SMPS inverter voltage feedback methods

[Jamie Morken]

Hi,

For this circuit:

http://rocketresearch.nekrom.com/ne...inverter/240VAC 60Hz split phase inverter.jpg

For measuring the "120VAC-x" signals at 200kHz bandwidth what would a
good way to do this?

So far for possible methods I have, voltage divider to opamp/ADC,
optically isolated error amplifier, ie. FOD2741BSDV
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=FOD2741BSDVCT-ND

isolation barrier opamp (transformer based or capacitive):
ie. AD215BY (transformer chopper based)
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=AD215BY-ND

ie. ISO122 (capacitive coupled)
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=ISO122JP-ND

I think making a circuit like the AD215 might be the way to go, or maybe
just a simple resistor divider could work too.

cheers,
Jamie

 

[D from BC]

Hi,

For this circuit:

http://rocketresearch.nekrom.com/ne...inverter/240VAC 60Hz split phase inverter.jpg

For measuring the "120VAC-x" signals at 200kHz bandwidth what would a
good way to do this?

So far for possible methods I have, voltage divider to opamp/ADC,
optically isolated error amplifier, ie. FOD2741BSDV
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=FOD2741BSDVCT-ND

isolation barrier opamp (transformer based or capacitive):
ie. AD215BY (transformer chopper based)
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=AD215BY-ND

ie. ISO122 (capacitive coupled)
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=ISO122JP-ND

I think making a circuit like the AD215 might be the way to go, or maybe
just a simple resistor divider could work too.

cheers,
Jamie

How about this parts intensive idea?
Do it FM radio style

Vin>VCO(1Ghz)>xformer>FM demod>Vout isolated

RF Power Detector
http://www.linear.com/pc/downloadDocument.do?navId=H0,C1,C1011,C1074,P2219,D2368

xformer
small ferrite xformer

VCO
Crystek Corporation
(Yikes...around $25.00 for ~1Ghz)

Dunno if this has been done before.. Just made it up..

Propagation time ?
delta V to delta V response time ?
EMI ?
Phase response?
Nonlinearity ?


D from BC
British Columbia
Canada.

 

[MooseFET]

Hi,

For this circuit:

http://rocketresearch.nekrom.com/new/split phase 240VAC inverte...

For measuring the "120VAC-x" signals at 200kHz bandwidth what would a
good way to do this? [....]
just a simple resistor divider could work too.

This would be a good way to go since the output is not isolated from
what I can see.

Also:

If you make a bandpass filter that takes the voltage directly from the
switches, you can improve the loop responce. It gives you something
very like the "D" part for your PID controller.

 

[MooseFET]

How about this parts intensive idea?
Do it FM radio style

Vin>VCO(1Ghz)>xformer>FM demod>Vout isolated

If you go down to much lower frequencies, the idea works quite well.
At 1MHz or so, the VCO doesn't take very many parts.

http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1010,C1096,P2186
The suggested circuit plus one resistor makes not too bad of a VCO

The receiver side can be a PLL that uses a tri-state phase detector
built into your PCLD or FPGA or whatever and a second VCO like the
first.

 

[D from BC]

If you go down to much lower frequencies, the idea works quite well.
At 1MHz or so, the VCO doesn't take very many parts.

http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1010,C1096,P2186
The suggested circuit plus one resistor makes not too bad of a VCO

The receiver side can be a PLL that uses a tri-state phase detector
built into your PCLD or FPGA or whatever and a second VCO like the
first.

Cool...


D from BC
British Columbia
Canada.

 

[Paul Hovnanian P.E.]

How about this parts intensive idea?
Do it FM radio style

Nah. Solenoid on the sensed AC acting against a spring. Throw in a
dashpot to damp it if necessary. Connect to a suitable potentiometer via
a bellcrank type linkage.

 

[Joerg]

Hi,

For this circuit:

http://rocketresearch.nekrom.com/ne...inverter/240VAC 60Hz split phase inverter.jpg


For measuring the "120VAC-x" signals at 200kHz bandwidth what would a
good way to do this?

So far for possible methods I have, voltage divider to opamp/ADC,
optically isolated error amplifier, ie. FOD2741BSDV
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=FOD2741BSDVCT-ND


isolation barrier opamp (transformer based or capacitive):
ie. AD215BY (transformer chopper based)
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=AD215BY-ND

$$$ ... Ouch!

ie. ISO122 (capacitive coupled)
http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail?name=ISO122JP-ND


I think making a circuit like the AD215 might be the way to go, or maybe
just a simple resistor divider could work too.

To monitor the waveform? Why not just sample it? Send sampling pulse
through toroid xfmr, send some AC there if you can't feed the HV side
circuitry locally (toroid xfmr or something), sample, then send sample
across another toroid xfmr back to LV side. I use little #43 ferrites
for those jobs, sometimes #77. Mostly #43 because there always seems to
come a time when someone asks for even more bandwidth later.

Could also use LAN xfmrs. They are often rated 1500VAC but mostly not
agency approved for this kind of job. If they don't feature a split
bobbin don't use them.

 

[MooseFET]

Nah. Solenoid on the sensed AC acting against a spring. Throw in a
dashpot to damp it if necessary. Connect to a suitable potentiometer via
a bellcrank type linkage.

How about an electric heater boiling water to push a piston.

 

[Joerg]

How about an electric heater boiling water to push a piston.

Nah. A pump, waters a plant, plant grows, weight increases, click.

 

[Ben Jackson]

Nah. Solenoid on the sensed AC acting against a spring. Throw in a
dashpot to damp it if necessary. Connect to a suitable potentiometer via
a bellcrank type linkage.

You're making this way too difficult. You can do it with all off-the-
shelf parts. Just hardwire a fluke meter to the AC side, then use a
webcam on the sensing side connected to a laptop running computer vision
software.

Better make it a backlit meter. Oh --and a monkey to change the batteries
from time to time.

 

[Jamie Morken]

If you go down to much lower frequencies, the idea works quite well.
At 1MHz or so, the VCO doesn't take very many parts.

http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1010,C1096,P2186
The suggested circuit plus one resistor makes not too bad of a VCO

The receiver side can be a PLL that uses a tri-state phase detector
built into your PCLD or FPGA or whatever and a second VCO like the
first.

This is the way I'd like to do the voltage feedback, with a VCO feeding
a transformer, and then the PLL circuit, but I think that part is too
slow to react, the maximum VCO modulation bandwidth of the LTC6900 is
25kHz, I'm running the PWM at 200kHz.

cheers,
Jamie

 

[Jamie Morken]

$$$ ... Ouch!




To monitor the waveform? Why not just sample it? Send sampling pulse
through toroid xfmr, send some AC there if you can't feed the HV side
circuitry locally (toroid xfmr or something), sample, then send sample
across another toroid xfmr back to LV side. I use little #43 ferrites
for those jobs, sometimes #77. Mostly #43 because there always seems to
come a time when someone asks for even more bandwidth later.

Is this the same as what Moosefet is talking about with the VCO on the
transformer primary? Can the sampling be tied into the PWM signals sent
to the SMPS fets? So that the modulation and demodulation is done by
these signals, not sure if this would work with variable dutycycle PWM?

cheers,
Jamie

 

[MooseFET]

You're making this way too difficult. You can do it with all off-the-
shelf parts. Just hardwire a fluke meter to the AC side, then use a
webcam on the sensing side connected to a laptop running computer vision
software.

Better make it a backlit meter. Oh --and a monkey to change the batteries
from time to time.

The Fluke 45 plugs in the wall. Its display it lit so you don't need
the monjey.

 

[MooseFET]

This is the way I'd like to do the voltage feedback, with a VCO feeding
a transformer, and then the PLL circuit, but I think that part is too
slow to react, the maximum VCO modulation bandwidth of the LTC6900 is
25kHz, I'm running the PWM at 200kHz.

The servo feedback needs a bandwith much higher than 60Hz but not
higher than the 200KHz. For the AC feedback, you can just use a
transformer drivien with the output signal. The DC feedback is just
to ensure that you don't have a DC component in the output.

 

[Jamie Morken]

[...]
How about this parts intensive idea?
Do it FM radio style
Vin>VCO(1Ghz)>xformer>FM demod>Vout isolated
If you go down to much lower frequencies, the idea works quite well.
At 1MHz or so, the VCO doesn't take very many parts.
http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1010,C1096,P2186
The suggested circuit plus one resistor makes not too bad of a VCO
The receiver side can be a PLL that uses a tri-state phase detector
built into your PCLD or FPGA or whatever and a second VCO like the
first.

This is the way I'd like to do the voltage feedback, with a VCO feeding
a transformer, and then the PLL circuit, but I think that part is too
slow to react, the maximum VCO modulation bandwidth of the LTC6900 is
25kHz, I'm running the PWM at 200kHz.

The servo feedback needs a bandwith much higher than 60Hz but not
higher than the 200KHz. For the AC feedback, you can just use a
transformer drivien with the output signal. The DC feedback is just
to ensure that you don't have a DC component in the output.

The PWM dutycycle needs to be set each cycle, using the voltage
and current feedback, I'm using small inductors so it can change quite
rapidly. Would you be able to help me design this modulator/demodulator
ready to feed into the 200kHz ADC if I put it in the public domain and
pay you or at least send you some chocolates?

I already am going to use a bipolar ADC floating with the 120VAC and SPI
optocoupled, but its good to have a backup, as I'm not sure if I can get
12bits resolution from that ADC.

cheers,
Jamie

 

[Jamie]

You're making this way too difficult. You can do it with all off-the-
shelf parts. Just hardwire a fluke meter to the AC side, then use a
webcam on the sensing side connected to a laptop running computer vision
software.

Better make it a backlit meter. Oh --and a monkey to change the batteries
from time to time.

ha, what about Calibration certification when changing the batteries?
Who's going to do that ?

 

[MooseFET]

Joerg wrote: [....]

To monitor the waveform? Why not just sample it? Send sampling pulse
through toroid xfmr, send some AC there if you can't feed the HV side
circuitry locally (toroid xfmr or something), sample, then send sample
across another toroid xfmr back to LV side. I use little #43 ferrites
for those jobs, sometimes #77. Mostly #43 because there always seems to
come a time when someone asks for even more bandwidth later.

Is this the same as what Moosefet is talking about with the VCO on the
transformer primary? Can the sampling be tied into the PWM signals sent
to the SMPS fets? So that the modulation and demodulation is done by
these signals, not sure if this would work with variable dutycycle PWM?

From my reading of it, it isn't but it is worthy of consideration. A
balanced modulator and balanced demodulator will bring DC across with
no trouble. If you pick the right sort of running frequency, it will
be quite insensitive to all the switching noise.

With a bit of trichery, you could get any imbalance between the two
sides to come across on the same transformer. The second channel
could be phase or an even harmonic.

 

[Terry Given]

MooseFET wrote:

[...]

How about this parts intensive idea?
Do it FM radio style
Vin>VCO(1Ghz)>xformer>FM demod>Vout isolated

If you go down to much lower frequencies, the idea works quite well.
At 1MHz or so, the VCO doesn't take very many parts.
http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1010,C1096,P2186

The suggested circuit plus one resistor makes not too bad of a VCO
The receiver side can be a PLL that uses a tri-state phase detector
built into your PCLD or FPGA or whatever and a second VCO like the
first.

This is the way I'd like to do the voltage feedback, with a VCO feeding
a transformer, and then the PLL circuit, but I think that part is too
slow to react, the maximum VCO modulation bandwidth of the LTC6900 is
25kHz, I'm running the PWM at 200kHz.

The servo feedback needs a bandwith much higher than 60Hz but not
higher than the 200KHz. For the AC feedback, you can just use a
transformer drivien with the output signal. The DC feedback is just
to ensure that you don't have a DC component in the output.

The PWM dutycycle needs to be set each cycle, using the voltage
and current feedback, I'm using small inductors so it can change quite
rapidly. Would you be able to help me design this modulator/demodulator
ready to feed into the 200kHz ADC if I put it in the public domain and
pay you or at least send you some chocolates?

I already am going to use a bipolar ADC floating with the 120VAC and SPI
optocoupled, but its good to have a backup, as I'm not sure if I can get
12bits resolution from that ADC.

cheers,
Jamie

think Nyquist. no matter how hard you try, you cant organise closed-loop
voltage or current control beyond 100kHz. Fsmps/10 is about as fast as
you can go.

If its a current-mode controller then you get cycle-by-cycle current
limiting, and the voltage EA generates a (slowly varying) current setpoint.


Cheers
Terry

 

[MooseFET]

MooseFET wrote:

MooseFET wrote:
[...]
How about this parts intensive idea?
Do it FM radio style
Vin>VCO(1Ghz)>xformer>FM demod>Vout isolated
If you go down to much lower frequencies, the idea works quite well.
At 1MHz or so, the VCO doesn't take very many parts.
http://www.linear.com/pc/productDetail.jsp?navId=H0,C1,C1010,C1096,P2186
The suggested circuit plus one resistor makes not too bad of a VCO
The receiver side can be a PLL that uses a tri-state phase detector
built into your PCLD or FPGA or whatever and a second VCO like the
first.
This is the way I'd like to do the voltage feedback, with a VCO feeding
a transformer, and then the PLL circuit, but I think that part is too
slow to react, the maximum VCO modulation bandwidth of the LTC6900 is
25kHz, I'm running the PWM at 200kHz.
The servo feedback needs a bandwith much higher than 60Hz but not
higher than the 200KHz. For the AC feedback, you can just use a
transformer drivien with the output signal. The DC feedback is just
to ensure that you don't have a DC component in the output.

The PWM dutycycle needs to be set each cycle, using the voltage
and current feedback, I'm using small inductors so it can change quite
rapidly. Would you be able to help me design this modulator/demodulator
ready to feed into the 200kHz ADC if I put it in the public domain and
pay you or at least send you some chocolates?

I already am going to use a bipolar ADC floating with the 120VAC and SPI
optocoupled, but its good to have a backup, as I'm not sure if I can get
12bits resolution from that ADC.

cheers,
Jamie

think Nyquist. no matter how hard you try, you cant organise closed-loop
voltage or current control beyond 100kHz.

That isn't quite true. Depending on the sort of PWM modulator you
use, you get two samples per cycle because both edges of the on time
can be moved by the input.

Fsmps/10 is about as fast as
you can go.

This still ends up about the right limit. When you move both edges in
a PWM, you usually end up with a giant noise spike right at Nyquist.

If its a current-mode controller then you get cycle-by-cycle current
limiting, and the voltage EA generates a (slowly varying) current setpoint.

The is true for moderately large values of "slowly". Since the OP is
trying to construct a 60Hz sine wave, he has a lower limit on the
voltage feedback frequency responce of perhaps 1KHz or so.

If he has a microcontroller in the system, he can also use the
information from this point in the previous cycle to improve his
waveform for a constant load.

 

[Joerg]

Joerg wrote: [....]

To monitor the waveform? Why not just sample it? Send sampling pulse
through toroid xfmr, send some AC there if you can't feed the HV side
circuitry locally (toroid xfmr or something), sample, then send sample
across another toroid xfmr back to LV side. I use little #43 ferrites
for those jobs, sometimes #77. Mostly #43 because there always seems to
come a time when someone asks for even more bandwidth later.

Is this the same as what Moosefet is talking about with the VCO on the
transformer primary? Can the sampling be tied into the PWM signals sent
to the SMPS fets? So that the modulation and demodulation is done by
these signals, not sure if this would work with variable dutycycle PWM?

From my reading of it, it isn't but it is worthy of consideration. A
balanced modulator and balanced demodulator will bring DC across with
no trouble. If you pick the right sort of running frequency, it will
be quite insensitive to all the switching noise.

With a bit of trichery, you could get any imbalance between the two
sides to come across on the same transformer. The second channel
could be phase or an even harmonic.

Yep, works like a synchronous detector. Most ECG units operate that way
since there usually cannot be any conductive path to the patient.

Which frequency it is tied to doesn't matter. But if the SMPS FETs don't
consistently switch in excess of 500kHz Jamie isn't going to yield his
desired 200kHz signal bandwidth. So it might be better to send sample
pulses across. It is also quite common to use this same sample pulse
transmission to get some energy across to feed circuitry on the isolated
side. Saves a toroid.

Pick your frequency carefully and don't get into trouble with the FCC
because something was too leaky.