DC 12V DC to 9V AC 12W

[Buriedcode]

I've searched the forum and found a number of posts regarding low voltage, low power inverters for 50/60Hz AC.  Almost all resolved with either modify the original equipment (that probably just reictified it for multiple DC supplies) or use a small 12V to mains inverter, and the original supply.

I have a piece of audio equipment that accepts 9V AC, it uses a couple of voltage doublers and linear regs to make +/-12V supplies, along with a 5V and a charge pump for ~100V.  Whilst I could modify it by replacing the -12V and +100V supplies with their own separate supplies from 12V, I don't really want to modify it at all. 

The original power supply is just a 220V to 9V transformer, its quite heavy and bulky, but is obviously the simplest, cheapest, and most reliable solution.  It still works, so this isn't about looking for something that "just works", but I've repaired its power cord several times.

As another "corona virus project" (= time sink) I thought I'd investigate just how hard it would be to create a low power, 15W converter that outputs reasonably clean 50Hz AC.  Preferably powered by 10-19V DC so I can use a generic power brick.   Boosting 12V to something like 15-16V for headroom, then a H-bridge with a hefty filter on the output was my first thought, but one side of the power input to this device is connected to the chassis, and the audio in/out grounds.   This means when hooked up to other equipment could be earthed, and will be part of the signal chain.

Am I correct in thinking that as long as either the converter itself is isolated/floating, or the power input to the converter is isolated this wont matter at all?  Could I just ground one side of the H bridge output? 

I considered a half bridge, as one side could then be "actual" ground, but that creates problems of having to make a bipolar supply, and at 50Hz it'll need some hefty caps on the rails to ease the sag during the on cycle, or use a much bigger converter for 30W on each rail.  I even thought perhaps having separate converters, charging the one rails cap whilst the other one is PWMd to the output.  It becomes a bit of a mess when a H-bridge is by far the simplest way, and has plenty of off-the-shelf drivers.

Any better ideas?

[TimNJ]

How about a Class-D amplifier? You can drive it with a 50Hz sine wave from either DDS or analog source. Feedback should compensate for supply voltage fluctuation, within reason. At the end of the day though, I'm not sure how much smaller this would be than a 15VA 50Hz transformer.

[NiHaoMike]

Replace the diodes with fast recovery or Schottky diodes, then make a high frequency inverter operating at ultrasonic frequencies. It makes no sense to do power conversion within the audio band to power a sensitive audio circuit.

[Buriedcode]

Replace the diodes with fast recovery or Schottky diodes, then make a high frequency inverter operating at ultrasonic frequencies. It makes no sense to do power conversion within the audio band to power a sensitive audio circuit.

As I said I'd rather not modify the original unit, its semi-vintage and I've just spent a while refurbishing it.  If they are somehow destroyed in the future, I will consider this.

How about a Class-D amplifier? You can drive it with a 50Hz sine wave from either DDS or analog source. Feedback should compensate for supply voltage fluctuation, within reason. At the end of the day though, I'm not sure how much smaller this would be than a 15VA 50Hz transformer.

That's exactly what I was thinking, essentially a half bridge converter, but the feedback of a class D should keep the output relatively clean.  Given its for a power application rather than audio, some THD is acceptable, so I can increase the dead band to something sensible.  This ultimately leaves the "difficult" part to the bipolar power supply.  I think I'll play with LTspice for a bit..

[Buriedcode]

Right, so after re-acquainting myself with loop filters, poles and zeros, I have a reasonable LTspice simulation of a half bridge class D with feedback.  I also did a full bridge too, but given a 20ms simulation produced a pretty big raw file I didn't fully simulate it driving different loads.

If I go the half bridge route I can keep the ground, but require a split rail +/-15V supply, most likely at 1A on each rail.  If I go full bridge, I'll have to ensure its isolated.  As expected this just opens up a can of worms in deciding the best topology, but as its more of an exercise than actual practical use (it's been a while since I fiddled about winding my own transformers and dealing with poles/zeroes..) I guess its all part of the fun.  Given I'll need a split rail, I'll have to use a transformer anyway, so I might as well make it isolated, perhaps with some opto feedback, but regulation isn't all that critical (14-16V).

One thing I was thinking though, my class D isn't self-oscillating, but classical using a fixed frequency triangle wave for the PWM, at 200kHz.  Does anyone think it would be prudent to use this same frequency for the of 9-20V to +/-15V converter?  I was thinking that if I use an off-the-shelf device with its own oscillator, if thats around 200kHz, the two could generate some interference in the audio band.  I know many devices have a "sync" input, and as the load should be fairly constant, I'm hoping it won't have to go into discontinuous mode, so at least the switching frequency of everything will be the same.

Am I being paranoid?

[Unixon]

How about abusing a high frequency DC-DC converter with 50Hz sine mixed in feedback network and AC coupling that to the load?

[Buriedcode]

Well, essentially thats I'm doing with a class D, except AC coupling a power input with low impedance requires a whopper of a capacitor.

As its more of a test, I might just prototype both the half bridge and full bridge versions, as the full bridge makes the input boost/buck much easier, but the half bridge lets me keep my ground.  I was mostly asking if syncing the two converters would pre-empt potential problems with EMI.

[Unixon]

So, the solution is 12->9V DC + sine gen + D-amp + RLC or just sine gen + D-amp + RLC and 12->9V conversion is by limiting max duty cycle. Right?

[Buriedcode]

So, the solution is 12->9V DC + sine gen + D-amp + RLC or just sine gen + D-amp + RLC and 12->9V conversion is by limiting max duty cycle. Right?

Well, no, 9V RMS is +/-12.75V so I'll need +/-15 supplies for a half bridge, but only 15V DC for a full bridge.  The sine gen, is the input to the class D.  The 15V supply gives enough headroom, as the rails will sag under load, and some crude feedback shoud keep the output fairly stable over load variation, but the load I'm using is pretty much fixed.

I don't know if this will be particularly efficient, as I've got effectively two converters in series, and its certainly a lot of effort just to replace a largish mains transformer - but as I said in my first post, this is a challenge too, not something for economical purposes.

[Zero999]

If it just rectifies the 9VAC, then a 12V squarewave will work perfectly. No need for any filtering, or PWM, just a 50Hz square wave fed into an H-bridge.

[Buriedcode]

If it just rectifies the 9VAC, then a 12V squarewave will work perfectly. No need for any filtering, or PWM, just a 50Hz square wave fed into an H-bridge.

It doesn't, it uses voltage doublers for +/12V supply, and a multiplier for ~100V.

A perfect sinewave probably isn't needed, but given all the diodes internally are 1N4001, I'm hesitant to try a square wave.  That said, it can't hurt to test it, and if it creates problems with the audio I'll know.

[Zero999]

If it just rectifies the 9VAC, then a 12V squarewave will work perfectly. No need for any filtering, or PWM, just a 50Hz square wave fed into an H-bridge.

It doesn't, it uses voltage doublers for +/12V supply, and a multiplier for ~100V.

A perfect sinewave probably isn't needed, but given all the diodes internally are 1N4001, I'm hesitant to try a square wave.  That said, it can't hurt to test it, and if it creates problems with the audio I'll know.

That doesn't make much difference. Those circuits will work perfectly, if not even better off a 12VAC square wave, than a 9V sine wave. The only think you might want to do is limit the turn on/off times of transistors, if it causes noise, or perhaps just use ferrite beads.

[Buriedcode]

Well I just played about with LTspice, and when simulating the power supply section of the unit with a square wave, 100Hz +/- 12V, the outputs were fine, but there were some pretty hefty current spikes on the input.  I limited the rise/fall times a great deal and it dampened them a fair bit, but it was still 4A+.  But thats of course with an ideal voltage source with zero resistance.

So I simulated a full bridge, quick and dirty, picked the MOSFETs based on what I had to hand, but probably overkill on the MOSFET driver.  Dead time was crude, but seems to work, at least for the simulation/switches.  The pink trace is the input to the 7912, which is the most sensitive (least headroom) to input voltage.  The dull yellow is the power in current.

Although the units manual claims it draws 12W, I'm guessing by the schematic, and the temperature of the back of the unit after a while that it wastes a good chunk of that in the linear regs.

So I guess the only way to know if it a) works, and b) doesn't interfere with the high analogue on board, is to test it.  Given this is only running at 100Hz (or at least <1kHz) and relatively low power (<2A) should I slow down the switches by limiting the gate drive?



 

[Zero999]

Well I just played about with LTspice, and when simulating the power supply section of the unit with a square wave, 100Hz +/- 12V, the outputs were fine, but there were some pretty hefty current spikes on the input.  I limited the rise/fall times a great deal and it dampened them a fair bit, but it was still 4A+.  But thats of course with an ideal voltage source with zero resistance.

So I simulated a full bridge, quick and dirty, picked the MOSFETs based on what I had to hand, but probably overkill on the MOSFET driver.  Dead time was crude, but seems to work, at least for the simulation/switches.  The pink trace is the input to the 7912, which is the most sensitive (least headroom) to input voltage.  The dull yellow is the power in current.

Although the units manual claims it draws 12W, I'm guessing by the schematic, and the temperature of the back of the unit after a while that it wastes a good chunk of that in the linear regs.

So I guess the only way to know if it a) works, and b) doesn't interfere with the high analogue on board, is to test it.  Given this is only running at 100Hz (or at least <1kHz) and relatively low power (<2A) should I slow down the switches by limiting the gate drive?

A rectifier & capacitor will draw current spikes. A sine wave is worse than a square wave for current spikes, because the capacitor only charges during the peak of the sine wave and it will discharge, whilst the AC voltage is low.

Here's a simulation I did awhile ago with a bridge rectifier & capacitor, with a sine wave input.
https://www.eevblog.com/forum/projects/why-the-waveform-of-a-transformer-is-distorted-when-drawing-current/msg1199129/#msg1199129


Here it is with a 34V peak square wave input, rather than a 24VRMS sine wave.

[BrianHG]

Warning about such audio circuits.  Depending on their bridge and cap configuration, if they use 9v AC, they may be seeing it as 18v RMS peak to peak, IE ~25v peak-peak.  This means 12v with a half-bridge wont work as the output will be 12v peak-peak, 6v AC.  You need a full-bridge to convert the 12v to a 12v AC waveform, 24v peak to peak.

Also, now, your source 12v supply & your inverter circuit must not have a connection to mains earth on it's GND, otherwise that bridge will screw you if you audio device has a GND connection to another device, like a PC audio card where the PC's sound card's GND is wired to the EARTH GND.  You will drive a ton of current through the GND when the polarity goes backwards.

A solution would be generating a -12v from your +12v, and use a half bridge from the +12v to -12v to generate your output 12vac/24v peak-peak signal to drive the amp.  This way, you keep the DC out 12v adapter's GND at GND and have a +/-12v swing at the output of your inverter.

If you really want a sinewave 50/60hz, you can use a cheap power op-amp like this one 'TDA7293V' powered from the +/-12v you created and feed the it's input with a MCU generated sine-wave.  Or, if you don't want a heat sink, there are plenty of clas D audio amps in a chip which can do also this (ones which support positive & negative supply operation), but, you will need a choke inductor on the output.  MCUs with PWM outputs can drive your mosfet H-Bridge as well to make your square wave or sine wave output.

[Buriedcode]

Warning about such audio circuits.  Depending on their bridge and cap configuration, if they use 9v AC, they may be seeing it as 18v RMS peak to peak, IE ~25v peak-peak.  This means 12v with a half-bridge wont work as the output will be 12v peak-peak, 6v AC.  You need a full-bridge to convert the 12v to a 12v AC waveform, 24v peak to peak.

I've simulated as best I can the power supply circuitry on the device, its pretty simple, and it seems 9V RMS is a minimum to prevent ripple on the negative rail, so 12.75V peak for a sine wave.

Also, now, your source 12v supply & your inverter circuit must not have a connection to mains earth on it's GND, otherwise that bridge will screw you if you audio device has a GND connection to another device, like a PC audio card where the PC's sound card's GND is wired to the EARTH GND.  You will drive a ton of current through the GND when the polarity goes backwards.

That is why I originally was going with a half bridge (see previous post), so I could keep my "ground" - even though I might as well make the input converter isolated anyway.  A full bridge would require the input converter to be isolated, rather than having it optional. 

A solution would be generating a -12v from your +12v, and use a half bridge from the +12v to -12v to generate your output 12vac/24v peak-peak signal to drive the amp.  This way, you keep the DC out 12v adapter's GND at GND and have a +/-12v swing at the output of your inverter.

See previous post.  I originally intended to use a half bridge class D, albeit a rather crude one - just with opamps, drivers and MOSFETs, before testing out a standard square wave full bridge.

It seems, whilst i was worried about the input current spikes of using a square wave, a bigger concern is the ripple current in the capacitors in the doubler circuits for the +/12V rails. (see LTspice schem, right side, C18 and C19).  With a nice 50Hz sine wave @ 9V RMS there's about 450mA RMS ripple on the positive, and 550mA on the negative.  With a 50Hz square wave, thats pushing 1A RMS.  I checked a few datasheets for various generic 470uF 25V caps, and it seems the max ripple current is between 700-900mA depending on the manufacturer.  I know I could just feed it from a full bridge and see if anything gets too hot, but I don't like just "seeing if anything blows".

If I increase the frequency to 200Hz brings this down to ~500mA, which is more sensible, and shouldn't make much of a difference to the multipliers as its still a low frequency.

So, really I'll have to decide (or test) between:

1) Full bridge, square wave @200Hz, driven by an isolated converter to provide 13-15V from 9-18V in.
2) Half bridge, square wave @ 200Hz, driven by a converter (isolated or not) to provide +/-15 from 9-18V in.
3) Half bridge class D providing a nice sine wave @50-200Hz, driven by...  converter (isolated or not) to provide +/-15 from 9-18V in.

I realize that most if not all off-line SMPSs these days are isolated, but I cannot guarantee it, so it would be nice to have it isolated.

All this is certainly not worth the cost/effort just to replace a simple 230 - 9V 2.2A AC transformer, but as I said in the first post, its more of an exercise than anything else.  And the above options aren't easy, but by no means impossible, its just a question of how efficient it will be and how much hassle it really is.

[Zero999]

I was playing with LTSpice today and remembered this thread. How about just using a discrete BJT bridge, driven by the classic astable circuit?

[Buriedcode]

I was playing with LTSpice today and remembered this thread. How about just using a discrete BJT bridge, driven by the classic astable circuit?

I'm sure it could, provided its running higher than 50Hz (see above post about ripple current in voltage doublers).  You're right in that for the full bridge sim I was using the same driver/switches as I used for the Class D, but running at 200Hz (rather than 200k) so it was overkill.  BJTs should be a bit slower than MOSFETS too, so I imagine EMI/ringing wouldn't be a problem.

Right now I'm working on an Isolated supply - which is required for a full bridge, and a "nice to have" for the other options.  I've settled on an isolated flyback based on the TPS55340, with a dual winding on the secondary so I could have a split rail if needed.  I haven't used this device before, or worked out compensation with an optoisolator, so it'll be a learning curve.

Once I have something working though, I'll try the simple full bridge astable first. Cheers

[djerickson]

I went through a similar thing recently. The Mutable Instruments Ambika Synth uses a 9V AC transformer. I wanted to use +12V only. Apply +12V to the 9VAC input. That will take care of the internal +12V and +5V. No mods so far. For the -12V which is probably low current,  find a place inside for a +12V to -12V low-current DC-DC. I used a 3W MeanWell SPBW03F-12 that was lying around. There are many similar 1W to 3W parts for < $10. Add extra bypass caps if you want to reduce noise. If the -12V current is low enough (<50mA) then one of the 12V charge pump ICs will work.

Your choice: either use a nice cheap and simple internal change or a complicated outside solution. Took me many hours to think about it and about 2 hours to build. It works swell.

The mods to the unit can be easily un-done (3 wires: +12V, -12V and GND) if you ever need to return it to its original splendor.

Good luck,
Dave Erickson
www.djerickson.com

[Zero999]

I've had another idea. Rather than make the astable from BJTs, use MOSFETs, which can also drive the load.

[Buriedcode]

I went through a similar thing recently. The Mutable Instruments Ambika Synth uses a 9V AC transformer. I wanted to use +12V only. Apply +12V to the 9VAC input. That will take care of the internal +12V and +5V. No mods so far. For the -12V which is probably low current,  find a place inside for a +12V to -12V low-current DC-DC. I used a 3W MeanWell SPBW03F-12 that was lying around. There are many similar 1W to 3W parts for < $10. Add extra bypass caps if you want to reduce noise. If the -12V current is low enough (<50mA) then one of the 12V charge pump ICs will work.

The -12V rail is definitely >100mA so a chargepump is out, but I'll have to measure it properly when I replace its regulator.    I as hesitant to try anything other than a "clean" rail because there appears to be quite a lot of analogue on it - not just the power rails of opamps where PSRR will help.  So It would probably need to be -15V for headroom for a linear reg, or at least a filter.

Add to that I would need to add my own 100V boost converter in there to replace the rather large multiplier.

Whilst it would of course be cheaper, and easier to mod the unit, as I said, I'm wondering just how difficult it would be as my own project.  With all these minor mods, it adds up to being rather substantial, and I pretty much have the parts to hand for either solution (modding, or 9-15V -> 9V AC adapter).  So I'm picked the more challenging one.  It should also mean less harm to the PCB which is now 24 years old and has been repaired rather a lot.

There are however, other reasons for modding it as you mentioned - I'm guessing about half of the power is wasted in those three regulators, as the back of the unit gets quite toasty.  Replacing the 5V rail with a buck module (+post reg for lower noise), and the +12V with an LDO would allow one to use 13V ish, with little heating. 

As I said, I have the parts for both solutions, so I'm using the modding as a backup if the project stalls (likely). 

[Buriedcode]

I've had another idea. Rather than make the astable from BJTs, use MOSFETs, which can also drive the load.

Cheers for this.. funnily enough I believe I have both those MOSFETS in stock ( back when I salvaged parts from laptop boards..), so I could prototype that in an afternoon.  Could I increase the value of the inductor for rounding off edges more?

[djerickson]

Here is the schematic for my Ambika +12V mod. Ambika uses +/- 8V at 130mA for the analog circuits, from '7808/'7908 regulators. So +/- 12V worked fine.
Enjoy,
Dave

[Zero999]

I've had another idea. Rather than make the astable from BJTs, use MOSFETs, which can also drive the load.

Cheers for this.. funnily enough I believe I have both those MOSFETS in stock ( back when I salvaged parts from laptop boards..), so I could prototype that in an afternoon.  Could I increase the value of the inductor for rounding off edges more?

The inductor is there to limit the current due to shoot-through, when both the top and bottom MOSFETs turn on simultaneously. D1 is there to clamp the voltage, for light loads, otherwise the maximum voltage rating of the MOSFETs will be exceeded.