add that that was your biggest mistake.

4) You will be asked what you have to offer the new company. Have a positive answer based on what you know about the prospective employer. Be sure to research that employer before your application and interview.

5) Don't even consider suing that previous employer. The ill advised disclosure of its IP was certainly contrary to policy and was likely an offense that was specifically listed as a cause for termination. Even if that were not the case, filing a wrongful discharge lawsuit could ensure this is the end of your career in electronics. The risk/benefit is just not worth it.

Good luck.

John

 

[Flyback]

I wouldn't dream of trying to sue.............The feedback loop was dodgy, but I was initially told that the smps was out in the field on trials and that there were no problems driving the real load.........what I didn't realise was that these smps's were using the Vishay optocoupler.
I should have realised though, that optocoupler tolerances are shocking and that 45 degrees is nowhere near enough phase margin for a feedback circuit with an optocoupler in it...............and that holds even if theyd only ever used the Vishay optocoupler.

I saved the company million of pounds (-though indirectly, since it was really the apps engineer that was the hero). The Director (the one above the engineering manager) of that department was not very impressed that I had shared the schematic with the apps engineer......the Chief Engineer was trying to get them to haul me over the coals for doing that.

if I ever get to work in electronics again, and I see a 45 degree phase margin smps with opto feedback in prototype test, then how should I deal with it?

Supposing that they've never done gain/phase analysis with eg a AP300 frequency analyser (by the guy that writes switchingpowermagazine......I forget his name.......its something like "tony tinder"..........both forname and surname start with the same letter.)

Anyway, what if I ask for batch to batch opto pole frequency analysis on the chosen opto and the boss refuses.....do I go straight to the company owner, and tell him that not enough has been done to assure smps stability in the face of poor optocoupler tolerance?

What do you say is the minimum acceptable phase margin for an smps with optocoupler feedback? (without the opto being cascoded).

 

[jpanhalt]

Please step back and look at the situation. I was not responding to any electronic question. There are a lot of us here who were once, bright, young, brash, and eager. In my field, we were referred to as "young Turks" by the established individuals. (That was not meant as a derogatory term.) We still had to follow the rules. It is not against the rules for a company to do dumb things and waste millions of dollars.

It is against the rules to violate company policy and to go outside the chain of command. Those were your mistakes. From an electronic standpoint you could be right as rain, but that is not what is important when you apply for a job.

As long as this unpleasant period is bugging you, it will influence your performance on a job interview. Put it behind you and move forward.

I truly wish you the best of luck.

John