driving transformer primary

[johnfm]

Hi,

I need a circuit to provide a few mA at 3-5V across an isolated barrier. The hard part is that my total available power on my primary side is 5V at 3mA.

I am trying to debug a low power oscillator/xfmr/rectifier circuit. I seem to have problems driving my primary side xfmr with the output of a schmitt trigger inverter IC. The input waveforms look nothing like my oscillator output. Should I try driving a FET, rather than the primary directly (any idea where I can get low power FETs for driving inductive loads)? Any suggestions for improving the driver circuit? Thanks for your time...

 

[BrianG]

Hi John, can you provide a bit more information please.
What is the frequency of your oscillator? What logic family schmitt inverter are you using? Is 3mA your maximum power budget or just a target? Details of the transformer: primary inductance and d.c. resistance, turns ratio, type of core (i.e. iron laminations, pot core, etc.)

 

[OperaHouse]

Are you really limited by that 3ma on the input? That requires more sophisticated solution than I have used. Just wanted to mention that small line filter transformers operted at about 80khz are a great low cost solution for this purpose. Given the low voltage, diode losses are going to be a problem. I'm sure someone makes a little module that does this.

 

[johnfm]

Thanks guys,

I can adjust my oscillator frequency as needed, but it's around 100KHz at the moment. 3mA is my absolute max available current on the primary side. My schmitt inverter IC is the TI SN74AC14 (could be changed if necessary). The proto xfmr I'm using is the coilcraft Q4470-C (400uH, ~1 ohm DCR, 1.3:1 turns). I'm looking to use an off-the-shelf xfmr to keep costs down.

I haven't been able to find an all-in-one solution as yet, but will take a look at the line filter concept. I'm hoping a 1.3:1 ratio, xfmr efficiency and schottky losses will give me around 3.3V on my seconday side.

I'm not much of a xfmr guru... so I appreciate your input. Thanks for getting back to me...

 

[nbucska]

Is there a way to redesign/reduce the load?

<nbucska@pcperipherals.com>

 

[johnfm]

Unfortunately not. We've simplified the secondary side as far as we can go...

 

[nbucska]

Hi John:
In my 30+ yrs as engineer I learned that when a job is impossible, it is often worth to step back and take
a larger wiev; or do something else.

Can you give me more data ?


<nbucska@pcperipherals.com>

 

[johnfm]

You may be right! But for now, I have signed on to try present a solution with the aformentioned specs.

I provided a little more info on the xfmr part I am looking at in a previous post. Please let me know what additional data you would like, and I'd be happy to provide it.

Thanks for your time...

 

[nbucska]

What is the max. VTG difference across the barrier?
Could you use charge pumps instead of trafo-s?

What is the load circuit ? if it is a sensor, could you
operate it in sampled mode with lower duty cycle ?

Instead of diodes, could you use FETs driven with a second trafo ?

Could you use higher secondary VTG with lower current ?





<nbucska@pcperipherals.com>

 

[johnfm]

Hi, responses are embedded below...

What is the max. VTG difference across the barrier?

2V max difference across the barrier (5V in at primary is fixed, 3V out would be my minimum).

Could you use charge pumps instead of trafo-s?

If it can provide at least 500VAC isolation then it should be acceptable.

What is the load circuit ? if it is a sensor, could you
operate it in sampled mode with lower duty cycle ?

Load circuit consists of a small uP, A/D and some signal conditioning circuitry.

Instead of diodes, could you use FETs driven with a second trafo ?

Again, the requirements are: 500VAC isolation, ~3.3V, 2.75mA min output from 5V, 3mA input. The method for achieving this is negotiable as long as the requirements are met.

Could you use higher secondary VTG with lower current ?

No, the uP for example will draw the same Iq regardless of voltage (I believe 5VDC is the max voltage for the micro anyway).

Thanks again...

 

[walker1]

Could your problem with driving the trafo be due to DC in the primary coil.
That is, one end of the coil is connected to Vcc or Gnd and the other to the CMOS output ?

Have you tried to connect the trafo in push-pull with an inverted signal in the other end ?
That should reduce the required input current by a factor 2. (and the transformer ratio will need to be increased by a factor of 2 as well

Have you tried to boost your switch power by connecting two or more CMOS inverters in parallel ?

Is your trafo driven by the same schmitt trigger-inverter that is used as oscilator ?
(May not be the best idea)

 

[Skogsgurra]

There is no need to use anything else for a driver. The 74ACxx family is designed to source and sink 24 mA. So do not burn your precious milliwatts in unnecessary extra circuitry.

Use a series capacitor to block off any DC in the primary (or use an exact 50 % duty cycle and double 180 degrees out of phase drivers, as walker said). Is your power constraints due to fear for overheating the 74AC14 or is no more power available on the primary side? If the latter is the case, then you will have to look at the efficiency of the transformer and rectifier. If you have lots of power available then you could easily draw 10 mA and more from the 74AC14. No sweat.

 

[johnfm]

Thanks guys,

I did add the blocking cap and it made a big difference. I have not tried the push-pull approach with the inverted signal on the other end (I presently have one side of the primary coil connected to GND). Will this improve the efficiency of my power transfer?

My oscillator is not generated by the schmitt inverters... I am using a low-power RC/comparator scheme.

At the moment, I do have two inverters in parallel driving the coil, but I'm not sure I need it. The power constraints I mentioned (3mA) is the total available on the primary side... I am stuck with this number, so I really have very little margin to work with.

I will try the push-pull technique. Thanks for your input, I appreciate it...

 

[OperaHouse]

And where does the 5V come from, since 5V doesn't occur naturally on earth. If you could operate from the unregulated source you would have something. Are you trying to operate off something like an interface port?

 

[electricuwe]

I'm afraid that walker1's post is a bit misleading:

Using a push-pull configuration might reduce currents in the primary winding but not the DC input current requirement.

Thats just determined by the law of energy conservation. No load losses should be your major concern in such a low power configuration

I would recommend the following:

-Do not change to push pull
-Use 74AC.. to drive your primary as recommended by skogsgurra
-Use a center tapped rectififer with low volatge schottky diodes. This will result in a voltage drop of about 0,3..0,4V. Stay awa from the synchronous rectifier approach recommended by nbucska. This drive power you need will be higher than the gain due to reduced voltage drop
-taking the diodes into account you have a power of roughly 1,3 V * 3mA + 5 V * (3mA-2,75 mA)= 3,9 mW + 1,25mW = 5,15 mW to cover all the losses
-try to calculate the no load losses of your design as a function of frequency, taking into account oscillator and driver as well as core losses of transformer
-You might have to try several transformer designs with differnet flux densities and core sizes and materials.
-minimize no load losses

A tough project, to my mind, but not impossible to solve.

 

[logbook]

John,
there is a more fundamental problem here which the guys haven’t mentioned. The primary current is so high because the primary inductance is too low for the frequency you are using. I did a quick SPICE sim of a 400µH inductor in series with a 1 ohm switch, running from a 5V supply. I bled the stored energy off into a diode to reset the current to zero at the end of each cycle. The peak primary current was over 60mA. This is a good starting point for an analysis.

When you apply the rectangular drive waveform to the coil you force a voltage across the primary. The primary inductance then causes the current to ramp up. Let’s ignore the secondary completely. If the primary current gets excessive without worrying about the secondary then you have a problem.

E=L*di/dt
di= E*dt/L = 5 * 5E-6 / 400E-6 = 62.5mA.

The RMS value of this current is
Irms = Imax * sqrt(D/3)
where D is the duty cycle of the triangular waveform, in this case 0.5
Irms = 25.5mA.

Now you are not seeing this much current. One reason is that your drivers are limiting the current to perhaps 50mA. Another reason is that the primary energy is not being discharged at the end of each cycle so the flux becomes continuous rather than discontinuous (this is standard switched-mode converter terminology). The capacitor works by preventing the primary current from becoming continuous.

It is not possible to say that just increasing the frequency will reduce the power. This will require experimentation. The diode losses will increase with frequency, but at these low current levels, small-signal schottkys should be good for at least 1MHz. The core losses in the transformer will also increase with frequency. You could experiment with a much shorter on-period for the primary. This should cut down on the primary power.

The energy stored in the primary leakage reactance also causes a loss. If you can resonate this, rather than dissipating it, you will improve the efficiency. You would need to do a bit of study on switched-mode converters however.

You may have missed the part in electricuwe’s post about using a centre tapped output. The idea is to make a full wave rectifier using only one diode in the series path for each part of the conduction cycle. This is essential for efficiency at low voltage levels. So you use a centre tapped output and two diodes to make a "bi-phase bridge".

 

[lcsjk]

Whoaa there folks! Let's don't get too carried away with this one! The impedance of a 100 uH inductor at 100 KHz is 2*pi*freq * Inductance, giving an impedance of 251 ohms for this transformer for a sine wave. That tells me that 5 volts can only drive 20 milliamps into the primary with an open secondary.

As suggested earlier, the transformer must have a chance to reset itself. The volt-seconds for the ON cycle must equal the volt-seconds for the OFF time. This can be accomplished by using a film capacitor in series with the primary or driving with a 50 percent duty cycle source.

Now if the secondary has a load, the primary will see this load also, increased or decreased by the turns ratio.

To avoid losses, the secondary needs to be a full-wave center-tapped topology. YOu probably will have to insert an inductor before the capacitor to avoid the short pulse each half cycle as the output current is the capacitor is re-charged.

I think you will also have to drive the input from a push-pull source. Note that using a 0-5 volt input the transformer will have a +/- 2.5 volt primary voltage.
Using a push pull from two of the inverter outputs, the transformer will change between +5 and -5 volts giving an effective +/- 5 volts on the primary.

I suggest you drive the transformer primary from your oscillator circuit and 74ac14 and measure the current. If it is too much you will have to get a higher impedance transformer, or go to a higher frequency.

 

[johnfm]

Thanks all,

I really appreciate your input. I have implemented some of the design ideas described and we are getting close. I am not seeing the high currents described by logbook, I can get nearly 50% efficency with the blocking cap and the inverter-driven push-pull on my primary. This is better than I was getting with the single-ended driver, and gets my nearly to my goal... although I would think at these low currents I could do somewhat better than that.

I am curious to try the center-tapped secondary with bi-phase bridge, as I'd like to minimize my diode-losses... I am using a 4-schottky bridge on my output at the moment. I'm not sure I understand the concept of the inductor in series with the cap on the primary... lcsjk, can you expand on that a little? I will also adjust my frequency and watch my primary current.

Thanks again for all your help...

 

[Heydave]

If your isolation requirement is primarily DC, you can dump the transformer and use a complementary drive to matched caps. On the "secondary" side, use a schottky bridge to get back to DC. This should be good for a few ma. If necessary, use a common mode choke to filter the residuals and possible provide some AC isolation. You could use 4 phases, 4 caps and two bridges if you want to overlap and reduce filter cap requirements. Drive can be direct from CPLD, FPGA, or something like a 74AC74 clocked at about 1 MHz. It should be cheaper than a transformer. AC coupling between windings of a transformer can transfer noise across the isolation. The complementary cap technique can be used to transfer data as long as you use an encoding scheme that has no DC such a manchester. The technique can use less power and run faster than typical opto couplers but it depends on what your AC isolation requirements are.

 

[johnfm]

Thanks heydave. Any idea where I can find a little more info on the matched cap technique? Sounds like it's worth exploring this as well...