why do some appliance use only negative voltage regulators

[djacobow]

I've seen some washers that have stuff powered by a negative regulator rather than a positive one. I also have a simple 24Vac sprinkler controller that has only a LM337 to power its logic. Why would someone do that?

I understand the need for negative regulators in devices that need a bipolar supply. But if you only have one "pole" what does it matter which way it goes, except for turning your schematic upside down?

[Richard Crowley]

It is not nearly as common in modern equipment, but the original generation of transistors were only PNP, so a negative power rail (relative to ground) was not uncommon.  There is some vintage audio gear that is still highly prized by eccentric collectors that uses only negative power.

Also (in a previous era) the traditional telephone system was powered from -48V.  The explanation is that if the wiring is negative, it won't be subject to the electrolytic corrosion (in the earth) as if it were positive.  Perhaps that is the same reason for the sprinkler controller where the wiring (and the valve loads) are typically buried underground.

[T3sl4co1l]

Without knowing absolutely anything else about the applications, I would wildly guess:

- Sometimes, you just use a negative regulator for positive (you ground the "output"), because who knows.  Stocking, prices that day, who knows.
- Occasionally, for electrical characteristics.  LM317 is a 'high dropout' emitter follower configuation, but LM337 is necessarily a 'sort-of-low dropout' device, which also means it's prone to capacitance stability problems like most LDOs.
- Some circuitry work more logically with a negative supply and positive ground reference.  NPN ECL (the ancient MC10k series for instance) uses a -5.4V supply and 0V ground, because the signals (input and output) are referenced to the positive supply.  You can always run it from +5.4V, but all your signals are a few hundred mV difference, so it doesn't take much supply noise to cause problems.  In other words, PSRR is higher for the negative rail than the positive.  Much easier to have engineers design around that, than having to require extremely good supply planes and decoupling -- especially in those days (ECL was introduced in the 60s), when a lot of wiring was done with wire wrap for god's sake!
- Galvanic corrosion might be a motivation, which would be in common with the two potentially-wet applications.  Anodes corrode, so it's better to have larger things (like chassis or shielding) be sacrificial, or made of corrosion-resistant materials, than to require the circuit be made under the same limitations, or conformally coated, etc.

Tim

[SeanB]

Could also be so they can ground the supply lead, and using NPN transistors in open collector packages enables you to drive them easily. Then using the negative regulator makes for an easier power supply design, with a transformer that has a direct connection to ground ( with a split secondary and 2 diodes as rectifiers) for extra noise reduction.

[djacobow]

Well, in the sprinkler application, the circuit provides 24Vac to the solenoids in the ground, through ST Z01 triacs. So, I don't think galvanic protection of the sprinkler wiring is the driver. The negative voltage could maybe provide that for the circuity in the box.

I wonder if the triacs they chose can be triggered more easily or better, or something in Quadrants II or IV.

[richard.cs]

I wonder if the triacs they chose can be triggered more easily or better, or something in Quadrants II or IV.

That sounds very likely actually. I've used this chip for controlling triacs before: http://www.bucek.name/pdf/u217b.pdf it's not that obvious from the datasheet but it generates a -9V supply relative to mains neutral so that it can trigger the external triac in quadrants II and III. If the supply, and therefore the trigger output, was positive then it would trigger in quadrants I (good) and IV (bad, may not work at all). At first glance you look at the application circuit and think all the electrolytics are backwards but they're not.

[Seekonk]

It has been almost two years since I looked at it so I don't remember much.  I looked at converting a gas stove to propane at my 12V camp.  A tiny inverter for the burner ignitors was easy.  The controller for the oven and clock seemed upside down and it drove triacs.  50V for the fluorescent display seemed negative, not really sure.  It was all possible with conversion to FET and a small inverter.  What killed the project was the oven glow bar that consumed 400W all the time the flame was on.  That was more than the solar system could provide.  Will convert that to a pilot light.

[Richard Crowley]

It has been almost two years since I looked at it so I don't remember much.  I looked at converting a gas stove to propane at my 12V camp.  A tiny inverter for the burner ignitors was easy.  The controller for the oven and clock seemed upside down and it drove triacs.  50V for the fluorescent display seemed negative, not really sure.  It was all possible with conversion to FET and a small inverter.  What killed the project was the oven glow bar that consumed 400W all the time the flame was on.  That was more than the solar system could provide.  Will convert that to a pilot light.

What about an electronically-generated spark?  I have seen devices (like gas-fired stove/range/cooker) that use a spark (not unlike an automotive spark-plug) which repeatedly arc until the flame is established (or else shut off the fuel for safety).

[Seekonk]

They actually have an interesting circuit.  The glow bar is in series with the gas valve that either has a mechanical delay or the bar lowers resistance as it heats up.  There is a definite turn on delay.  So at the least I have to build an inverter to create low voltage/high current to operate the valve.  Current stove has just a pilot light and I now have the parts to recreate that setup.  My preference is that compared to continuous running of spark generator or sensing flame conduction (flames do conduct) and turning sparker off.

[David Hess]

- Occasionally, for electrical characteristics.  LM317 is a 'high dropout' emitter follower configuation, but LM337 is necessarily a 'sort-of-low dropout' device, which also means it's prone to capacitance stability problems like most LDOs.

I usually do it for cost reasons.  NPN and N-channel transistors can be used in low dropout negative regulators without extra complexity and for a given performance, they also cost less than their PNP and P-channel equivalents.  If the input is floating then the negative regulator can be used to generate a positive output.

You can see this reflected in old power supply designs where every output was generated from a floating transformer winding and every regulator, positive or negative, used an NPN power device.  Some even older designs did the same thing but used only PNP germanium power transistors.

I have occasionally also done this for easier frequency compensation using only only NPN or n-channel emitter/source followers on both the positive and negative output to take advantage of less expensive transistors.

[Paul Price]

If the device is using TRIACs for power/solenoid/motor control, a negative power supply for TRIAC triggering gives the best triggering with the least amount of trigger current, thus the reason for a design using a positive supply ground for the control circuits. The control circuit is almost identical in its design and use of parts to one using a negative ground, but just using the positive output of the power supply as the common circuit ground point and connected to one side of the 120VAC/240VAC line and TRIACs. This allows reliable triggering and simple trigger circuits for the control TRIACs without using TRIAC optocouiplers for triggering. Quite common in washing machines, for instance.

[djacobow]

OK, this is starting to make sense to me now; the other place I had seen this was in a washer. I think it is about driving the triacs without optos.

Which begs the question, if you use a negative supply, can you hook up a "regular" uC, just with the rails reversed? The uC's Vcc attached to ground and the uC's ground attached to -3.3V or whatever? I'm assuming that the uC would not know the difference and would be perfectly happy -- you just have to be careful when you try to attach your debugger. :-)

I'm interested in projects that have triacs controlling low voltage AC and trying to design out the opto. I've done some experiments, and even with a positive power supply it seems to work OK. The only thing is that the power supply for the digital logic controlling the gate needs to be referenced to one side of the AC waveform. This requires that the PSU only be half wave rectified, I think.

[T3sl4co1l]

Most TRIAC datasheets I've seen, quadrant IV operation is only double the bias required of the others, so it shouldn't be too terrifically hard.  That's static, so I don't know if dynamic (i.e., turn on speed) still sucks.

Standard CMOS MCUs will hardly care, because the PSRR to VCC is only slightly more than to GND, and both have reasonable noise margins.  Ideally, the threshold range is 30-70% of supply, i.e., symmetrical, but NMOS are about 2.5 times better performing than PMOS*, meaning either the NMOS transistor is physically smaller (it is -- normally CMOS is fabbed with a 1:2 area ratio) and has lower capacitance, or the same size (and capacitance) but much lower resistance (normally the NMOS is ~30% lower resistance, which, hey, that's almost exactly the remainder of 2.5 / 2 expressed as percent, so the system works).

So, as far as sensitivity to supply rails, GND is probably slightly more sensitive due to the lower resistance to it (when devices are active).  The threshold voltages should be matched as specified, so that's really the only difference there is to note.  CMOS is otherwise pretty symmetrical, as it aims to be.

*A handy rule of thumb, which is based directly upon the material properties of the semiconductor: namely, electron mobility is always higher than hole mobility.  They're still fairly close in silicon, but CMOS never worked out for GaAs because the hole mobility is something awful like 5 times lower than the electron mobility.  Which is, in turn, several times better than it is for silicon.  This is both reasons for why 1. GaAs generally goes much faster than Si, and 2. all the big logic circuits that were fabbed in GaAs (e.g., Cray-1) were designed with good old fashioned NMOS (N channel FETs plus dumb old pull-up resistors!).  Needless to say, all that logic simply gobbled up power!

Tim

[richard.cs]

Most TRIAC datasheets I've seen, quadrant IV operation is only double the bias required of the others, so it shouldn't be too terrifically hard.  That's static, so I don't know if dynamic (i.e., turn on speed) still sucks.

For applications with inductive loads there exist "snubberless triacs" which essentially don't trigger in quadrant IV at all and are more immune (apparently) to spurious re-triggering at turn-off.