This topic comes up for any modular owner at least once. For many, power woes pop up later requiring expensive replacements or a full redo of their case. But worst of all is nasty bleed, crosstalk, noise, and other non-musical interjections that interrupt the creative process along the way. Fortunately, all of this can be mitigated or completely avoided by following some basic knowledge and taking the time up-front to plan out your system.
Note: there's a bit of oversimplification here but this should help with intuition and keep the article as cut-n-dry as it needs to be. I want everybody to be able to understand this so they can select components wisely. No calculus, just some algebra.
Voltage is the potential difference between charges... it's the pressure of the water in the garden hose. It's measured in volts, (V).
Current is the flow of charged particles... it's the flow of the water in the garden hose.
Resistance & Impedance
These are not the same thing but of the same category. Resistance is the opposition of the flow of current. Higher resistance means less current will flow. Resistances are like kinks in the garden hose or a partially closed valve. The garden hose itself is somewhat resistive too. Resistance is measured in Ohms.
Impedance is similar to resistance but it is frequency dependent. For now, just think of impedance as a frequency dependent resistor. Maybe like a potentiometer whose rotation is proportional to frequency. An interesting aside is that it is very common in electronics and nature that impedance ('resistance') increases with frequency. This is good. If most things didn't operate like this, we'd have quite a chaotic and unstable world.
Ohm's law relates voltage, current, and resistance.
where E is voltage, measured in Volts, I is current, measured in Amps, and R is resistance, measured in Ohms.
Despite the simplicity of this expression, it is incredibly powerful and simple enough for intuitions to develop that describe the relationships between these quantities. Think about Ohm's Law a bit. What happens if you keep R fixed but increase I? What happens if R increases but keep I fixed? What happens if you fix E but increase R?
Power is the real stuff. It's what makes things do the things we more intuitively understand as humans. It's what makes your modules turn on. It's what makes them heat up or generate light. It's what makes a fan turn or your car move. Power is defined as:
Constant Voltage, Loading, and Current Draw
Nearly all power supplies (including the ones that power your synth) are constant voltage output. This means the PSU will try its best to keep the voltage constant (it will regulate it) to a fixed voltage regardless of the load. Key word: try.
What is a load? The load ends up being whatever you connect to the power supply's output. The PSU doesn't know specific details about the load(s), just what it 'sees' looking at its own output terminals. It's a pretty shallow individual. Loads can be amplifier circuits, LEDs, microprocessors, your body, a screwdriver, and so on. The load the PSU sees includes all of the electronics connected to its output terminals. Schematically speaking:
Now this bulk load presented to the PSU terminals is going to try and draw some current from the PSU based on it's equivalent resistance. The flow of current is denoted by the green arrows. This current, multiplied by the fixed, regulated, constant voltage the PSU is the output power the PSU must deliver. P = I * E. So a +12V rail with a load that wants to draw 2A from the PSU will require 24W output power from the supply. The equivalent DC resistance of the load would be R = 12V/2A = 6 Ohms. If you had a bigger load, its equivalent resistance would go down and draw more current. This equivalent resistance is dependent on how many modules you plug into the distribution system, which is connected to the PSU terminals.
The current rating of the supply is just a maximum current it can deliver. The actual current draw from the PSU will be determined by whatever the load is. Pretend all of your modules draw 100mA (that's 0.1A) on the +12V rail. 5 modules powered from the same supply will draw 500mA (P=6W). 20 of them will draw 2A (P=24W). If your PSU is rated at a maximum output current of 1A, let's say, you could only power 10 modules if each drew 100mA. Add more and the PSU will no longer be able to regulate that constant 12V output and it will begin to dip or start to do weird things. You want to stay away from operating PSUs at or above max rated output.
Turns out the load (the R in the above picture) is not a fixed resistance and is an impedance. Additionally, the power supply output is not a perfect, stable, noiseless DC voltage. In reality, it could look closer to this:
In this picture, you see a bunch of noise riding on what looks like a poor sinusoid (it has lots of harmonics). Note the timebase and the amplitude. This is 25mV peak-to-peak at a frequency of ~ 240kHz. Also note those huge transient spikes. This is the output of a SMPS. If you were too zoom out OR (!!!) measure this waveform incorrectly, it would appear as if it were a flat line or 'DC.' Regardless of how good a PSU is, there is always noise and ripple. The better the supply, the less of it there is. The PSU output will be some DC level (the +12V or -12V or whatever) with some sort of AC waveform riding ontop. The AC portion is this noise and ripple. We will elaborate on this further but for now realise that the PSU output is not perfect and it's not just a DC voltage.
Ground is one of the most misunderstood concepts in electronics. And I have some guesses why it is so. The term itself is a shortcut taught at an early stage in an engineer's education. Providing this simple shortcut, at this point, I feel is a big mistake. In circuit analysis courses, ground pops up (even when running a circuit off a battery or isolated PSU) and is used so folks don't have to keep drawing wires everywhere on their schematics. It also makes their math 'something minus 0' which gets everybody excited as one unknown drops out of a linear system of equations. Along the way, as the circuits get more complicated, ground seems to always remain this perfect conductor that is 0.000000000... volts. The reality is that it is not. In some cases, not even close.
I like to tell folks having trouble with the concept of ground to first forget about the term itself and all the associated things they think of. Some of this knowledge is likely incorrect and I won't know which parts are correct. So start fresh and we'll see if we can't do better:
We'll first stop calling things ground. I like this approach and was recommended by Hinton to make things even clearer. Let's call it '0V' just as you would call the 12 volts rails '12V.' After all, it is a nominal value just like the others. Will it be actual 0V? Probably not, but it has to be 0V for measurement purposes when we look at the +12V, +5V, and -12V rails.
Ground (0V from here on out unless we speak about true grounds) is just another conductor or set of conductors - maybe it's a wire, a PCB trace, a metal enclosure, or maybe it's the braided shield surrounding a patch cable. It's typically a return for a bunch of common currents from all different sources to return. You can't throw out a voltage and current over a conductor (the 'resistance') and not have a way for it to get back. You need at least 2 conductors to your load. Take your synth power, for example. You probably have a +12V rail, a -12V rail, a +5V rail, and 0V (it may be called 'ground'). Your power supply sends a DC voltage of +12V and supplies some current to your module. The supplied current has to get back to the PSU through the 0V line (else you'd have an open circuit and no current would flow). The same is true for the -12V rail. And same for the +5V rail, too. All of these rails provide the module with the required voltages and currents (currents are drawn by the module - the PSU doesn't 'force' current into the module) and they all are returned on the same 0V conductors.
In schematic form, instead of drawing an ideal, zero loss wire to return the 0V current back to the PSU, you would instead draw a resistor (you add inductors and capacitors, too, once you realise that whole impedance thing). The 0V path is never a perfect conductor. But at some point, it's good enough (which means it's losses are low enough).
0V returns are important in audio systems - particularly for an unbalanced system as are most synths - because they ultimately control how much noise will be injected into your audio. Better, lower resistance (or really, impedance) 0V returns make it harder for noise currents to enter your system. Why?
E = I * R
If your 0V conductor(s) have higher resistance (really it has a frequency dependent resistance or impedance), then E, the voltage drop is also higher.
We will now stop using R so much and start using the term impedance (Z), instead. Unfortunately, impedance is where things get hairy if we want to keep the math simple. Impedance is a complex term (complex meaning there are complex numbers involved) but we can just intuit it from here on out to avoid all of that. This Z term is like a frequency-dependent resistance. At some frequencies, the impedance will be higher and at other frequencies it will be lower. The reason is because any conductor not only has a fixed DC resistance which creates losses, but it also has capacitance to other conductors and inductance. This means the conductor - even a wire! - will have reactive components that all depend on the frequency of the stuff running through them.
I shouldn't go too much further. Just stick with frequency dependent resistance for now. If that's too simple for you or want to know more, just give a quick search on the web!
An Attempt to Explain Noise Currents without Fancy Math
The noise currents I'm talking about can be introduced into nearby conductors either directly or 'magically.' The magic induction of noise currents are caused electro-magnetic fields floating all around us that cannot be seen by the human eye (well, light is still an EM wave but you get what I mean). Put a conductor near them and they form a current inside that actual conductor! It's important to note that most sorts of noise currents are not musical. They may present themselves as white noise or hiss to your ear, or hum, or other oddities. That is because they are forming voltage drops that are proportional to the noise signal themselves!
A reason you may hear 60 cycle hum is because that strong field near the transformer is flying through the air around all of your patch cables (which have 0V returns that are the shields of the cable). This field induces some current to flow in that cable shield. If you have a current and a resistance (which all conductors still have), you will have a voltage drop. This is a voltage drop on what was supposed to be zero volts, right? This drop is not constant but follows the shape of the noise signal. This is why you can hear this hum frequency. You can also get hum from many other mechanisms; this is but one example.
So our 0V return on our patch cable is not steady. It has some additional signal on it caused by the introduction of additional currents. These are not the actual signal we are trying to pass through the cable. It is noise.
What about the signal we are trying to pass? For an unbalanced system (which most modular synths are), there are only two conductors. On your patch cables you have the center conductor, some plastic around it to insulate it (which you can't see), then the outer conductor which is the shield, then some more plastic to insulate it from you and other stuff (this outer insulation is the outside of the cable - the part you see). If you don't know what I'm talking about sacrifice a cheap or non-working patch cable and carefully dissect it. Anyways, that shield conductor is also the return path for the audio or CV signal you are passing through it. You sent some out and it has to come back. It comes back on that shield conductor. Ok simple enough.
But wait, we have that other signal riding in that shield as well. The noise signal. Because there are only two conductors and any signal needs a return conductor, both the noise signal and your audio signal are combined. There is no way to separate them. This is why balanced audio is the bees-knees (look it up with this knowledge here and you'll see how they solved this problem).
Ground loops are something that also pop up often. The term is silly to me and another generalisation that is, in itself, the cause of much confusion. Ignore it and instead sketch out the return paths for your signals... draw any wires connecting things together as impedances and you'll see what drops can be generated through a system. To me, this is the best way to understand how noise can enter a system and can be utilised at both a macro and micro level.
OK, that went on too long. Main thing is that Ohm's Law works. It works for simple DC values and it works for AC impedances, voltages, and currents. Currents across resistances or impedances form voltage drops. If the current is changing, the voltage drops will change too. If the impedances are changing, so will other values (remember the relationship?) Remember signals are signals. The inputs of other modules or recording interfaces don't care where they came from. If it's a changing voltage, that is the signal. If it has 16kHz whine on top of your beautifully crafted complex oscillator patch, it thinks it's supposed to be there - it gets recorded. Ok. I think we can move on.
Safety & Earth Ground
Ground can also be used for safety purposes. The third prong on your wall outlets is connected to a big spike that goes into the actual earth. This is the real zero volts or ground - and now you know why it's called ground. It's also why we've called 'ground returns' in circuits '0V' instead of 'ground' - they may not actually be tied to true ground. Panels and any metal that is exposed to a human that contain electronics considered hazardous or that could present a hazardous fault should be connected to safety ground. This is so that a failure can be 'caught' by some electronics and you don't get electrocuted.
Most isolated SMPS bricks are Class II devices so the earth grounding opportunity is inherently lost. While this is passable by code for safety reasons, there are issues it presents when used in an audio environment. The reason is that you eventually want to connect your gear to other, earthed gear. Class II bricks mean the stuff coming out of them is isolated. So where is zero volts? There is no true zero volts. Why? Because it has no absolute reference to it. There is no wire connecting to the rod jammed in the ground (or to other gear that is). The voltages at the output are only relative to it's own 0V line and that's it. It's like having a battery with the negative terminal being the 0V terminal. It's not in an absolute sense - that is, relative to earth ground - it's only zero relative to the positive terminal.
The brick's 0V is connected to all of the sleeves of the patch cables through the distribution system and the jacks. Now connect just one patch cable into a studio compressor or recording console (which is grounded to earth - or real zero volts) and now you have tied those two system grounds together... sounds great. Well, it would be if the tie point wasn't a tiny, lossy shield conductor in a small patch cable. Now these two systems are at the mercy of the losses of this tiny conductor to keep these two at the same 'zero' volts. If you assume the studio gear is properly grounded (it is probably is), then we can say that side of things really is 0V. But connect that through a tiny conductor with lots of loss, run any current through it and what's on the other side? Ohm's Law it up. Any Z of that cable will have the modular 'zero volts' at something 0V + Z*I. It will not be zero volts like your studio rack gear. Noise currents will develop a drop across that tie-in patch cable. You have to have an entirely isolated system or have the grounds between systems at the same potential to have a clean, noise-free system. Note that studio grounding is a whole 'nother can o worms to get into. All you need to remember is that if the grounds are at the same potential between two pieces of gear, there is no possibility for noise current to flow.
Some other interesting side effects of ungrounded equipment... ever touch a panel and it feels fuzzy? Well, likely it's not made of fur and you're probably not tripping - that is a field being generated on the panel caused from not being grounded. Turns out safe devices still allow some capacitive coupling and some maximum level of leakage current to flow (touch current).
Class II bricks can work and are obviously used all over the place. My point is that you need to understand where they can be problematic in a studio environment - especially on unbalanced gear.
Moving on to the actual goods!
Most power supplies fall into 3 categories:
1) Linear, 2) Switched-Mode (SMPS), and 3) Hybrid (SMPS with Linear post regulation).
1) Linear supplies have been around forever. They are simple circuits, don't operate at any high frequencies, and are very easy to design to get excellent specifications. The downsides are that linear supplies are heavy, large, and not all that efficient (which is why they are large and heavy in the first place).
2) SMPS are what you will see almost everywhere in the consumer electronics space (which is HUGE). They are cheap as all holy hell, very efficient, and very small/lightweight. Without a question, they are great for powering non-critical electronics. The SMPS, as it's name implies - switch. They switch very fast edge rate signals at very high frequencies. Switching fast rise-time signals very fast into inductors generates a TON of transient spikes that have to be filtered somehow. Much of the SMPS design is in reduction of this switching noise and associated transients. All you need to know is they can't remove them completely (see the scope screenshot above). A linear supply, on the other hand, doesn't switch and doesn't need to filter this out because it's not there to start. When it comes to audio or other applications where high performance (as it relates to noise, regulation, etc.) is required, they are very difficult to get right. To get them right makes them start to approach and exceed the costs of really good linear supplies. Maybe one day linear supplies will be a moot point, but now, a really good SMPS with very low noise is a very expensive proposition.
Note that not all bricks or wall warts are SMPS. Some are linear (though they often have pretty cheap regulators and filtering). Just wanted to make that clear.
3) Hybrid supplies are a combination of linear and SMPS technology. They attempt to take some of the benefits of 1) and couple them with the some of the cost, efficiency, and weight/size benefits of SMPS. This is all fine and dandy but it will still not outperform a good linear supply. Why? PSRR of linear regulators is essentially 0 at the switching speeds of most SMPS. You can look that up or just know that a lot of the SMPS switching noise can get through this linear post-regulation stage - and at most SMPS switching frequencies and the associated harmonics (oh, there be many!), linear post regulation is pretty much an open door.
On the main systems I use for recording music, I only use full linear power supplies. Why? Because I don't really have the time mess around and I get very upset and grumpy when things interrupt me while I'm trying to be creative - esp if you are recording. Recording a highly volatile instrument like a modular requires things to always work right to capture the moment. Linear supplies are high performance and reliable. They're also easily repairable. That being said, there are tons of SMPS and hybrid switcher/linear supplies that will work. But they cannot touch the performance of a linear supply. So it depends on what you want and what you'd like to deal with. Keep in mind, you can also find poor linear PSUs. But if an engineer can't design a good linear PSU, then they surely cannot design a good SMPS which are far more complex. It's up to you to looks at the specs of each.
As mentioned earlier, the downsides of linear supplies are their size and weight. There's no getting around the dissipation of linear pass regulation, weight of large VA transformers, and BFCs.
To get around the portability issue, you can use a separate enclosure that houses all of the PSUs and put low-loss distribution in the synth cases. Sense lines are also very helpful in this scenario.
In a nutshell, if you don't want to screw around, just go full linear. I use Power One supplies. I do some basic modifications to increase reliability and precision but even the stock models are very nice power supplies. Hinton also makes a linear PSU more directly suited for synth use that many folks rave about although I do not have direct experience with them. I do know he is an experienced designer.
SMPS should be selected only when they have to be. The only benefit I see of a SMPS is portability. Need to put your modular in your overhead? That's going to be very hard to do with linear-based systems. So it's important to really spend time figuring out what is OK for you and how you use your modular.
NOTE: DIY power supplies are a BAD IDEA if you are not comfortable and knowledgable about electronics and the associated safety concerns.
Power supply specs are notoriously measured incorrectly. Some of this is due to lack of proper equipment or knowledge and other times the specs are fudged on purpose a to make them look better on paper. Sure, XYZ PSU can supply 5A at +/-12V! I would immediately ask for how long, how stable, and how noisy. It means nothing without other specifications and still means nothing if those specifications are measured incorrectly. Without proper measurements, you are left to trial-and-error and this can be an expensive and risky proposition. Many folks have gone down this road and have ended up disappointed.
The main specifications for a PSU are ripple & noise and load regulation (besides the obvious voltage/current ratings).
Power supply ripple is the periodic fluctuation of the 'DC' voltage output of a PSU. It's riding on top of the DC level.
Load regulation is how well the power supply maintains its rated output voltage under various conditions.
An example: Load regulation of 0.05% (very good!) for a 50% change in load. This means that if you change the load by half, let's say, worst case you would expect 12V rail to change 0.006V or 6mV. Very, very good! The worst regulators are in the many percents. 5% would mean the 12V rail could be as high as 12.6V and as low as 11.4V and you wont know what it'll be day-to-day.
The reason you want good load regulation is because many modules' 'references' are based on these voltages directly (not the best way to design a module, but it is what it is). Things like all of your tuning controls are usually referenced across the +/-12V rails. If those rails change their value, you may find yourself retuning your modules or, in some cases, recalibrating them! This is more relevant today where manufacturers are designing more and more digital modules and not giving the option to power those sections with a 5V rail... you get lots of heavy, noisy current draw from the 12V line and virtually none on the -12V. If modules all drew the same balanced current from each rail, then the drop would be the same on each rail and calibration settings based across these rails wouldn't care. If you are a manufacturer reading this, please spend the extra 80c and put the option of powering your onboard digital system from the 5V. If for nothing else, there isn't amps of wasted capacity on the negative rail.
Main point: if load regulation isn't good enough, you're gonna have a bad time, OK?
Distribution (the conductors that get from your power supply's output terminals to the module power cables) is very much overlooked in the modular world. Eurorack systems really push how poor a distribution system can go due the cost-driven nature of the format. Hey, I like when stuff is cheap just like anyone else... it means more modules! But there is a limit to all of this if you want a solid system. The distribution is, in many cases, more important than the power supplies themselves.
To start, let's use the water hose example again...
Let's say the power supply is a fire hydrant that outputs a constant pressure (voltage). You hook up a big fire hose to the hydrant. On the other end you are measuring the pressure (voltage). Open up that hydrant!
What happens? You get a huge powerful blast of water coming out. Lots of pressure at the output (it will shoot it far enough to put out apartment complexes on fire). But now say a bunch of people gather around the fire and accidentally start standing on the fire hose. It 'kinks' a little under each foot. These kinks are analogous to a more lossy distribution system. The more people standing on the hose, the lower the pressure (voltage) you will get at the end. If enough people stand on the hose, they will completely cut off the flow of water (current) and the pressure at the output (voltage) will be zero.
Even if the hydrant had perfect pressure regulation, it doesn't matter. It doesn't know you have a bunch of people standing on the hose leading to the destination where you, not the power supply, have expected a certain voltage. So it thinks it's doing a great job. So how's this relate to the synth and why do we care?
We should first describe two kinds of issues: DC voltage drop and 'noise' (which are really frequency-dependent voltage drops).
DC Voltage Drop
Since we know E = I * R, plugging another module on the same distribution board will create a larger voltage drop (for fixed R, if I increases, so will E). If you've already tuned certain modules, you may find you have to retune or recalibrate them - and all you did was plug another module into the distribution board. You can even move modules around and have them affect other modules if distribution isn't low enough loss.
Noise ('AC Voltage Drop')
Remember that long diatribe on grounding? Turns out that same 0V return is the return for any audio signals (in an unbalanced system, as are most modulars). Any differences in potential at the ends of the 0V return line will allow current to flow. This isn't a current you want... you didn't ask for it. It's a noise current. Its characteristics are dependent on many things... line noise, other audio signals, digital noise, any SMPS switching noise, etc...
Your distribution system's goal is to keep the 0V points all at the same exact potential across the entire distribution system. If they are not, you get voltage drop and noise currents flowing (which generate frequency dependent voltage drops).
Let's look at some examples of distribution schemes:
Ribbon-cable flying leads
Ribbon cables were never really meant for handling anything 'power.' That's another discussion. The real issue is that you are daisy-chaining all modules' 0V currents through very lossy 26 or 28AWG wires. How lossy?
6 conductors at 26AWG gives a DC voltage drop of 55mV at 3A (a lot but some digital modules these days are in the 100's of ma each!) across 400mm. I'm not including any connector contact resistance losses in these calculations... just the end to end loss. For comparison purposes...
This distribution scheme works so long as the modular system is very small...
Using PCB traces to distribute is better, but not by much. One benefit is that you can actually mount it and your power cables aren't flopping around. For comparison, our 3A current will generate about 35mV drop across a 400cm board with traces doubled up on both sides.
Rabid Elephant Low-Loss Distribution Board (LLDB)
We use 5x solid copper bars 0.062" thick and rugged Samtec gold-plated connectors. It's mounted to a PCB but the PCB is doing very little of the actual distribution - it's mostly for holding the connectors and busbars. Note how simple this system is. Copper busbars, a PCB, and some connectors. That's it. You don't need all sorts of other stuff on a distribution system. No band-aids, just solid performance.
The drop at 3A is a few mV.
This is our solution for getting a (very!) low loss distribution system in a small form factor:
The other thing to mention is that we've also decided to ditch the ribbon cables completely. We have our power cables custom made for us. This format change from ribbon cables doesn't mean it's incompatible with any existing euro module. The system is fully compatible. They come in two flavours:
RBE to Euro
This cable features the 'RBE' Samtec Rugged Mini Mate on one end and a regular, unkeyed 0.1" female header on the other. Each side features long-life gold-plated contacts (for more than 3x the life of a typical ribbon cable's contacts).
You use this cable in one of 2 situations:
1) to go from a normal Eurorack distribution 0.1" header to a Rabid Elephant module.
2) to go from the Rabid Elephant Low-Loss Distribution Board to a regular Eurorack module (with regular 0.1" header)
RBE to RBE
This cable features the RBE Samtec Rugged Mini Mate connectors on both ends. It is the preferred cable because now you have completely eliminated the possibility of plugging in a module incorrectly. Currently, this cable is only for plugging a RBE module into our LLDB. Maybe some other manufacturers will start using the same connector (it's an excellent connector - check it out!)
System Reliability & Longevity
All of the components of your power system contribute to how reliable it will be and how long it will last before you start seeing issues. There are many points of failure and I'll cover the main ones:
The PSU itself is the thing that is doing most of the work and has the highest parts count of the Power/Distribution system. You want to pick supplies that have long life! If they fail, bad things may happen like putting excessive voltage on ALL of your modules simultaneously! Look for PSUs that have been fully burned-in and tested. You also want to use PSUs that have been UL/CE listed since it means they have passed some base level tests. They are also deemed 'safe' by professional testing agencies so you can be a litte more assured that they wont burn down your house, injure you, or kill you.
Also, I've heard some people leave their systems powered-up at all times. Don't do that, please. Electronics life is mostly limited due to use (heat). If you keep them on, you're reducing their lifetime. Keep in mind all electronics can fail. Even the most stringent military or medical electronics cannot be guaranteed to work with 100% certainty. You probably have a system with 10+ different manufacturers parts installed... do you trust every single one of them not to cause any sort of problem? What about imported - no name - non inspected power bricks? Do you trust them to be plugged directly into a 15A service. Enough so to keep your house and loved ones (your cats) safe while you are away? Probably not.
Cables & Connectors
Ribbon cables and their associated IDC connectors are really only good for somewhere between 10 and 30 mating cycles before their specs (contact resistances, load ratings, etc) are no longer valid. That number also depends whether the platings on the contacts were done properly and in a certified, QC controlled environment which is probably not even the case if the connectors are purchased on the cheap from an unreputable supplier.
They are also not intended for handling often. The strain relief system is poor and many folks remove them by pulling against where the cable meets the connector since there's no other place to grab.
The connectors on your distribution boards are even more important because this is the part of the system that is 1) much more expensive and time consuming to replace and 2) it is common for all modules you'd ever want to install in your case.
You really want to think about how long you want to keep your modular instrument and how reliable you want it all along the way. The older things and the more use they get, the more you will expect failures.
Another benefit of using proper connectors (that are keyed) is that you completely eliminate the possibility of improperly connecting a module, which may damage that module irreparably.
There are lots of different power supplies and distributions systems available. Since the end-user more often than not is also the 'system integrator,' it's up to you to figure out what your goals are, how to implement them, and what hardware is required. If you want a large system or even a medium sized system with capability for lots of future growth, you want to take more care in selecting good power and good distribution. If you are just starting out and are unsure whether the modular is a good long-term venture for you, then maybe a low cost entry level system is appropriate.
It's important to note that for any of these discussions 'how much' is dependent on MANY variables like: PSU type and specs, load, module bleed into the rails, under-filtered modules, distribution DC and AC losses, power cables, power connectors, how much noise you can stand, and so on.
At the end of the day, better power and distribution does accomplish a few real things. It lowers the noise floor, reduces crosstalk/bleed, and it eliminates or reduces PSU noise from entering audio. It can also provide increased reliability and life. But you have to ask yourself how good is good enough. Is -40dB noise floor good enough? Are you willing to filter out any bleed or just get rid of sensitive modules? Do you want to calibrate your VCOs when you plug in another module or move things in your case around?
It's all a balance of pros/cons - write them down, weight them, and figure it out! Be sure to ask questions if anything isn't clear. I hope this article wasn't too much or too little. Have fun patching!