I apologize for this rather long post but I need to explain what I am trying to achieve. I have a number of vintage cars (1920s and 1930s) all of which have 12 volt third brush generator systems. This system effectively outputs a constant current into the battery via an electromechanical cut-out. The early cars have an on/off switch so that the charging can be stopped when the driver judges that the battery is fully charged whereas the later cars have a three position switch which switch resistors in series with the generator field coil to provide a low or medium charge rate or a high charge rate when the headlights are in use. These systems work well enough with batteries that can be refilled with water on a regular basis but it is not safe to use them with maintenance free batteries. There is a real danger of explosion due to a build up of gases with nowhere to go. Unfortunately, at least here in regional Australia, maintainable batteries are no longer available.

There are a number of possible solutions such as converting to a two brush system and fitting a conventional voltage regulator. Another method involves shorting the generator output with a variable mark-space ratio controlled by the battery terminal voltage but this requires an electronic cut-out (diode). Neither of these solutions appeal to me as I wish to keep the cars original – any modifications must be easily reversible.

Accordingly I have designed a simple voltage regulator that emulates the switched resistor scheme described above. The generator output voltage controls the gate-source voltage of a MOSFET which acts as a variable resistor to control the generator field current – the schematic is attached. The device can be installed or removed in minutes and requires no other modification to the car.

And now, at last!, to my problem. The system works very well but on the oscilloscope I can see an instability in the 10 – 50 Hz range with an amplitude of around 200mV peak to peak on the output. The small capacitor shown on the schematic filters out any commutator noise. If I increase this to 500 microfarad the instability worsens considerably and the frequency is at the sub Herz level. Obviously this is a negative feedback design and obviously there is a time constant involving the MOSFET resistance and the field coil inductance. That and the time constant associated with the 500 microfarad capacitor is presumably giving me a phase reversal at a low frequency. Unfortunately it is not clear to me what is causing the instability with the small capacitor.

Perhaps someone with a deeper knowledge of DC generators could enlighten me or, better still, come up with a solution!

 

I think you mean they are 6V systems.

What's likely happening is that when the voltage gets high enough and the MOSFET turns off to switch in the resistor, the battery voltage drops slightly.
This turns the MOSFET back on and the cycle repeats, giving the oscillation you see.
That's not necessarily harmful and is similar to how most automobile voltage regulators work.

To minimize that you can add some hysteresis (making the turn-on voltage slightly higher than the turn-off voltage) by connecting a resistor between the collector of Q2 and the base of Q1.
100kΩ will give about 0.1V of hysteresis.
You can adjust its value until you get the desired circuit operation with no oscillation.

But note that under normal operation, depending upon the resistor value and the battery load, the circuit may always switch on and off at some low frequency rate to keep the battery properly charged.

 

Hi Crutschow. Thanks for your reply. Definitely 12V system - British cars not American. Resistor R8 is not actually fitted, as noted. I thought it might be necessary to provide some initial field current to get the MOSFET to turn on but it isn't. There is enough remanent magnetism in the field coils to generate a volt or so to turn on the MOSFET which then acts as a variable resistor.

When the battery voltage rises, Q1 and Q2 both conduct more which reduces Q3 gate source difference which increases the drain source resistance. In turn, the field current falls, the generator output falls and the battery voltage falls to maintain equilibrium. In a linear system the output should settle at a fixed level unless the negative feedback turns positive(i.e. the open loop transfer function frequency response exceeds 6dB/octave). Other than the obvious time constant of the MOSFET resistance and the field coil inductance I have no idea what the transfer function of the generator is - particularly with respect to the magnetic field distortion caused by the armature current in the field coil.

Thanks for the hysteresis idea but I think the cure might be worse than the disease . It will entrench the variation I am seeing rather than eliminate it. Nevertheless I will try it next time I have the regulator out of the car and in my test rig.

 

If you want a linear system then you certainly don't want to add the hysteresis.
I was envisioning a bang-bang type control with the MOSFET full on or full off.

For linear control, depending upon the field current, you can end up dissipating a fair amount of power in the MOSFET (worst-case is at 1/2 the maximum current).
To avoid that dissipation you may want to consider PWM control of the current.
The control loop would still be linear but the control of the current would be efficient PWM.
(Here's a reasonable simple circuit to generate the PWM signal. Your control signal would go to the Mod_In node, pot not needed.)

To stabilize the loop you can try compensation with a Miller capacitor between the base and collector of Q1, value experimentally determined.

 

Yes, I agree that my circuit is not very efficient, but no less so than the switched resistor scheme I outlined in my original post. The MOSFET power dissipation is around 10 watts worst case and causes only a small heat rise in the diecast box to which it is attached, enough however to persuade me to mount the temperature compensating thermistor external to the box. The whole system is very inefficient. My test rig uses a 1/3HP motor driven by a VFD and it is barely up to the job . That's around 250 watts to get 100 watts out of the generator.

Thank you for your idea of the PWM circuit. Not dissimilar to a class D amplifier circuit with which I am familiar. The only snag I initially see is that the circuit has to be self powered. As there is no conventional ignition switch (magneto ignition), I can't switch power to a regulating device - it must self excite. But it shouldn't be too difficult to arrange the FET to be on to allow the generator voltage to build up before the control circuit takes over. I will give it some serious thought.

I think adding a Miller capacitor will be no different to increasing C1 which I mentioned in the original post.

Thanks again for your suggestions here. Much appreciated.

 

..causes only a small heat rise in the diecast box to which it is attached, enough however to persuade me to mount the temperature compensating thermistor external to the box.

I'm confused.
What exactly are you compensating for?
I thought is was the charging voltage by placing the thermistor near the battery, but you seem to be implying something else.

I think adding a Miller capacitor will be no different to increasing C1 which I mentioned in the original post.

True.
It's just that the Miller capacitor value would be reduced by the gain of the transistor for a given compensation time-constant.
You should be able to brute-force compensate the loop by increasing the capacitor value sufficiently.
Sometimes a small resistor in series with the capacitor (a few tens of ohms to a few k ohms) to add a little lead compensation can help the stability.

 

I'm confused.
What exactly are you compensating for?

Sorry to confuse you! A lead acid battery needs temperature compensation for optimum charging voltage, -26mV/deg C or thereabouts for a 12V battery. My circuit provided overcompensation until I moved the thermistor out of the box.


Excellent point about the Miller capacitance. I'd stopped at 500 microfarad with C1 as the next step of 5000 microfarad was impractical. I am familiar with lead-lag compensation so will give this a try. It might be some time before I report back as the regulator is in one of my cars which is in use .

 

Success! If I had been a bit more methodical from the outset I could have saved a lot of time. Whilst the regulator was in the car (working well I might add), I measured the impedance of the field coil of a spare generator and calculated its inductance. Knowing its resistance and that of the MOSFET in its working range I could calculate the first pole frequency - around 80Hz. I figured that rolling off the amplifier gain at around 8Hz (lag) and flattening it again at 80Hz (lead) would probably fix the problem

Today I pulled the regulator out of the car, connected 20uF in series with 120 ohms across C1 (still needed to filter out commutator noise) and tried it on my test jig. The instability was gone.

Thank you Crutschow for reminding me of the usefulness of lead-lag compensation.

 

I apologize for this rather long post but I need to explain what I am trying to achieve. I have a number of vintage cars (1920s and 1930s) all of which have 12 volt third brush generator systems. This system effectively outputs a constant current into the battery via an electromechanical cut-out. The early cars have an on/off switch so that the charging can be stopped when the driver judges that the battery is fully charged whereas the later cars have a three position switch which switch resistors in series with the generator field coil to provide a low or medium charge rate or a high charge rate when the headlights are in use. These systems work well enough with batteries that can be refilled with water on a regular basis but it is not safe to use them with maintenance free batteries. There is a real danger of explosion due to a build up of gases with nowhere to go. Unfortunately, at least here in regional Australia, maintainable batteries are no longer available.

There are a number of possible solutions such as converting to a two brush system and fitting a conventional voltage regulator. Another method involves shorting the generator output with a variable mark-space ratio controlled by the battery terminal voltage but this requires an electronic cut-out (diode). Neither of these solutions appeal to me as I wish to keep the cars original – any modifications must be easily reversible.

Accordingly I have designed a simple voltage regulator that emulates the switched resistor scheme described above. The generator output voltage controls the gate-source voltage of a MOSFET which acts as a variable resistor to control the generator field current – the schematic is attached. The device can be installed or removed in minutes and requires no other modification to the car.

And now, at last!, to my problem. The system works very well but on the oscilloscope I can see an instability in the 10 – 50 Hz range with an amplitude of around 200mV peak to peak on the output. The small capacitor shown on the schematic filters out any commutator noise. If I increase this to 500 microfarad the instability worsens considerably and the frequency is at the sub Herz level. Obviously this is a negative feedback design and obviously there is a time constant involving the MOSFET resistance and the field coil inductance. That and the time constant associated with the 500 microfarad capacitor is presumably giving me a phase reversal at a low frequency. Unfortunately it is not clear to me what is causing the instability with the small capacitor.

Perhaps someone with a deeper knowledge of DC generators could enlighten me or, better still, come up with a solution!

View attachment 132807

Motorcycle alternators vaguely resemble current sources - Early British motorcycles regulated with a dirty great Zener. Near as I've been able to discover - it was a 100W 15V part. You'd probably need more to regulate a car dynamo.

Maybe a lower voltage Zener to compensate the Vbe drop and multiple parallel shunt power transistors on plenty of heat sink.

 

Motorcycle alternators vaguely resemble current sources - Early British motorcycles regulated with a dirty great Zener. Near as I've been able to discover - it was a 100W 15V part. You'd probably need more to regulate a car dynamo.

Maybe a lower voltage Zener to compensate the Vbe drop and multiple parallel shunt power transistors on plenty of heat sink.

Thanks Ian. My generator puts out only 8A which at 14V would be 112 watts so you are not far off. Your solution would work but it's a bit crude! I am now very happy with the circuit I have developed so a PCB is the next project.

 

Thanks Ian. My generator puts out only 8A which at 14V would be 112 watts so you are not far off. Your solution would work but it's a bit crude! I am now very happy with the circuit I have developed so a PCB is the next project.

A circuit to control field current in an old style dynamo isn't impossible, an alternator type might be converted as long as you know whether one end of the winding is grounded or hot.

Electromechanical regulators have disconnect contacts that open when the armature winding voltage is less than the battery, a solid state regulator requires a series diode on the dynamo output to stop the battery discharging back into the coil. It should cope with the Vf of a regular silicon diode. A Shottky barrier diode has lower Vf, but the come in low voltage ratings and might not survive in that corner of the automotive electrical environment.