Warpverter

[Hedges]

DC current draw is rectified AC current draw (less some smoothing by capacitors).
What waveform do you see in your application?

 

[Warpspeed]

Yes, the switching bridges draw current in half sinusoids from the bank of big electrolytics.
Current from the battery is pretty much smooth dc at low inverter loads, with ac ripple increasing in amplitude with higher inverter loading.

The ac output voltage from the inverter is pretty much tied to the instantaneous dc bus voltage on the electrolytics and voltage regulation of the transformers.
A sudden transient surge on the ac side may well be catered for by stored energy in the electrolytics, in which case very little actual voltage correction by the feed forward system may be required.

The electrolytics slow down voltage changes across the dc bus, and give the feed forward additional time to correct.
Its common on commercial PWM inverters to have minimal sized electrolytics for cost reasons.

One of the main features of a home brew Warpverter is the massive (no cost spared!) bank of low ESR electrolytics.
That goes a very long way towards achieving very good step load change performance, both increasing and decreasing.
An ultracapacitor bank would be the ultimate if costly solution.

 

[Warpspeed]

Just had a look at my Hall sensor, its an HSTS21 with +/- 100 amp range.
There are a whole family of these things, including a +/- 500 amp version and a +/-400 amp version.
https://www.ebay.com.au/itm/235319655247?itmmeta=01HVMTRCWV6BHQ1NYGYGMXFP46&hash=item36ca24fb4f:g:lJkAAOSwkP1d5iuw&itmprp=enc:AQAJAAAA4DZFDtc1Q2ir8IgrU/hWxXTWReC0pMDDM/jr56eJM/i6bgxsijQzaUBL1pKYfLXe61c/B9DPFA8Imto031U3iy5uPnbr6gMw3nWNZkXxPFevcj9l8xWflAGUvmEy9qXUEyVRVZiALFzKMKV4V5TVO2NtcWAbDp9rYcpR4+i+v78lgPzxKJouK/G3eBTWuGg26o+zygCqS5mM2twymx50tFpDR0iiZaSQUBt9UW6ulJ21TwxAfvNNs+yKI4p5nstlCUVN2ofdE1vOD4MdNJyEl5q5UuhZ5GP5X48vPVX8Mykb|tkp:Bk9SR8DO4ZrdYw

These use a +5v supply, and they have inbuilt potentiometers for adjusting the +2.5v zero output, and also the output span.

My previous explanation of it having a +/- 2v output span was not correct, its +/- 625mV which is more than enough.

There is also this one, which is the same kind of thing but requires +12v supply.
https://www.directindustry.com/prod/shenzhen-socan-technologies-co-ltd/product-132598-2600179.html

 

[rogerdw]

Thanks guys I've been enjoying the dialogue. Hopefully some of it sticks and I understand it better.

Here's the specs from the range I bought from. I've got two of the middle ones ... 300 amps. I planned to fit the other in the combined outputs from the four charge controllers.

It came from here ... https://www.aliexpress.com/item/1005001367453644.html?

 

[Warpspeed]

Reading through the instructions at the link:

It came from here ... https://www.aliexpress.com/item/1005001367453644.html?

It says (cut and pasted):

Easy installation ano operation: This product is easy to install and use, just plug it into the outlet of the tv or computer.

Ah so !

 

[rogerdw]

It says (cut and pasted):
Ah so !

Haha, yeah I read that too.


Those are the ones that KeepIS used on his inverter build and for his peak current meter. You can see some of his comments here ... https://www.thebackshed.com/forum/ViewTopic.php?TID=15417&P=21#205106

and another post here ... https://www.thebackshed.com/forum/ViewTopic.php?TID=15417&P=21#205172

Pretty sure Klaus used the same ones for his version of the peak meter. He sent me a pcb so I can build one too.

The ones I bought need a 12v supply and as far as I know the O/P varies between 0-4v ... with 4v being 300A in this case.

I just need to power mine up (it's already fitted) and monitor the output pin and compare with the clamp meter. Should have done that last night.


EDIT: On the link for the current sensor, look for any of the "View more" buttons hidden between the adverts and you'll be able to see all the details.

 

[Warpspeed]

Those D-C2T sensors appear to have a zero to four volt output, although that is not exactly clear.

For the Warpverter feed forward current compensation to work, you need to have +2.5v output at zero current, then plus or minus current reading either side of the +2.5v zero.
If I remember, Klaus used an HSTS-400 amp sensor, and I used an HSTS-100 amp sensor which worked very well for me here.

For measuring surges and peak inrush, a digital CRO triggered on the rising edge is what I use.
Some of the better quality multimeters can sometimes have a peak hold function too.

 

[rogerdw]

Those D-C2T sensors appear to have a zero to four volt output, although that is not exactly clear.

Yes I think that's how they work though I need to hook one up to check for sure.

For the Warpverter feed forward current compensation to work, you need to have +2.5v output at zero current, then plus or minus current reading either side of the +2.5v zero.

Ahh yes ... in all the excitement I had forgotten that key point ... and also that they need to read both ways to cover the backfeeding. Doh!

If I remember, Klaus used an HSTS-400 amp sensor, and I used an HSTS-100 amp sensor which worked very well for me here.

Are they HSTS21-400? I found a datasheet for them ... and they look the same profile as my existing one, so should be easy enough to swap over.

For measuring surges and peak inrush, a digital CRO triggered on the rising edge is what I use.
Some of the better quality multimeters can sometimes have a peak hold function too.

Yeah I can use mine but I get a bit lazy dragging it around to the inverter and then 5 mins later needing it on the bench, though I do have a peak hold function on the clamp meter. I'm pretty sure that eventually I won't look at any of that stuff ... but while I'm learning how everything works some actual figures and levels have been very useful.

 

[Warpspeed]

Are they HSTS21-400? I found a datasheet for them ... and they look the same profile as my existing one, so should be easy enough to swap over.

You are quite right, just went out to the shed and had another look.
The one I have here is an HSTS21-100. That is plenty for me, my full load current is only 50 amps.
There are a whole family of these including 400 amp and 500 amp.
https://html.alldatasheet.com/html-pdf/1154780/YHDC/HSTS21-400A/115/1/HSTS21-400A.html

 

[MattiFin]

Yes, the switching bridges draw current in half sinusoids from the bank of big electrolytics.
Current from the battery is pretty much smooth dc at low inverter loads, with ac ripple increasing in amplitude with higher inverter loading.

The ac output voltage from the inverter is pretty much tied to the instantaneous dc bus voltage on the electrolytics and voltage regulation of the transformers.
A sudden transient surge on the ac side may well be catered for by stored energy in the electrolytics, in which case very little actual voltage correction by the feed forward system may be required.

The electrolytics slow down voltage changes across the dc bus, and give the feed forward additional time to correct.
Its common on commercial PWM inverters to have minimal sized electrolytics for cost reasons.

One of the main features of a home brew Warpverter is the massive (no cost spared!) bank of low ESR electrolytics.
That goes a very long way towards achieving very good step load change performance, both increasing and decreasing.
An ultracapacitor bank would be the ultimate if costly solution.

Isn't the DC bus voltage already one part of feedforward? DC bus current is just additional feedforward to compensate for I*R losses in transformers/wiring?

As far as I can see both options have their positives and negatives:
DC bus load current compensation: slower to react, compensates transformer/wiring losses only partially in transients (can't compensate voltage loss on transformer resistance if it doesn't show in DC bus current due to huge DC bus capacitance!)
AC load current or DC current between bus and H-bridge compensation: faster information about load current but more noisy, possibly need to be filtered heavily so that ends up as slow as the DC bus measurement.

 

[Warpspeed]

Isn't the DC bus voltage already one part of feedforward? DC bus current is just additional feedforward to compensate for I*R losses in transformers/wiring?

As far as I can see both options have their positives and negatives:
DC bus load current compensation: slower to react, compensates transformer/wiring losses only partially in transients (can't compensate voltage loss on transformer resistance if it doesn't show in DC bus current due to huge DC bus capacitance!)
AC load current or DC current between bus and H-bridge compensation: faster information about load current but more noisy, possibly need to be filtered heavily so that ends up as slow as the DC bus measurement.

Yup, you have this pretty well all figured out.
Feedforward attempts to compensate for conduction losses and the resulting voltage droop within the inverter and inverter wiring.
Its not perfect, but it has much better overall dynamics than voltage feedback for sudden large step load changes.

faster information about load current but more noisy, possibly need to be filtered heavily so that ends up as slow as the DC bus measurement.

Its certainly faster, and might be potentially more prone to noise.
The noise problem can be largely overcome by using a dual slope integrating analog to digital converter that is synchronized to the inverter operating frequency. The integrating function largely removes impulse and high frequency noise.
Having a large low ESR capacitor bank also goes a long way to improving dynamic response to sudden load changes from the large stored energy, and that too helps remove noise from the incoming dc.

The final result of using feed forward, trades off some static voltage regulation accuracy for speed of response to sudden load changes.
The grid voltage goes up and down by a few volts during day and night, and everything still works just fine.

What is far more objectionable is light flicker when something like a large induction motor starts up (refrigerator).
Voltage feedback slowly ramps and corrects, but it still a comparatively slow process, which can become unstable if it over corrects.
Feedforward is faster, and unconditionally stable, but output voltage consistency is less than perfect.

Ac voltage that slowly wanders up and down by a couple of volts, is never a real problem.
Sudden voltage surges and dips that cause light flicker is far more objectionable.

The wandering voltage issue could be fixed by using very slow feedback in addition to the much faster feed forward.
Its just that the added complexity would be difficult to justify.

 

[MattiFin]

Yup, you have this pretty well all figured out.
Feedforward attempts to compensate for conduction losses and the resulting voltage droop within the inverter and inverter wiring.
Its not perfect, but it has much better overall dynamics than voltage feedback for sudden large step load changes.

--

The noise problem can be largely overcome by using a dual slope integrating analog to digital converter that is synchronized to the inverter operating frequency. The integrating function largely removes impulse and high frequency noise

You could at least in theory also do the feedforward to reference voltage from the AC output current. Analog domain only, no need to get the AD-converters in the mess.
But I also envision it as more unstable or prone to oscillations, especially if you over-compensate with the feedforward.

 

[Warpspeed]

--

You could at least in theory also do the feedforward to reference voltage from the AC output current. Analog domain only, no need to get the AD-converters in the mess.
But I also envision it as more unstable or prone to oscillations, especially if you over-compensate with the feedforward.

If you use the output for control, then it becomes by definition feedback !
Getting a fast precise measurement of ac is always tricky, because it never stays still, always swinging up and down, and as you say, a PID control feedback loop can very easily over correct and become unstable.

If you overcompense with feedforward, all that happens is that voltage droop turns into a voltage rise with increasing load.
The inverter then has a negative output impedance. I wondered if that could ever become unstable, but it turns out that it does not.

The inverter load will always be (mostly) resistive, and that swamps out any slight negative output impedance of the inverter.
If you go absolutely crazy with feedforward, I am sure you could provoke oscillation, but with a practical inverter I have never seen any tendency towards instability with the current correction tweaked right up.

Its all been a quite fascinating thing to play around with.

 

[MattiFin]

If you use the output for control, then it becomes by definition feedback !

I'm not sure about terminology and this goes to maybe irrelevant nit-picking for this thread.

But I would still call it feedforward as output current is not really the end goal (output voltage control is the "end goal")
It is still open-loop "best guess" of the needed compensation unlike feedback.
Also it doesn't have any error amplifier-part that I'd imagine as a part of feedback control.

(You could have also feedback from output current if you had sinusoidal reference current that you could feed to the error amplifier but then your inverter would have current output)

 

[Warpspeed]

The whole concept of feedforward is that all the inputs drive some algorithm or lookup table to produce as you say a "best guess" at providing the desired result at the output.
The advantage is to provide a theoretically instantaneous correction of multiple rapidly changing input conditions, sacrificing some accuracy of output for a huge improvement in overall reaction time.
That was my purpose, faster response to large step changes in inverter load.
Its also much easier to detect and measure sudden changes in dc rather than ac.

Measuring the water depth of a still lake, is a lot simpler than measuring the precise average water depth with a strong ocean swell running.