Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.
tinyt Guru Joined: 12/11/2017 Location: United StatesPosts: 438
Posted: 03:14pm 27 Jun 2018
Copy link to clipboard
Print this post
I found the 54v 5.2A DC-DCs, I did not know that I bought 20 of them and I don't remember when, it must have been cheap at that surplus store. I will test them and if all are good and if the model I have has current sharing function, I can connect them in parallel for a total of 104 amps. But this will be another project.
Back to testing, I increased the choke number of turns from 8 to 14 and the waveform distortion became worse. 54V current draw remains the same at 0.18 amp.
I also 'scoped the junction of the transformer and choke. Here it is, looks like an alien's heartbeat to me.
I changed back to 8 turns choke and 'scoped the junction again. Here it is, cousin alien.
I then changed to 4 turns choke and the 115vac output has less distortion than with the 8 turns choke. Here it is. But the 54V current draw increased to 0.22 amp.
Without the choke, the current is 0.31 amp.
I varied the VFB trimpot to the CW and CCW stops, the 108 usec gap of the SPWM from zero crossing did not change. The AC voltage and 54V current readings at the stops are: 103vac - 0.16 amp. DC. 123vac - 0.22 amp. DC.
I set the trimpot again for 115vac output and then lowered the 54vdc supply. The 115vac remained steady until the supply reached 40vdc, at which point it started to drop. At about 34vdc, the EGS002 started its undervoltage off and soft-start cycle. End of test for now. Next I will remove the 10 ohm series resistor and connect the bulk capacitors. When the toroid transformer is finished, I will play again with the choke.
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 07:53pm 27 Jun 2018
Copy link to clipboard
Print this post
Tinyt, Using an oscilloscope that takes short samples, of a rapidly switching waveform can create some ugly illusions and be less than helpful in trying to work out what is really going on. An analog CRO may be better, but only slightly better. The irony here is that the better your oscilloscope bandwidth is, the more difficult it will be to actually see anything.
One effective solution to this is to use an averaging circuit (a low pass filter) consisting of a series 4K7 resistor and a 100nF capacitor down to ground between what you are monitoring, and the tip of your oscilloscope probe. That should remove most of the high frequency PWM so you can see the shape of the real averaged waveform hidden underneath all the "fuzz".
Put the larger choke back in. A higher inductance choke is a good thing, even though it may increase the waveform distortion, it will isolate the PWM switching frequencies from the transformer, and both the transformer and mosfets will benefit mightily.
The waveform distortion is caused by a resonance in the transformer, and every transformer will be different to how it distorts the waveform. Harmonics in the waveform can excite this resonance and generate kinks or wobbles in the waveform, especially around and just after the zero crossing point.
The choke decouples the transformer from the mosfet switching bridge, the more inductance the choke has, the more the transformer can run wild and do crazy things to your waveform. Taking out the choke, or reducing its inductance is definitely the wrong way to go. You need to fix the resonance problem at the transformer.
The transformer resonance can never be eliminated altogether, but it can be brought down in frequency much closer to 60Hz and actually become helpful.
If you hypothetically tuned the transformer to exactly 60Hz that would create another problem of a huge resonant buildup of energy, especially if there is no load on the inverter. The voltage feedback is going to have a very difficult time trying to tame that huge buildup of resonant energy.
The "trick" is to tune the transformer to 90Hz by fitting a suitable tuning capacitor directly across the secondary winding.
By making the resonant frequency x1.5 the operating frequency, that is low enough to give excellent and increased attenuation to any residual PWM in the output.
But much more importantly, any gradual buildup of resonant energy at 90Hz will be out of phase with the next coming cycle at 60Hz, and tends to be damped out.
Try to tune as close to 90Hz as possible, ideally within 1 or 2 Hz.
**WARNING** Fitting a large capacitor directly across the transformer secondary WITHOUT first having a decent NON SATURATING choke may potentially create dangerous current spikes in the mosfets. So get the choke problem sorted out first, even if it makes your waveform look like total crap.
Don't worry about the kinks and the wiggles initially, just get a decent man sized choke that is still going to have plenty of inductance even at full flat out power and beyond.
Then tune your transformer to 1.5 times the inverter output frequency. 90Hz in America 75Hz in Australia. The weird distortions on your waveform should all go away.
There may still be a slight kink right at the zero crossing, that is also an artifact introduced by the transformer. Its caused by using unipolar drive, and having some unfortunate capacitive coupling between primary and secondary. An electrostatic screen should fix that problem if anyone wishes to try it.
But make absolutely sure your new electrostatic foil layer does not add an unsuspected shorted turn. That would be most unfortunate...
Its much easier and more effective fitting an electrostatic screen to a normal E/I transformer. You only need one turn to do it. A really big toroid needs a screen with many tuns, and each turn can generate a volt or more of voltage. So use wide foil and few turns if at all possible, and ideally ground the centre of the foil winding, not just at one end. A screen should help a lot, but it needs some thought if its going to be truly effective on a large toroid.
It all sounds rather complicated, and I suppose it is. But if you take great care and play by the above rules, you should avoid most of the problems that a great many people here are still having.
Edited by Warpspeed 2018-06-29Cheers, Tony.
tinyt Guru Joined: 12/11/2017 Location: United StatesPosts: 438
Posted: 08:39pm 27 Jun 2018
Copy link to clipboard
Print this post
Thanks Tony for all the good stuff. So after I have built the toroid transformer, can I measure the inductance of the secondary and use the resonance formula for 90 Hz to get the capacitance needed? Do I need to connect it to the unpowered mosfet bridge?Edited by tinyt 2018-06-29
Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498
Posted: 08:54pm 27 Jun 2018
Copy link to clipboard
Print this post
Changing the choke from 8 to 14 turns is a big jump and had a big effect. Adding or subtracting a turn or 2 to see if you are getting close to the ideal is how I would do it.
This not in the EG8010 Datasheet but is in the EGS002 Datasheet.
There are only 10 types of people in the world: those who understand binary, and those who don't.
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 09:11pm 27 Jun 2018
Copy link to clipboard
Print this post
Yes, that is the CRO input filter circuit. The values are not at all critical.
Best way to test your transformer resonance is to first completely disconnect it from the inverter. Connect it up to an audio oscillator through a series resistor, to one side of the secondary, then ground the other end of the secondary. Connect your CRO directly across the secondary winding.
Try tuning the winding with various capacitors, start with about 1uF.
As you tune the oscillator frequency, a definite peak in voltage will be seen at one particular frequency. Try a few different resistor values, start out with 1K. A lower resistor value will give a higher voltage, but it may not peak so well. A higher value resistor may give a slightly sharper peak, but may become noisy and difficult to see.
A digital frequency counter would be really nice to have for accuracy.
But lacking that, there may be enough mains hum pickup to see superimposed with the oscillator. As you get very close to 90Hz that should become stationary on the CRO trace. That is probably going to be a lot more accurate than the frequency markings on the knob of a really low cost oscillator.Cheers, Tony.
tinyt Guru Joined: 12/11/2017 Location: United StatesPosts: 438
Posted: 09:46pm 27 Jun 2018
Copy link to clipboard
Print this post
Actually here is a picture of the RC probe I sometimes use. The capacitor is 0.1 uF surface mount. I was just lazy and just wanted to know if I can see the difference between the two choke turn configurations thru that noise.
But I like the bandwidth filter of that expensive 'scope I borrowed before.
I will check in my junk garage if I have not yet thrown out the audio oscillator I have and use it to tune the toroid.Edited by tinyt 2018-06-29
wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156
Posted: 10:31pm 27 Jun 2018
Copy link to clipboard
Print this post
Hi Tinyt, I was thinking about your crossover distortion whilst trying to sleep last night and developed a possible theory. Why would someone design a sinewave inverter that had two dead bands 100uSec either side of zero and the answer was that they wouldnt (would they?). Thats why I asked you to vary the feedback pot to see if it changed. Can you please scope the input side of either the FET driver output (24kHz pwm) IR2110 or at the gate resistors 4r7 5r6 - whatever you are using (not the FET side but the driver side) and see if there are some narrow pulses a lot closer to zero than the 100usecs you see at the transformer? Could inadequate gate drive be swallowing the pulses as they get close to zero ie not getting a high enough gate voltage to turn on from ever decreasing narrow pulses.
Warp could be right and it is just a triggering problem causing the pulses to vanish, but if pulses are missing 100uS each side of zero would cause some crossover distortion. Good luckEdited by wiseguy 2018-06-29If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike
Tinker Guru Joined: 07/11/2007 Location: AustraliaPosts: 1904
Posted: 09:31am 28 Jun 2018
Copy link to clipboard
Print this post
Thank you for that suggestion Tony. I must be one of the very few inverter build posters here who *does* have an electrostatic screen on my toroid. Since the donor Aerosharp toroid did have such a screen and the makers of it put it there for a reason I just re used it (testing for short turns of course). This was boo booed here as unnecessary by a well known contributor.
Now I know why my inverter's sine wave has very little by way of zero cross over 'kink'.Klaus
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 10:01am 28 Jun 2018
Copy link to clipboard
Print this post
Its not so much a triggering problem as that a digital oscilloscope just takes very short periodic samples of what is connected to it.
If its looking at extremely short narrow pulses, unless the pulse coincides exactly with the exact instant of sampling, it may be missed completely.
So as the PWM pulses become increasingly narrow near the zero crossing, more and more may be missed, and the oscilloscope assumes there is nothing at all there to see.
Digital oscilloscopes can be very nice to use, and have some wonderful features that an old analog CRO simply cannot compete with. But a digital CRO can, and does, sometimes tell you complete lies due to the sampling effect.
I have both types of oscilloscopes and I much prefer the analog scope for everyday use. But there are odd times when I have to fire up the digital CRO to do something a bit special or unusual that the analog scope just cannot do.Cheers, Tony.
Madness Guru Joined: 08/10/2011 Location: AustraliaPosts: 2498
Posted: 10:59am 28 Jun 2018
Copy link to clipboard
Print this post
You're not Robinson Crusoe Klaus.There are only 10 types of people in the world: those who understand binary, and those who don't.
wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156
Posted: 11:57am 28 Jun 2018
Copy link to clipboard
Print this post
But if that was the case surely not all of the pulses in that 100uSec dead gap would coincide with the exact timing of sampling to miss out on being seen, they are asynchronous events. I would have expected to see at least a pulse or two get through?
I have operated a TDS220 like Tonys for over 25 years (still do) and I cant ever recall not finding a pulse I need to see. Im not trying to be argumentive - I am always prepared to learn ,but when I hear something that is counter to my experience I flag/query it ? However I also agree with "But a digital CRO can, and does, sometimes tell you complete lies due to the sampling effect" you do develop a gut feel of when its lying to you ! I remember when Agilent had a little circuit that they developed to show other digital CRO users how crap their CRO's were, but Agilents CRO had it in spades rock solid - it smelt a bit like the VW pollution software that knew when it was being watched - all a bit incestuous.
I will but out until I get my own boards and get some real hands on experience, was just trying to help tonyt find the cause of a crossover distortion issue.
BTW Warp - your post on the series choke & finding a compromise resonance to the 25kHz and 50Hz was excellent work! Edited by wiseguy 2018-06-29If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike
tinyt Guru Joined: 12/11/2017 Location: United StatesPosts: 438
Posted: 04:04pm 28 Jun 2018
Copy link to clipboard
Print this post
Did more 'scope captures. This time at the pins of the EGS002:
CH1 is at Test1 and CH2 is at Test4. Trigger at CH1 falling. Picture 1.
CH1 is at Test1 and CH2 is at Test4. Trigger at CH1 rising. Picture 2.
CH1 is at Test2 and CH2 is at Test3. Trigger at CH1 rising. Picture 3.
CH1 is at Test2 and CH2 is at Test3. Trigger at CH1 falling. Picture 4.
All four captures above have the scope being triggered continuously. Looks like the SPWM dead time gap at zero crossing is not symmetrical depending upon the direction of the zero crossing of a particular phase of the fundamental.
Being a visual guy, I made a sketch of the relationships of the screen captures. I hope I got this right. The bracketed numbers refer to the picture numbers above.
And then there are these three musketeers popping only once right after power on. CH1 at Test1 and CH2 at Test4. They don't appear at the Test2 and Test3 pair. Scope is set for single trigger mode this time.
I hope somebody else can confirm/or not what I found. Note that the EGS002 I tested as received, is jumpered for 50 Hz and 1.5 usec dead time which I confirmed between Test1 and Test2 outputs. It also has cross-conduction prevention transistors. I will see if I can check right at the EG8010 outputs.Edited by tinyt 2018-06-30
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 08:25pm 28 Jun 2018
Copy link to clipboard
Print this post
Difficult to know for sure.
But I will tell you one thing, none of those oscilloscope traces look anything like what real PWM signals should look like. Those look more like random bar codes !
Where are the flat tops between the tops of the pulses ? Your CRO is really struggling, and trying to join up some scattered sampling dots and coming up really short....
Cheers, Tony.
tinyt Guru Joined: 12/11/2017 Location: United StatesPosts: 438
Posted: 09:36pm 28 Jun 2018
Copy link to clipboard
Print this post
The pictures above are just focusing on the SPWM gaps at the zero crossing, nothing else. Here is an old capture of a 50Hz and the unfiltered SPWM.
For 50 Hz and an SPWM of 23.4 kHz, I think there will be 234 pulses for each 10 msec (half fundamental) if the pulses are equally spaced, less if generating a sinewave that will be filtered downstream. I am trying to understand the spec of my old TDS 220: It has a record length of only 2500 samples. Sample interval is 50S/s thru 1GS/s, I think this depends upon the SEC/DIV setting. The sample interval is SEC/DIV / 250. The display has a resolution of only 320 x 240 pixels. Assuming it can capture an entire 468 pulses, I don't think it can show all details on the screen. Hence, the picture above looks like that
I remember the older analog scope TEK 453 has A and B sweeps where you can magnify a very small section of a waveform and also pan left-right of the waveform. But this TDS 220 does not have it. Unfortunately the TDS 220 is the only one I have. So for the picture above, I just use my imagination and try to picture in my mind the actual pulses that are changing in duration.Edited by tinyt 2018-06-30
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 10:14pm 28 Jun 2018
Copy link to clipboard
Print this post
Its an interesting problem.
My own Tektronix 2430 digital CRO is the very first digital that Tek ever built, and one of the very first commercial digital oscilloscopes to come onto the market. Its a real dinosaur from the early 1980's.
It has a very good and very fast high resolution CRT, the same as in their 400 Mhz analog scopes of the same era, but the sample rate of 100 mega samples/sec on each of two channels is pitifully slow by todays standards.
But it does really bring out the limitations of the sampling method of data acquisition. Its perfect for capturing very slow one time events, and the math capability is excellent. But for capturing fast digital signals on a data/address bus, or looking at something like PWM its absolutely hopeless.
Even the very low end digital oscilloscopes of today now have at least ten times the sampling rate of my old antique digital, but the exact same problems still remain.
What you are seeing on the screen can be very different from reality in some situations, which can sometimes lead to some very wrong conclusions.Cheers, Tony.
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 10:23pm 28 Jun 2018
Copy link to clipboard
Print this post
Glad you liked that. I am afraid very few people really understood what I was on about. Many people here want to stridently argue that a real choke is just not required, and PWM turns into a clean sine wave by pure magic alone. Cheers, Tony.
wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156
Posted: 12:28am 29 Jun 2018
Copy link to clipboard
Print this post
In a simplified way to look at it, these things are an overgrown buck converter with active flywheel diode, a modulated reference voltage being fed through a very stiff transformer into near as dammit short circuit at 25kHz (filter cap) so an output inductor is a good plan!
wiseguy Guru Joined: 21/06/2018 Location: AustraliaPosts: 1156
Posted: 01:07am 29 Jun 2018
Copy link to clipboard
Print this post
Whoops duplicate postEdited by wiseguy 2018-06-30If at first you dont succeed, I suggest you avoid sky diving.... Cheers Mike
Warpspeed Guru Joined: 09/08/2007 Location: AustraliaPosts: 4406
Posted: 01:07am 29 Jun 2018
Copy link to clipboard
Print this post
Yes, YES !! Absolutely. Cheers, Tony.
johnmc Senior Member Joined: 21/01/2011 Location: AustraliaPosts: 282
Posted: 02:08am 29 Jun 2018
Copy link to clipboard
Print this post
I find these posts, most informative and there is much to learn in the world of electronics (inverters) many thanks to all the contributors.
Notice. New forum software under development. It's going to miss a few functions and look a bit ugly for a while, but I'm working on it full time now as the old forum was too unstable. Couple days, all good. If you notice any issues, please contact me.
Print this post
