Some input/help on this circuit

[jgauthier]

Design goals:
I have a 9VAC (1A) power supply. I want to power some microprocessor boards with this, as well as monitor it's signal. (the whole scope is larger, but this portion is isolated to the power supply and interactions with it)

My plan is to rectify the 9VAC into DC (then use DC/DC converters for 5V/3V items). I also want to sample the AC signal with one of the MCPUs, so I need to get it down to ADC levels (0-3.3V)

I put together this simulation on LTSpice, and while it "works" there could definitely be some improvements. (I've also attached it)


For the rectifier I am planning on using this. This simulation has a single 4700uF capacitor in it. At a full load current, that will cause a 1.7V ripple. The DC/DC converters I am planning on using are these. They shouldn't have a problem with the ripple. I'll use one for 5V and one for 3.3V. However, it's unlikely I will be at full load current. 500-600mA is more realistic, so that would be less ripple.

With regards to the circuit: I use a voltage divider on the 9V RMS signal to get it to 3.3v pk-pk. I understand there are multiple ways to shift this signal up. I decided to go with a summing amplifier. An issue I ran into was the impedances caused by both the 3.3V->1.65V divider and the 9VAC ->1.65V divider. So, I buffered both of them. Finally, using the same op amp, I sum them together.

There may be a way to use a single rail, but I just used two rectifiers to get +/-12V to use as the voltages for the op amps. From my reading using a single rail would require doing some other level shifting, and referencing a VCC and a VCC/2 to get it done. I felt like if I already have 12V, then why not use it.

This circuit "works" with some caveats.
The 3.V output works with some "startup" issues:


The +/-12V signals work, but I would have expected to see some ripple. Perhaps the sim is limiting that somehow:


The -12V signal takes 10ms to get going, which is.. strange to me.

Lastly, and this is a concern to me, I have the "grounds" tied together.
The connector coming in is a barrel jack, so I call them "tip" and "ring" on the schematic.
In this sim, "ring" is connected to "ground". I cannot get the simulation to work otherwise.

Definitely welcome suggestions on that.

Thanks for looking!

 

[Harald Kapp]

Assuming the signals 9VAC and ring are the same (from the same AC source), then this circuit won't work. Here's why:

When 9VAC is positive, then ring is negative and there is a direct path between the two via the connection marked red in the image. The same goes for the inverse signals via D6 and D3.
This will work with separate AC sources only. f you have only a singel 9 V AC source, the single supply approach may be simpler. Just create a "virtual ground" and reference all signals to this. Use a capacitor plus resistive divider to couple the AC into your circuit for monitoring. You may need to use rail-to-rail opamps to be able to measure the full range of input voltages

This technique creates a dual supply from a single AC source but requires bigger filter capacitors since it uses half-wave rectifying.

Another option uses additional capacitors to create a dual supply from a single AC source. This method is somewhat limited in output current by the necessary size/capacity of the isolating capacitors.

Or use a transformer with 2 secondary windings or a single secondary winding with center tap.

 

[Alec_t]

Provided your 9V AC source does not have one terminal grounded, this arrangement would give you positive and negative supplies :-

 

[Harald Kapp]

@Alec_t : This is like a virtual ground but without a low impedance path for current. I'm sure you'll experience issues when the load is not absolutely symmetrical. For example when the load is connected from V+ to GND ...

 

[jgauthier]

@Harald Kapp Thank you for the your detailed response. I learned about the "dual supply" trick after posting. But that page, and the virtual ground page were both excellent. I am going to explore both options.

 

[Alec_t]

I'm sure you'll experience issues when the load is not absolutely symmetrical. For example when the load is connected from V+ to GND .

Agreed, but should suffice for low-current use such as an op-amp supply.

 

[crutschow]

Agreed, but should suffice for low-current use such as an op-amp supply.

No.
Any output load on the op amp will unbalance the current and cause a problem, as it has no path to the AC supply.

On way get a plus and minus supply is to ground one terminal of the AC supply, and use two diodes connected in inverse parallel at the output to get two half-wave rectified outputs of opposite polarity.

 

[Harald Kapp]

Plus no DCpath at all. Try connecting an LED + resistor from "+" to the virtual ground ...

 

[Alec_t]

Ok. I give in .