SMPS inverter voltage feedback methods

[Joerg]

Joerg wrote:
Jamie Morken wrote:
Joerg wrote:
Jamie Morken wrote:
Joerg wrote:
Jamie Morken wrote:
Joerg wrote:
MooseFET wrote:
MooseFET wrote:
[....]
http://en.wikipedia.org/wiki/Peak_detector
I think maybe this is what was the idea from the start but
I am
a bit thick headed
Also consider:
+12V
!
!VCC\ LM339 ---/\/\---+12V
IN ----!+ \ !
! >---+----+
--!- / ! !
! !VSS/ ! ===
! ! ! !
! -12V ! GND
------------
For the accuracy of this transformer chopped voltage
measurement,
what are some techniques to improve accuracy, or does it
require
per unit calibration because of the variances in components?

The leakage inductance of the transformer is your enemy. So
long as
it is small and the transformer is well manufactured, the
errors in
the transformer will repeat nearly exactly.


Two ways to deal with this:

a. Switch to a known voltage once in a while. Such as a
reference on the isolated side. For 12 bit accuracy that won't
exactly be a cheap one or has to be calibrated.

b. Use an FM scheme to transfer the information. That
eliminates the influence of the leakage inductance.

What would an FM method look like for this application?


Two methods come to mind:

One would be a simple voltage to frequency conversion.
Unfortunately the old standby LM331 only goes to 100kHz and you
want more BW so you'd have to roll your own. It also doesn't
come in SMT which can be a drag. Now this gets transferred
across a toroid and on the other side you translate the
frequency back into a voltage which can often be done using the
same chip. Or a micro just counts the pulses against its timer
which is pretty easy. For some reason I always run out of timers
:-(

Method two would be a PLL where the voltage you want to log
modulates the loop. On the other side you place a detector.
Remember AFC? That used a detector scheme that measures the DC a
station created at the end of the IF chain. The DC was an
indicator how far your radio was out of tune and its oscillator
was regulated to a point where it's right on. So you need the
first part of the AGC, sans loop.

Thanks, I have an FPGA on the secondary side so counting pulses
would be
a perfect application for that, the tricky part is getting the
primary
side voltage to frequency, it might require its own clean power
supply
for that section I think, like if using the PLL to generate the
frequency.


Well, yeah, but that's just a matter of a TLV431 or TL431. Under
10c a pop.

For using a PLL, would the 74HCT4046 work for this 200kHz bandwidth or
is there another PLL that would be better to use? I've never used
a PLL
directly so am not sure the best way to implement the voltage to
frequency generation for feeding the transformer the FM signal.


Yes, the 4046 would be able to do that. However, in order to obtain
a reasonable conversion factor precision and linearity you may have
to servo it. That gets old pretty quickly. IMHO a pulse modulation
scheme with DC restore on the system side is more economical.


Which form of pulse modulation would be good to implement to convert a
120VAC signal for this and then send it over the isolation transformer
or perhaps opto?

http://en.wikipedia.org/wiki/Pulse_modulation

That link shows 5 different pulse modulation methods, PAM, PWM, PCM,
PPM, PDM

This one is easiest:
http://en.wikipedia.org/wiki/Pulse_amplitude_modulation

Back to this one then I think?

http://rocketresearch.nekrom.com/new/transformer chopper voltage measurement/circuit2.jpg

I usually do it with FET switches or multiplexers. Clock enters xfmr
over to isolated side, ping-pong the mux between signal-to-be-measured
and ISO-GND. Back across another xfmr, then DC-restore on the system
side followed by sample&hold or track&hold. The clock can also be used
to derive a low DC voltage on the iso side to run some circuits there.
But mostly I need to run some big stuff there in which case there'd be a
third (larger) toroid for the isolated supply. If the BW can be small
this big xfmr can again be used to supply a clock.