phil99 said  

As with FET cemetery's issue a DC ammeter in series with the primary may help see what is happening.

I have a DC amp meter measuring current in the Primary?

Any attempt to use a resistive ammeter will fail, 20kHz components and hard switching noise levels are extremely high, only a differential probe combined with a fully isolated DC hall sensor will give a clear display of current.

And, the current sensor is indicating the beginning of saturation.

I'm thinking Phil's thought is to put a dc ammeter in the ac primary cct?

Should read zero unless there's an imbalance in the ac drive?

I accidentally posted in the middle of an edit from a couple of posts ago and am pasting it completed here as it is now on the previous page and would probably get missed

  wiseguy said  

  KeepIS said  

A VERY quick Look at Induced buzz:
Put a 1uf across the AC output and the spark does the same thing, it stops the buzz.

Place a microphone near the toroid, if it hears a buzz, close a relay that applies a momentary capacitor across the secondary lol  

I would have expected the momentary capacitor would have a 50% chance of curing or continuing the buzz. Is it a resonance issue with the main components (culprits) being transformer choke and output capacitance and adding the other capacitance reduces the Q (detunes it) back to the edge of stability?

I have not experienced this issue exactly as described. When my Toroid has a buzz or hum episode it drifts in and out at random but it never stays for any length. Fascinating that a momentary capacitor connection will "fix" it.

Maybe in software at idle we could employ an MPPT - minimum power point tracking that can toggle (suppress) one or two of the narrowest SPWM bits on the appropriate half sine to keep the core magnetics (in a dynamic loop) balanced near to the mid point.  Maybe the noise - which does not appear to do anything bad - and only occurs at no/low loads should just be ignored? We might kill the patient whilst trying to cure it.

Edited 2025-05-08 08:53 by wiseguy

  Quote  

I'm thinking Phil's thought is to put a dc ammeter in the ac primary cct?
Should read zero unless there's an imbalance in the ac drive?

Yes, that was the idea and as pointed out by KeepIS the 20kHz issue means it would need to be an old analogue type.

An automotive centre-zero moving magnet type would do. They don't use a shunt, just connect in line. Not very accurate but that doesn't matter for this.

It would likely indicate the small positive hump detected at zero crossing seen by the DC Hall sensor.

The buzz is independent of the Nano AC regulation loop:

Starting with an induced Buzz at 53V supply in Test Mode.

Very slowly lowering the DC voltage will hold the buzz down to 33V input, 99% PWM and 7 volts DC below AC regulation.

Slightly below 33v, Buzz stops and cannot be induced at these low voltage for obvious reasons. The small current hump causing the buzz coincides with ONE side of the AC zero crossing cycle.

Controller in RUN mode:

Sometimes a slight buzz appears during soft start ramp-up and then disappears, at PID handover there is no Buzz and idle is quite.  

Again, a sudden DC disturbance will induce a Buzz, restart inverter and all is quite again. However the inverter is more resistant to Buzz from small DC step changes when the Nano is in PID control AND AC is regulated.

@Wiseguy:

I agree with your thoughts, I wonder if we are just trying to heard cats again?

I can only speak for the Main Inverter: I would not change a single thing, the layout, wiring and assembly is the result of eliminating all of the unforeseen errors and mistakes I thought I have allowed for in my first few inverter builds.

The Test bench unit is a lash up, controller way to close to the power board, choke cabling close to the controller board, slightly incorrect chokes, Toroid not wound correctly and more, but even then it runs quite well in this compromised state.
.
Edited 2025-05-10 15:59 by KeepIS

Chokes and Kinks

As mentioned on another thread, low inductance below 20uH total can really start to push Idle power up.

As the TOTAL inductance moves down from around 80µH, Idle current starts to slowly increase, the value of 42µH (2 x 21µH) was chosen because it gave the best wave shape and less ringing under "very high" current loads, it also reduced noise from Harmonics/Bad loads reflecting back into the Toroid, far better than a higher inductance "most of the time".

This is why trying to select a choke value based around Idle and low power waveforms is not such a good idea.

Idle power is slightly raised and Idle waveforms aren't quite as pretty with a "Total" of 42µH Inductance, however I chose to keep the FETS safe at very high currents over having pretty images at Idle. And a 42µH choke is still plenty to buffer the FETS from high Idle step Front switching currents (Idle current is still relativity low).  

Way back when almost every Inverter had the typical wiggles around zero crossing at Idle, modelling and later investigation concluded that some of this was from a slight ringing as the first PWM pulse-front hit the toroid after zero crossing.

Higher Inductance chokes, usually Ferrite, were often used back then and I found these could make the wiggles worse, and often masked Toroid buzz. The slight ringing will make a buzz, but even back then a few Tech builders were also noticing a slight rise and fall in Toriod/choke sound at Idle, sometimes taking from seconds to a few minutes to appear, it's worth noting that these designs were very early EG8010, a lot different to what we are playing with here - so IMHO this is nothing unique or new to the NANO.
 
When I induce a saturation buzz in the test inverter, I do not have a kink or wiggles in the AC waveform, I guess it depends on how deep you push the toroid into saturation on the offending half cycle before the AC waveform reflect this.

In any case the Toroids do not appears to move very far into saturation in this situation, apart from slightly higher Idle power and less efficiency, it does not seem to hinder the Inverter in any other way AFAIKT.

Various technical writings suggest variation in hardware switching devices, FETS, Optos, Drivers etc will cause slight timing uncertainty in H-bridge symmetry, I would imagine timing imbalance would vary with temperature as well.  

Dead time values can introduce harmonic distortion and give the appearance of a kink or wiggle around zero crossing on the AC waveform, usually on both zero crossing points, a lot of the early CRO screen captures had this, likely from being overly cautious (with good reason) with dead time values back then, and accompanied by a buzz or hum.

I personally feel that any attempt to play with SPWM generation in order to overcome this random noise might likely end up creating an even bigger problem sometime in the future    But that's likely just me being overly cautious.
.

Footnote added 2025-05-11 18:56 by KeepIS
.


Should read "and often masked Toroid buzz because of the noise made by the Ferrite choke"
.

  KeepIS said  

Inverter idling, I force a small sudden jump in DC volts, buzz begins and stays there, assume DC voltage step is moving the flux to the beginning of saturation.

A: Small current spike has appeared coinciding with AC zero crossing. (start of saturation)

Interesting.  Thanks for doing the test.  For clarification, how are you measuring the current spike?  Using the 700A Hall Effect battery current sensor in your build?

I have many of these isolated Hall current sensors, the three in the main Inverter are permanent.

For these tests with the Bench test Inverter, a 200A unit was used for measuring current flow between the Toroid and the Power board. It's more a small half cycle mound rather than a spike, obviously amplitude depends on how far into saturation the toroid is driven and I could not push it very far in these tests, enough to make an constant audible buzz though.