The Fun Projects voltage regulator

It is a known limitation that the Model T Ford can suffer from an overcharged battery because of the absence of any voltage regulation. While the 3rd brush generator allows one to set the charge current, which remains fairly constant over a wide generator speed, there is nothing to reduce the charge when the battery is full. So, if one has set the charge current to say 10A in order to run the headlights continuously, when the headlights are off the battery will receive this 10A, even after it is fully charged. That's not good for the battery, apart from it losing more electrolyte than normal. In practice, the best compromise with the stock standard set up is to set the charge current to 5A which won't damage the battery, and will allow moderate headlight usage.

Principles of Automotive Voltage Regulators.
The principles of voltage regulation for automotive electrical systems were known back in the 1920's, but Ford with its economy focussed ideals, ignored it.
The method used, which has remained to the present time, is to reduce the generator (or alternator) field current when the battery has reached full charge.
Until solid state regulators started to become common in the 1970's, the regulator was an electromechanical device built along the lines of a relay. One set of relay contacts was placed in series with the field coil supply. Shunted across the contacts was a resistor; the purpose of which set the minimum field current and also protected the contacts against the arcing that would result from the rapidly collapsing magnetic field of the field winding.

This set of contacts was pulled in by the voltage coil. Effectively, the voltage coil is connected across the battery and its pull in power is obviously dependent on battery voltage. Therefore, the contacts are set to open when the battery has reached about 7V for a 6V system, or 14V for a 12V system.
The charge current is severely reduced, and the battery does not overcharge. Along with the voltage coil and contacts, are also a current coil and cutout with their own sets of contacts. The current coil senses battery charge current and its contacts open, again reducing field coil current, if the charge current is too high - such as when the battery has a low charge. This is required to protect; a) the battery, b) the generator brushes and commutator, and c) the generator windings from excessive current. Finally, the cutout coil senses current direction. Its contacts are between the generator output and battery. If current flows from the generator to the battery the contacts close, and if the current from the battery tries to flow back into the generator (e.g., when the engine is stopped) then the contacts open. This prevents the battery discharging into the generator when the car is not in use.
The design of alternators is such that they are inherently current limited, by means of winding inductance and that they generate AC. The rectifier diodes used to change the three phase AC to DC automatically prevent the battery discharging when there's no output. Hence, alternators do not require the cutout and current coils in the regulator assembly.
Electronic regulators which replaced the earlier mechanical types use a power transistor to control the field current. This is controlled by a zener diode voltage reference and comparator circuit. As alternators were standard by this time, there was no need to provide current control or a cutout in the electronic regulator assembly.

Regulation and the Model T.
An immediate problem trying to connect an accessory voltage regulator to the Model T, or any other car using a 3rd brush generator, is that there is only one terminal on the generator which is for the output.
The field winding is energised from this internally. So, to use a standard regulator involves having to separate this connection and provide a second insulated terminal. Indeed, this has been done. However, it looks out of place and involves wiring modifications. For those who just want a drop in regulator with no wiring modifications, an electronic regulator was developed by Fun Projects.

First Principles - Grounding Switch.

Switch short circuits generator output when no output is desired.

Well known in the Model T world is the so called "Grounding Switch". It made its appearance in an issue of "Tinkering Tips" many years ago. By means of this switch the generator could be shut off, if it was felt the battery did not require further charging. It was also thought that by shutting off the generator, some extra horsepower could be obtained from the engine.

To shut off the generator, we cannot merely open circuit the connection between the generator and battery, because of the excess voltage generated under no load, which can burn out the windings. But, because the generator is a constant current source, we can short out the output. Taking the output terminal to earth immediately robs the field coil of supply. It is perfectly safe to run the generator continuously this way, because as the field coil is not energised, the armature cannot produce any output.
Shorting the generator output to earth would cause a huge current from the battery to flow if it wasn't for the cutout. shown as a diode in the above diagram. The diode merely becomes reverse biassed when the switch is closed and thus no current from the battery flows through it. Importantly, when a grounding switch is installed, the cutout needs to be replaced with a diode type, if a mechanical type is present. This is because a mechanical cutout makes a direct connection to the battery when activated, and the switch would be shorting the battery current, limited only by the wiring resistance, until the cutout opened. Obviously, the switch will not last long used like this. It is true that the switch could be closed prior to the engine coming up to full speed, before the cutout closes, but human nature being what it is would eventually result in this point being forgotten.

Another method of controlling the charge rate is to simply divert more or less of the generator output to earth, as done by this period accessory:

(Photos courtesy MTFCA forum).

Here, a high power variable resistor is used to shunt the generator output. If, for example, assume the third brush is set to 10A. If the shunt resistor is set to draw 4A, then the battery will be charged at 6A. The resistor appears to be a carbon pile type. This is
a tube filled with carbon discs (hence the term "pile"). If the pressure against the discs is increased, they are pressed more firmly together, and the resistance drops. In this case, the adjustment is made by a screw threaded plunger.
While this method no doubt can save the battery from being overcharged, it is very inefficient with all the diverted current going up as heat. Additionally, the generator output is not actually reduced. The input power from the engine remains the same, as does the rate of wear on the generator. However, unlike the grounding switch, this one can be used with a mechanical regulator.

Principle of Regulation.
Having acquainted ourselves with the grounding switch, we can see that if the switch was closed when the battery voltage increased to 7V, and was opened when it fell below this, the battery would remain fully charged without risk of overcharge. The concept of the regulator is to do this automatically.
To take the generator output to earth, either a power transistor or a MOSFET could be used.  To control this, a switchmode controller or a comparator could be used. A switchmode control would gradually increase its 'on ' time (i.e. earthing the generator) as the battery reached 7V, and give the appearance of a variable current control when observed on the ammeter. A comparator would provide full charge until 7V was reached then shut the charge off altogether, until the battery voltage dropped to say 6.5V. This would show more of a pulsing type of indication on the ammeter.
Note that the charge current is still determined by the 3rd brush setting.

The Fun Projects Design.
A well known and popular accessory for the Model T is the Fun Projects Voltage regulator. It can be purchased at various Model T parts suppliers. It eliminates the potentially unreliable mechanical cutout, and prevents battery overcharge. It is a drop in replacement for the original cutout, looks identical, and needs no additional wiring. A further advantage is the crowbar protection to prevent generator burn out if the battery should be disconnected.

However, there is no technical information available on how it works, or the the internal design. As I'm often asked about the Model T electrical system, it was necessary to investigate and provide my own information.
The particular regulator in my possession was given to me on the premise that it was faulty, and could I fix it? As I discovered, the regulator was actually quite OK, but other wiring in the car was at fault. Nevertheless, I had been curious about the design and wanted to confirm my theory as to how it worked. So what better way than to examine the unit, and trace out the circuit. It could be of use for others who have a faulty regulator, or just want to build their own. Opening the unit simply involved a slight grind into a couple of spot welds. The cover then slips off.
I used a variable, current limited power supply, in series with the secondary winding of a 14V power transformer (primary not connected), to test the unit, which indicated it was functional. Upon raising the input voltage, a high frequency tone could be heard from the power supply, and connecting a CRO showed that indeed it was rapidly switching on and off.

Waveform at generator terminal. As battery voltage rises, the pulse width shown here decreases so that the generator output is earthed for a longer period. Frequency is about 6kHz.

Inside the unit, the cutout diode and switching device (both TO220 packages) are mounted on the chassis, the diode between the generator and battery terminals, and the switching device between the generator terminal and earth. This immediately confirmed the regulator is based on the well known 'grounding switch'. Mounted above, is the PCB with the other components, including a 16 pin IC. Remarkably, the IC,  diode, and switching device were devoid of type numbers. Thankfully, that need not present a serious obstacle.

Component side of the PCB.

The cut out diode.
Obviously, the diode was a dual Schottky type. I've serviced enough switchmode power supplies to have recognised it as such. Plus, logic tells that this would be the ideal type to use here. The reason being that a Schottky diode has less of a voltage drop across it than a silicon type. This means less heat dissipation for a given current, and less heatsinking is required. In fact, the body of the regulator housing is sufficient for this purpose. Typical types would include MBR30H100CT, rated at 30A and 100V.

The Switching device.
A bipolar power transistor could have been used, but given the current it would have to switch, and the lack of heatsinking, a MOSFET was the likely candidate. As the 'on' resistance of a MOSFET is a fraction of an ohm, it means very little voltage drop, even at high current. Hence, low power dissipation. Besides, MOSFETs are easy to drive straight from regulator IC's, as this one was. A further clue was the 470R gate pull down resistor between gate and source terminals. A bipolar transistor would not need such a low value of base resistor because the the base is current driven. With a MOSFET however, the gate presents as an open circuit, and the drive source needs to be low impedance to switch off the MOSFET very rapidly. A slow turn off will cause overheating due to operation over the linear part of the curve.
Indeed, a quick test confirmed it was a MOSFET. A good choice would a type such as IRF640 or STP60NE06-16. One important specification of the MOSFET used is that the gate voltage required to turn the device on is no more than about 5V.

The Integrated Circuit.
As to the IC, several options suggested themselves. But first, before getting carried away with guessing, it was necessary to trace out the circuit. As there are few external components, this was straightforward. Having done this, several clues became apparent. Two things indicated it was a switchmode regulator IC - this being gleaned from the .0047uF and 49.9K resistors - obviously the time constant of an RC oscillator. The one and only zener diode was 15V, and with a voltage that high, was clearly part of the crowbar protection. It was obvious that the voltage reference was internal to the IC; another feature of purpose designed switchmode regulator IC's. And of course, with my previous testing of the unit, the high frequency oscillation present was another confirmation.
Now, to get a block diagram of this IC to analyse operation requires knowing the exact type number. 16 pins immediately narrows it down to a few types. Then by looking at the function of the pins we can match them up. For example, the timing resistor and capacitor need to be on pins 6 and 7 respectively.
Also the output to drive the MOSFET needs to be pin 14, and/or 11. We can also see the output voltage is monitored at pin 2 via a voltage divider - obviously the input to the error amplifier. Using the values of resistors connected to this terminal, and knowing the regulation should commence at about 7V, the reference voltage of this IC was calculated to be around 5V. This immediately eliminated several types from the list of possibilities. Pin 9 would appear to be able to shut down the IC should the 15V zener conduct.
Type SG3524 fits these specifications.

Track side of the PCB. IC pin numbers and MOSFET pins are labelled.

The Circuit.

The MOSFET and Schottky diode have already been discussed. The IC receives supply to pin 15 via the chain of diodes D1 to D3. These are type 1N5819 Schottky diodes.
Apart from reverse polarity protection, this chain of diodes, and their associated  100uF 50V capacitors, also function as a voltage multiplier. Once the battery voltage reaches about 7V and the MOSFET starts switching, there is effectively an AC component present at the generator terminal. By multiplying this, we can get sufficient voltage for the IC to perform properly. Apart from increased drive to the MOSFET, the internal voltage reference is more accurate. There is about 13V present at pin 15 once regulation commences.
The charge retained in the 100uF connected to pin 15 would keep the IC powered up during 'dead time'; i.e., when the MOSFET is switched on.

 The output of the IC drives the MOSFET gate from two emitter follower output stages, at pins 11 and 14. In a push pull type power supply, these two outputs are used anti phase, but in a single ended regulator they can be paralleled. The supply also has to feed the collectors of the driver transistors, via pins 12 and 13.

The battery voltage is sensed via a voltage divider feeding pin 2; the non inverting input of the error amplifier. This is compared to the reference voltage (5V) internally generated in the IC, and made available at pin 16. Note that the input to the voltage divider is not the actual battery voltage, but the battery voltage plus the drop across the Schottky cutout diode and D1. We can work out the voltage at which regulation commences. With 5V across the 240R resistor, there will be 7V at the junction of D1 and D2. Adding 600mV onto this gives 7.6V at the generator terminal, and losing about 400mV across the cutout diode (Schottky) means about 7.2V at the battery, which is what we want. It would appear D1 and the associated 100uF decouple the voltage sensing input from commutator noise, and again provide reverse polarity protection to this part of the circuit.

The ramp generator to generate the variable pulse width relies on the resistor and capacitor at pins 6 and 7 to set the oscillation frequency. In this application their values are not critical, and they appear to have been merely selected as being mid way of the specified range of recommended values.
Crowbar protection for the generator (and the IC) is performed by the 15V zener. If the battery should be disconnected, the generator voltage will rise to an excessive figure. The zener conducts, and pin 9 starts to go high. This then forces the MOSFET to switch on and short the generator to earth. It may be wondered why pin 10 is not used despite being the official shutdown pin. This is because taking pin 10 high actually turns off the switching device, as would be required in the conventional kind of regulator circuit. Here, we want it to turn on the MOSFET to shut everything down. Pin 9 is the inverse of pin 10.

The 6V regulator could be simply modified should the owner wish to change to a 12V system, having already purchased the 6V model. The 100R resistor would need to be increased to 413R. The 100uF connected to the junction of D2 and D3 should be removed as we don't want the voltage multiplier functioning at 12V.

Likely Faults.
The most vulnerable component is the MOSFET and cutout diode. If the negative earth version of regulator was installed in a car that had been converted to positive earth, the cutout diode would immediately conduct. D1 and D3 will prevent the reverse polarity reaching the IC or the 15V zener (which would fully conduct on reverse polarity). However, there is an internal diode between drain and source inside the MOSFET. This would conduct. Assuming no fuses between the battery and regulator, a rather high current will flow until one of the diode junctions is blown open. So, expect to replace the cutout diode and MOSFET if this has happened.
Another scenario is if the generator should be 'flashed' after this regulator has been installed. If the MOSFET should be conducting when the battery and generator terminals are bridged out, the MOSFET will be blown apart internally.
When installed correctly, neither a short circuit to earth at either the generator or battery terminal will harm the regulator.

Positive Earth.
There is a positive earth version of this regulator for Model A owners. As I have been asked about this, I am presenting a theoretical circuit of what needs to be done to change the negative earth version to positive earth.

The circuit is essentially flipped upside down in the electrical sense.  Essentially, the cutout diode and 2200uF electro need to be reversed, as well as the supply connections to the PCB.
Note that I have not tried this so cannot confirm if it works!

IC not SG3524.
Subsequent experimentation to reconstruct the regulator on a breadboard revealed that unfortunately the IC is not an SG3524. Trying an SG3524 in the actual FP voltage regulator was not successful either. The problem is largely to do with the error amplifier. The unknown IC has about 3V at pin 9 (the error amp output), whereas there is nothing with the SG3524. Aside from that, it was found that the internal oscillator operated with a level of 3V p-p at pin 7 with the unknown IC, and oscillation started when pin 15 reached 3V. The SG3524 required pin 15 to be at 5.4V for the oscillator to start, and the oscillation level was only 2V p-p. There was no gate drive to the MOSFET unless pin 9 was taken slightly high with a resistor, but there was no regulation. The crowbar circuit does work with the SG3524.
Despite these differences, the IC still appears to be of the SG3524 family. There are a few other pin compatible types such as the LM3524D and NTE1720 but these have not been tested.
Update Feb. 2019: A reader has drawn my attention to the LM3524D in that the common mode voltage rainge of the error amplifier has been raised to 5.5V to eliminate the need for a resistive divider for the 5V reference. Interestingly, the Fun Projects circuit does not have the resistive divider that appears in the application notes.
It also appears that the Fun Projects regulator is in short supply, if recent comments in the MTFCA forum are to be taken note of. Also, information from another source informed me that Fun Projects was sold to its new owner knowing that the IC was no longer in production.
The LM3524D fits this description. I was keen to obtain one for testing, but turns out it is obsolete.
Update May 2019: Another reader has confirmed that the SG3524 works as a replacement, after a generator fault destroyed the original, by including the resisitve divider at the error amp input, and readjusting the divider at pin 2. This was for the 12V version of the regulator.

An experimental comparator type regulator.

Prior to examining the FP VR, I had assumed it used a comparator instead of a PWM controller. In this regard it would be like a regulator in a modern car in that it switches of the generator when the battery reaches full charge, and then switches it back on as the voltage drops. This switching frequency and duty cycle would be dependent on the load; the off time being quite long if the battery was charged and lightly loaded.
Experiments were made to see if my initial assumption was plausible, and the circuit above was constructed and tested on the bench. Like the FP VR, this circuit could also be built into the body of a mechanical cutout and be used as a drop in replacement.

It uses the "grounding switch" method for regulation. A MOSFET was also chosen as the switching device because of the low "on" resistance (.013R) meaning that little heatsinking is required, and the high resistance gate is easy to drive. A Schottky diode was chosen for the cut out also due to the low voltage drop and thus minimal heat generated.
The heart of the circuit is an LM393 comparator. The non-inverting input is held at a constant 2V by a red LED which is powered from a 470R resistor from the battery. Also across the battery is an adjustable voltage divider, comprising of a 10K pot and 560R resistor. When the voltage at the pot wiper (inverting input) exceeds 2V, the output at pin 1 goes high. The output drives the MOSFET gate high which then shorts out the generator. It can be seen that if the voltage divider is set so that pin 1 goes high when the battery reaches 7V, the battery will never be over charged.
Pin 1 is connected to the battery supply via a 2.2K resistor because the LM393 has an open collector output. Because of this, an emitter follower using a BC548 is used to drive the MOSFET gate. This is because the MOSFET must have a gate pull-down resistor to ensure it switches off completely when the LM393 output is low. If it wasn't for the emitter follower, gate voltage would be reduced because of the voltage division that would occur, and the MOSFET might not switch fully on. If this happens, it will overheat because of operating in its linear mode.

The switching would be unstable if not for one other component, the 15K resistor. This introduces hysteresis so that the turn off voltage is slightly less than the turn on voltage. When the output of the comparator goes high, the positive input of the comparator is raised slightly by means of the extra current flowing through the 560R. This means the turn off voltage must be made somewhat less than 7V (e.g. 6.5V). This provides a definite and clean switching point. Without it, the circuit oscillates at a high frequency around the comparator triggering point.

At this point the circuit is quite practical, but as the voltage divider, LED, and LM393 draw current, there would always be a drain on the battery. Unlike the FP design, the regulator circuitry operates from the car battery, not from the generator directly. If the car was used every few days this would not be a problem, but we cannot count on this. And besides, the FP VR does not have a drain on the battery when the car is not in use. Given that my design was to be at least as good, then the battery drain must be eliminated. This is easily done with a second BC548. It will be seen that the earth return of the regulating circuitry is via this transistor.
When the generator is rotating, the BC548 is biassed on by the 10K base resistor connected to the generator output, and thus the earth circuit is completed. When the generator is stopped, the BC548 appears as an open circuit and there is no drain from the battery.
The 47K base resistor ensures the transistor is really off when it should be, although is true the 10K and low resistance of the generator would do the same thing. The 100uF filters out generator noise that might interfere with the BC548 switching.

The circuit was bench tested and provided excellent results. With a 12V 5W lamp instead of the 6V battery, the circuit switched at about 100Hz. In this situation, the 2200uF simulated the battery time constant. No doubt with a proper battery the switching would be slower. The generator was simulated by a 10A variable power supply, fed into the regulator via a 12V 36W bulb for current limiting, and the secondary winding of a 14V transformer to simulate generator inductance. Over voltage protection is inherent in this design, so that nothing is damaged if the battery is disconnected with the generator running. Output remained at 6.8V with the lamp disconnected and the input increased to 16V.

I have not tested this in the car, but the circuit is presented here for those that want to experiment.