semi-isolated feedback without optocouplers?

[zilog]

My scenario is a battery powered push-pull smps with dual complimentary outputs of +-45V, where I want the secondary GND to be isolated from the primary GND to avoid ground loops in the amplifiers I will feed. It is however not important to keep pri/sec completely isolated, a few hundred ohms is enough, and the pri/sec grounds could be limited to float within 2-3V from each other without causing any problems.

Is there any feedback topology that accomplishes what I want without involving optocouplers to sense the secondary voltage? Optocouplers suffer from both bad linearity, bandwidth and temperature sensitivity, so I would really like to keep them out of my loop to speed up frequency response. I will need to sense both rails (due to reasong not only involving voltage regulation), so temperature variations will affect symmetry aswell.

I have thought about sensing using OP-amplifiers on the secondary side, and somehow transferring a balanced (one posititve and one negative copy) current to the primary side to avoid any net current acusing the pri/sec sides to drift in voltage. Is this an idea that could be further developed on?

[janneman]

If you use a differential-in, differential out opamp you can sense the difference on the sec side between output and sec ground, and transfer it as a siganl referred to prim ground, which I guess is what you want?
The opamp will have a common mode range of maybe up to 7 or 8 V if fed from +/-15V.
Look at the OPA1632 or THS4030 IIRC.

Jan Didden

[Fob]

I hope this help. The signal is scaled to 2.5 Vref on pwm error amplifier for+/-45V.

[Claude Abraham]

I don't understand why you dislike using an opto. Bandwidths of 1 to 2 MHz are quite available with today's opto parts. The control loop BW is generally less than 50 kHz, so the opto speed will not be the bottle neck. As far as linearity and temperature sensitivity goes, if the opto is embedded inside the servo loop, those are not issues.

I just completed a flyback topology using the 6N1136 opto coupler with great results. The output is fully isolated, and the line and load regulations are 0.1 to 0.2%. The converter works very well over the input range of 9 to 32 volts and the extended industrial temp range of -40 to +105 deg C.

Don't rule out an opto. The classic TL431 servo loop control scheme in use since the '80's is well proven and high performance. I hope I've helped. BR.

Claude

[lineup]

Ask Steve Eddy
he might know the signal transformer to carry feedback

[Eva]

You are wrong about optocouplers. You can approach 10Khz bandwidth even with the most common parts like 4N25 or PC817 (depending on bias current and load resistance). You can even simulate a short circuit load with an op-amp to avoid miller effect and further extend bandwidth.

Differential amplifiers in a noisy environment are likely to bring trouble, while the HF roll-off from optocouplers has the opposite effect.

[lineup]

There are special analog opto couplers
They are not too bad, when comes to linearity & bandwidth
Eell worth a try, for 'normal bw' power amplifiers.

Of course, speed is nothing like a normal feedback methods ...
and precision can not be like best, either. But still not bad!

==============================================
TIL300 .... >200 kHz
Precision Linear Optocoupler (Rev. A) (til300.pdf, 86 KB)
http://focus.ti.com/lit/ds/symlink/til300a.pdf
search also for
IL300 = Siemens
http://saturn.uni-mb.si/~bojan/pdf/il300.pdf


and here is my 5 year old topic about it:

Audio Signal Isolation Techniques - Analogue Optocouplers
- halojoy January 2003
Audio Signal Isolation Techniques - Analogue Optocouplers



Lineup = Halojoy

[janneman]

Originally posted by Fob
I hope this help. The signal is scaled to 2.5 Vref on pwm error amplifier for+/-45V.

This is a very noisy, sub-optimal solution. 26dB opamp gain and S/N ratio is thrown away.

Jan Didden

[zilog]

Originally posted by lineup
TIL300 .... >200 kHz
Precision Linear Optocoupler (Rev. A) (til300.pdf, 86 KB)
http://focus.ti.com/lit/ds/symlink/til300a.pdf
search also for
IL300 = Siemens
http://saturn.uni-mb.si/~bojan/pdf/il300.pdf
[/B]

Thanks, hadn't tought about using dual couplers for linearization.

I will be measuring a dual rail system, with one IL300/TL272 for the V+ <-> GND portion, and one IL300/TL272 for the GND <-> V- portion. Am I assuming correctly that linearity is good enough to get readings from both these IL300:s to be within one diode drop from eachother given identical input? I know this is overkill for a car-audio smps, but the circuit also serves as the test bench for some future projects that are much more stringent.

Is there any cheaper solution to accurately measuring both rails than using 2xIL300? Using my normal supply channels, this will cost ~25 Euro just in components..

[Eva]

You don't want wide feedback bandwidth. If you go that way, you will find all the undesirable output ripple above twice the switching frequency disturbing the modulator.

Operational amplifiers will also have trouble handling output ripple if low-pass filtering is not employed. In the end you will find yourself limiting the bandwidth of those fast optocouplers to well below 10Khz in order for the circuit to work properly.

What is the resonant frequency of your output filter? At what control-loop crossover frequency are you aiming? Gain up to frequencies too close to the switching frequency always causes trouble in push-pull derived converters.