MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

Module power entry: Filtering? Reservoir capacitors?
MUFF WIGGLER Forum Index -> Music Tech DIY  
Author Module power entry: Filtering? Reservoir capacitors?
jakobprogsch
I'm self educating my way through various aspects of power for modulars, including the ones on the module side.

Most modules have some form of polarity protection (series diodes or parallel diodes+ptcs) and a 10-47uF capacitor at the input. Some also add a resistor or ferrite bead forming an explicit lowpass filter. But if you do the math these measures only really filter significantly above audio frequencies and even above switching frequencies of SMPS. I feel I don't understand the purpose of these capacitors/filters. I haven't really found much discussion about this other than one or two offhand remarks in PSU related threads.

Putting RC or LC filters on the module that cover most of the audio range doesn't seem practical because it would require unreasonably sized components. So I was wondering if maybe a capacitance multiplier type circuit would work there? Those provide attenuation of ripple down to 100Hz and less even with sensibly sized components. I did some SPICEing with BC337, 1kOhm and 10-47uF which attenuates 100Hz by about 30dB. As a convenient side effect the pass transistors act as polarity protection diodes. This however doesn't isolate the power rails against nasty load variations inside the module. But then, I'm in control what happens on the module and should design it in a "non nasty" way.

Does any of this make sense? Am I missing something obvious here or am I misunderstanding the purpose of these capacitors/filters?
555x555
I did an analysis of this a while ago [1]. Here's a quick run-down.

These capacitors *can* be thought of as a way to filter out power supply and bus noise, but it's maybe better not to think of them that way. The problem is that if you think about it as a filter, you only tend to think unidirectionally, instead of bidirectionally. Hence the term "decoupling" capacitors. The idea is that they can handle changing current requirements at a frequency that the bus can't easily handle. So quick drops in bus voltage won't propagate as much across the capacitor to the module, and quick increases in the module's current needs won't propagate as much to the bus. But don't totally throw the filter idea out either; it's not like you can use the word "decoupling" and then a resistor and capacitor suddenly stop behaving like an RC filter.

The idea is that, if we imagine a perfect power supply, that perfectly regulated voltage still has to get around to each module, and it can only do so via the bus and the power connectors. But the bus/connectors can't have nonzero resistance. What's more important, the bus/connectors can't have nonzero *inductance*. These terms get translated into voltage differences by V = IZ. So even if your power supply would regulate a perfect +12V at any frequency, this means that as the frequency of a current draw goes up, the inductive reactance also goes up and at high enough frequencies your bus is essentially unable to deliver power. But you can compensate for the inductive part of the bus by adding parallel capacitance closer to the place where the power is consumed differentially in higher frequencies. At low frequencies, the capacitors don't do much, but also the bus is very efficient. At high frequencies, the bus stops being efficient, but the capacitors become efficient. This is the same principle, btw, for the "bus" made up of a trace on the board between the power inlet and an IC's power pins. Usually with a 10nF decoupling capacitor, though, because this bus, since it is shorter, has way less inductance and resistance.

Now, about noise/ripple, component values and frequencies, essentially, you're correct. The 10uF capacitors do next to nothing to stop bus noise from propagating in the audio range. But also, this audio range noise is not so bad, because so long as your bus's resistance is low, it shouldn't create much of a voltage difference. That is, there shouldn't be much audio range noise on the bus in the first place, and if there is, the problem is probably better fixed either at the power supply, at the bus, or in whatever module is ridiculously current hungry. But synthesizers are extremely bad at being properly bandlimited. A true sawtooth wave oscillator has energy up to infinity, and increasing the slew rate of the theoretically vertical line is gonna be important for getting the oscillator to track across many octaves. When this high frequency noise hits various ICs in other modules, it interacts and ends up creating noise in the audio spectrum. This is mostly the noise you're gonna be filtering out at the module inlet.

Perhaps counterintuitively, you're mostly going to keep that noise from going *to* the bus, rather than coming *from* it. And you do this by *raising* the impedance of the bus as seen from your module. This means your module is more reliant on the capacitors to provide its transient current needs, but also less reliant on the bus to deal properly with all the transient current it would otherwise dump there. You should do this if your module has some near vertical lines of current draw, whether from digital, from a saw-core oscillator, from a comparator that's drawing a lot of current, etc. Since you're dealing mostly with high-frequency noise, you should add impedance in the form of inductance or a ferrite bead, with the bead generally a better choice unless you switch a *lot* of high-frequency current. The resistor makes very little sense IMO, but it's traditional, cheap, makes a good fuse, and is usually not a real problem. The even better way to do this, though, is to set up your module so that high-frequency high-current is shunted and the overall current draw remains at DC.

The capacitance multiplier is an interesting idea, but I'd first want to think through how it decouples the bus from your module, and not just vice versa. I'll let someone else weigh in on that one.

[1] https://www.muffwiggler.com/forum/viewtopic.php?t=185331
EATyourGUITAR
555x555 is %100 correct. modules polluting the power bus is really the biggest problem.

if you have to build modules that have good noise floor while connected to a noisy power supply then that should really be fixed with a better power supply. a lot of these problems are compounded by flying bus ribbon cables on a happy ending kit. or daisy chained bus boards coming from a switching supply. torroidal transformers with linear regulators are the best power supplies you can get.

I like the results I get from 10nF + 10uF. I always size the resistor to match the current draw of the module and what I would consider an acceptable voltage drop across my RC filter resistor. if my circuit takes 10ma max I will use 220R or 330R. if my circuit uses 100ma max I would go down to 47R or maybe 22R. I always use 250mw resistors. this is where magic smoke comes from. I don't care about fire protection as much as I care about a early warning smoke signal that tells me I have installed something incorrectly or maybe a blown mosfet something like that. it really only matters that it has a "fuse" the first time I plug it in to test what I have built. it lowers noise to my end users. that is what really matters more than the fuse. personally I never spend the money on ferrites. the idea that these RC filters do nothing in the audio range is completely false. the corner frequency could be way up high but I can hear more noise when I cut the cap out with the device under test. this depends on so many factors that I can tell you that my results are anecdotal and your mileage may vary. if money and pcb footprint is available then you should definitely use ferrites. I would never spec $10 in giant caps to filter noise from a cheap power supply. I went to 10nF SMD 1206 on the last commercial product I designed. you can also put more of them very close to the offending IC on the PCB. too much capacitance in a modular synthesizer in aggregate will eventually cause you to blow a fuse or simply not turn on due to the huge in rush of current when power is applied. the only acceptable supper complicated solution I approve of is a DC to DC converter feeding a linear regulator and ONLY IF the the instrumentation is extremely sensitive to power noise.
555x555
Two more things, briefly. The extra impedance you add before the capacitor filters noise (uncorrelated, unwanted signal) from the power supply, but it also "filters" the power supply's ability to give power to your module. This results in a voltage variation that's caused by your own module's current draw, but since this is correlated with the signal, it manifests not as noise, but as distortion. So you don't really want to increase the impedance to the bus, unless you have to, and unless your capacitance is ready to make up the difference in the supplied power. This is at high frequencies. The reason that ferrites tend to be used rather than inductors is that they have a low impedance at low frequencies, while practical inductors contain a good deal of resistance. Again, all of this says: don't use a resistor. But of course people do and they're happy so seriously, i just don't get it. I mean, they do make nice, cheap, smoky fuses. This also tells us that more often than not this stuff is not all that important.

The other thing to say is that if you *do* have to deal with a noisy bus, a better way than filtering out power supply noise is to just increase the power supply rejection ratio (PSRR). For example, are you taking a voltage from a rail plus a resistor and a potentiometer? That's a simple circuit but it's going to couple noise into that voltage with no rejection at all. If you instead add a zener and a capacitor as a voltage reference, you just got rid of most of that power supply noise. If you use a real voltage reference or regulator, even better.
jakobprogsch
Thanks for the detailed answers. I'll have to take some time and look into all of these new pointers and do experiments. I guess I'll have to build some badly behaved "modules" for some torture testing.

Quote:
if my circuit takes 10ma max I will use 220R or 330R. if my circuit uses 100ma max I would go down to 47R or maybe 22R.

wouldn't that cause voltage drops in the range of 2V to 4.7V? That seems a bit high?
555x555
jakobprogsch wrote:
I'll have to take some time and look into all of these new pointers and do experiments. I guess I'll have to build some badly behaved "modules" for some torture testing.


If you do, please share the results. There are a few people on here who have shared measurements of various bus designs and power supply designs, but if anybody has done detailed measurements of various module power inlet configurations along with badly behaved modules and such, I'm not aware of them. I did it in simulation, but stopped once I'd satisfied myself with some basic principles. It'd be nice to have real numbers.
EATyourGUITAR
jakobprogsch wrote:
Thanks for the detailed answers. I'll have to take some time and look into all of these new pointers and do experiments. I guess I'll have to build some badly behaved "modules" for some torture testing.

Quote:
if my circuit takes 10ma max I will use 220R or 330R. if my circuit uses 100ma max I would go down to 47R or maybe 22R.

wouldn't that cause voltage drops in the range of 2V to 4.7V? That seems a bit high?


I never actually did the calculation till 2 minutes ago. yes, 100ma at 22R on the +12v rail is a 2.2v drop. this means my circuit will have rails of +9.8v,-9.8v,0v. for a TL072 DC mixer it will not crash the rails. for a digital module that has a 12v to 5v linear regulator, it will not matter at all. that is actually more drop than I would like but in reality, the TL072 circuits are all < 10ma even if there are 20 opamps on one PCB. those numbers are max current draw per rail. quotient current is always less.
https://www.maximintegrated.com/en/glossary/definitions.mvp/term/Quies cent/gpk/1009
555x555
I don’t want to be too critical as I’m sure you’re getting satisfactory results or you’d be doing something else—and that’s all that matters. But as far as these numbers go... If your quescient current is 10mA and your peak current is 100mA, that’s 90mA of current you have to pull from a cap or through the resistor. With 22 ohms, that means you’ll have a potential 2VAC “signal” on the rails that the capacitor will have to filter out or will show up as distortion or bleed within the module. Because it’s resistance and not reactance, this means that even at 20Hz that value is the same. And the crossover for 22 ohms/10u is 720Hz, which means the cap isn’t going to provide much current to your module under that value. Likely, there’s more PSRR in the lower frequencies, but also likely that some of it is frequency independent. So the question isn’t really whether you can afford 2.2V less of DC—I’m sure you could—but what’s the differential current draw and how much self-noise at the mid and lower audio frequencies you can handle. Like I said, if this works it works so I won’t criticize the results, but we would have to find some other explanation of why it’s fine. Maybe you’re not getting 10mA quiescent and 100mA max, but more like 20mA max, and then it’s just 0.2 (which is still way more than the 50mV ripple of a relatively mediocre power supply). Maybe the PSRR is just that good. Maybe whatever distortion/bleed is happening is pleasing. Etc.
jakobprogsch
555x555 wrote:

If you do, please share the results...


I have no results yet but I worked on that unruly test module. Behold the (probably horribly amateurish) "rippleable" constant current load:





It can be adjusted up to 2A and has a primitive oscillator that can produce sawtooth shaped load variations between ~50Hz and ~4000Hz with adjustable amplitude. I'll see if I get some experimentation in this weekend.

EDIT: The above schematic has issues with the opamp+mosfet part oscillating so I replaced the mosfet with a big darlington NPN and increased R9 to 10kOhm.
555x555
Awesome! 2A!! That should be...plenty Dead Banana.

So...one of the reasons this is tricky to test is it's a whole system...so power supply, distribution, etc. all are going to give different answers to what the best module inlet configuration is going to be. You should be able to test your power supply's behavior directly and account for it in the results. The distribution is a bit harder. It might be good to test worst case---flying bus---and best case. For the latter, I actually have a few extra low-impedance four layer bus boards I made a run of. They have some problems... but should be fine for something like this. I'll PM you.
555x555
Damn, I just checked and I think I got rid of them. Nevermind. Anyway, general agreement is that flying bus is the worst, and for the "best" there are two contenders: aluminum busbars, or a low-impedance capacitor + multilayer PCB solution like the genus modu. There's also some level of disagreement about exactly how much practical difference there is between these best solutions and just a generic pcb busboard.
EATyourGUITAR
You changed my mind. I'm going to eliminate the resistors on all modules with the exception of clocks, LED drivers. That shit needs to have some DC resistance. My next case will have a 15v dual supply outside of the case and 12 linear regulators inside the case. I can use switching supplies for the dual 15vdc. This should keep the spring reverb quiet.
EATyourGUITAR
555x555 wrote:
Damn, I just checked and I think I got rid of them. Nevermind. Anyway, general agreement is that flying bus is the worst, and for the "best" there are two contenders: aluminum busbars, or a low-impedance capacitor + multilayer PCB solution like the genus modu. There's also some level of disagreement about exactly how much practical difference there is between these best solutions and just a generic pcb busboard.


Low ESR electrolytic capacitors if you want the best.
555x555
Glad to hear it ;-) And yeah, ideally for quickly switched LEDs I'd use an inductor, but a resistor is a lot cheaper and should work just as well. There's also the trick of using a constant current source and a transistor to route the current around an LED.

About low-esr caps...I'm actually a bit confused about this. Looking at Nichicon's site, there's "low impedance" and "low ESR." Judging from the tan angles, the ESR on the low impedance caps is not particularly low. So these must be...low inductance? Is this maybe more important than low esr? And anyway, esr is not actually measured at dc, so seriously, i just don't get it
Mungo
555x555 wrote:
So...one of the reasons this is tricky to test is it's a whole system...so power supply, distribution, etc. all are going to give different answers to what the best module inlet configuration is going to be. You should be able to test your power supply's behavior directly and account for it in the results. The distribution is a bit harder. It might be good to test worst case---flying bus---and best case. For the latter, I actually have a few extra low-impedance four layer bus boards I made a run of. They have some problems... but should be fine for something like this. I'll PM you.
Thinking about the entire system is the best approach.

555x555 wrote:
Damn, I just checked and I think I got rid of them. Nevermind. Anyway, general agreement is that flying bus is the worst, and for the "best" there are two contenders: aluminum busbars, or a low-impedance capacitor + multilayer PCB solution like the genus modu. There's also some level of disagreement about exactly how much practical difference there is between these best solutions and just a generic pcb busboard.
To be able to measure the contribution of the distribution directly requires a specialist power supply, or multi-port measurements (the distribution system has many "ports" and the interesting one here is module header to module header). Even with an almost perfect supply its difficult to get below 10m Ohm of system impedance:
http://mungo.com.au/euroPowerZ.html
There are some other links there to what difference is possible when all this is considered and parts selected for best performance.
EATyourGUITAR
I googled around for the answer. it depends on the frequency of your switch mode power supply and or your load or appliance or synthesizer module. if your power supply has a 100KHz oscillator then you want low ESR. if you are using a linear supply then you can use low impedance maybe. I'm not an expert in this area. just repeating what I read online. I don't think the low ESR caps would be a bad choice for a linear supply but I do think that it would be more expensive therefor a waste of money. it would probably be good to look at the impedance in the 20KHz to 40KHz range. anything above this will be beyond human hearing anyway. this is pretty easy to test with a 40KHz clock feeding a switched dummy load.

you can probably get a 555 to oscillate in the 40KHz with the right resistors and caps. it doesn't need to be exactly 40KHz. in fact, you don't even need to add a transistor to a load resistor if your 555 is the kind that sucks a lot of power. TCL555 is low power. NE555 will put huge ripples into your power supply. small resistors to get the load bigger. small caps to get the frequency higher.

this is known as a 555 astable multivibrator. also known as a 555 clock or a 555 square wave oscillator.

https://www.electronics-tutorials.ws/waveforms/555_oscillator.html

you would test various capacitors by putting them on the + and - rails after the power supply, before the 555. you could make the test more extreme by putting 10R resistors in series to the leads of the power supply to simulate a longer bus board or ribbon cable flying bus. you would want to put an oscilloscope or AC voltmeter directly on the leads of the capacitor under test. if the low ESR shows less ripple compared to the low impedance cap, you will need to weight the financial cost to the functional benefit.
ricko
This is a really good thread. It would be great to have a summary thread. (But not by me, I know nothing real about power supplies.)

It strikes me that there are several things we want from a power distribution/module system:

1) It should not burst into flames or kill you.

2) It should not run out of juice or wobble

3) It should not introduce bad noise, hums or signals

4) It should not introduce frequency-dependent variations into modules

5) After these, it should be as efficient as it can be


Three small questions:

Really, should the primary way to address 1) be to use shrouded headers or other polarized (preferably earth-first) connectors that make it physically impossible to connect things the wrong way round in the heat of the moment? Putting in resistors to act as fuses should be Plan B, not Plan A, surely? Prevention is better than detection/protection. And why not just use fuses, when protection is important? (B.t.w. I use the Neutrix power connectors to my supplies, and always shrouded headers on the Euroracks. )


Second, for 3) could we say that while a flying bus cable is best avoided, it should be used to get a star configuration (i.e. attach it to a bus board as close as the entry point of the ground signal as possible,with as few modules as possible, and to keep badly interacting modules apart (such as two VCOs that want to sync with one another, or digital modules and analog modules) ?


Third, for the rest, should we be thinking about what is the right size (or range) of decoupling caps for the particular slew of each module more? If a module has a long slow DC draw (like an ADSR or super-slow LFO) or needs reference stability then some larger cap is appropriate (10 uF? 100uF??) but perhaps such a large cap serves no useful purpose on an audio oscillator (100 nF to 1 uF?) that needs faster power. And something with sharp edges needs a smaller cap again (100 pF to 10 nF?). ( And put in each if your module has each behaviour.)
Mungo
ricko wrote:
This is a really good thread. It would be great to have a summary thread. (But not by me, I know nothing real about power supplies.)
Please no, every time there gets a bit of good information collected people come in with uninformed points of view and claim the opposite. You've just done exactly that without realising.
ricko wrote:
Really, should the primary way to address 1) be to use shrouded headers or other polarized (preferably earth-first) connectors that make it physically impossible to connect things the wrong way round in the heat of the moment? Putting in resistors to act as fuses should be Plan B, not Plan A, surely? Prevention is better than detection/protection. And why not just use fuses, when protection is important? (B.t.w. I use the Neutrix power connectors to my supplies, and always shrouded headers on the Euroracks. )
There are a growing number of manufacturers using 10pin headers, which when not shrouded add many other possible ways to connect "wrong" and ability to reverse the polarity of +/- without any way for the power supply to detect and protect from that.

ricko wrote:
Second, for 3) could we say that while a flying bus cable is best avoided, it should be used to get a star configuration (i.e. attach it to a bus board as close as the entry point of the ground signal as possible,with as few modules as possible, and to keep badly interacting modules apart (such as two VCOs that want to sync with one another, or digital modules and analog modules) ?
Star connections applied religiously are usually the worst solution, they should only be used in very specific situations.

ricko wrote:
Third, for the rest, should we be thinking about what is the right size (or range) of decoupling caps for the particular slew of each module more? If a module has a long slow DC draw (like an ADSR or super-slow LFO) or needs reference stability then some larger cap is appropriate (10 uF? 100uF??) but perhaps such a large cap serves no useful purpose on an audio oscillator (100 nF to 1 uF?) that needs faster power. And something with sharp edges needs a smaller cap again (100 pF to 10 nF?). ( And put in each if your module has each behaviour.)
Capacitance is best when on the module, rather than trying through the power cables. As mentioned above if modules need a stable voltage they should have a regulator right where it is needed and not rely on the input voltages being exact or stable. But many common power supplies have problems starting into a large capacitance, so modules can't just add more capacitance or risk being "less compatible".
ricko
Quote:
There are a growing number of manufacturers using 10pin headers, which when not shrouded add many other possible ways to connect "wrong" and ability to reverse the polarity of +/- without any way for the power supply to detect and protect from that.

So that is a yes? At least use shrouded headers.

Or are you saying that because of slack manufacturers not using safe shrouded headers, the busbars etc themselves need to have the protection on them, not the modules?

Quote:
Star connections applied religiously are usually the worst solution, they should only be used in very specific situations.


Hmm. This is the opposite of everything I have read. Which is not to say it is wrong. What is the rationale?

Quote:
But many common power supplies have problems starting into a large capacitance, so modules can't just add more capacitance or risk being "less compatible".


So that is a yes: use capacitors on the module but avoid putting ones that are unncessarily large. (What I am thinking of is that I frequently see the willy nilly advice that all modules should have 1uF or even 10uF caps on each module, regardess of their specific purpose or requirements.)
Mungo
ricko wrote:
Mungo wrote:
Star connections applied religiously are usually the worst solution, they should only be used in very specific situations.


Hmm. This is the opposite of everything I have read. Which is not to say it is wrong. What is the rationale?
The goal in the power system in a eurorack case is to:
a) deliver +/-12V volts of power (and possibly 5V too)
b) keep the 0V as uniform and equal as possible across the entire system (possibly multiple cases)

Most of the noise and problems are from b). If you try and wire every single module with separate wiring to a "star" point then you will have a mess, and/or use thin wires to fit it all in. If you instead take the same amount of metal and wire everything together with the shortest connections possible you almost always end up with a better result, particularly for the modules which draw a lot of current and cause the problem in the first place. Take some measurements and build an electrical of your case to compare a busbar vs a star wiring.

Star is perfect if the modules never connect to each other. but the entire point of a modular is to interconnect them which through the unbalanced interconnection defeats the star immediately. Contrast euro rack with unbalanced cables to other modulars that use banana plugs and require the user to explicitly set out the 0V connection.
555x555
I knew I shouldn't have mentioned distribution, no matter how neutral I tried to stay, lol.

@ricko Part of the issue with summary threads is there are a ton of different considerations in all this, so there's no real summary possible and no single best practice, just the best practice for a purpose. Are you constructing a whole system of friendly interlocking modules? Are you making a module that fits in well with a badly constructed power system? Are you designing a power system that works well for badly constructed modules? In what *way* are these modules/is this power system badly constructed, and what compromises are you willing to make? etc. etc.

To do something really well, you need to understand the principles and make your choices accordingly. If you don't want to do that, or if you want to do that but haven't quite gotten there yet, most of the existing common design patterns will do well enough: 10 ohm resistor, bead, etc. Just don't start doing something weird like raising the resistance a ton or using a huge capacitor.

In a community of all kinds of skill levels, it's actually a pretty good idea to have some default design patterns, but expanding them into a more complex system of rules, instead of just getting enough understanding to be able to make the choice yourself, usually leads to worse, rather than better, practices.
jakobprogsch
I stay off Muff Wiggler for two days...

Anyway first I edited the post with my schematic above since there were some issues with the unwanted oscillations which lead me to replace the mosfet with a more appropriate BJT instead.

I'm aware that all of this (PSU, distribution, power entry...) is interconnected and that is exactly what I'm educating myself about by experimentation. I just created a thread about this specifically because I felt I could't figure out what exactly justified the approach/sizing of the common solutions implemented on modules.

Realistically I'm not the person to deliver a formal study of all of this and I was just going to test things I already have or consider using. I figured if I learn something from it I'd eventually document exactly what I (amateurishly) measured with what means. Then the internet is free to (mis)interpret it at will. I assume this will happen in a different thread since it would cover more area. The general gist is that I wanted to experiment with permutations of power supplies (I have some commercial ones and built a bunch of prototypes using different regulators), distribution (flying buscables, Busboards, intentionally shitty diy striboards and I also have some solid copper bars) and power entry as per this thread.

I wanted to do all of this for the reasons mentioned above about there being many diverging opinions etc. So I figured I should just build and measure this stuff myself and iterate towards a solution I like and more importantly understand intimately.
555x555
You may think you’re not the person to do a formal study, but you’re the one actually doing it, so you are! thumbs up This all sounds excellent and I eagerly look forward to hearing about your results.
Graham Hinton
ricko wrote:
Really, should the primary way to address 1) be to use shrouded headers or other polarized (preferably earth-first) connectors that make it physically impossible to connect things the wrong way round in the heat of the moment? Putting in resistors to act as fuses should be Plan B, not Plan A, surely? Prevention is better than detection/protection. And why not just use fuses, when protection is important? (B.t.w. I use the Neutrix power connectors to my supplies, and always shrouded headers on the Euroracks. )


The best way of addressing the problem is to use a proper polarised power connector and not a data connector.
Placing components in the dc power line will produce a voltage drop so you have to tailor the solution to the current draw of the module. Polyfuses are better than resistors as long as you factor in their resistance and warming characteristic.

Neutrik mains power connectors are unnecessary. A fused IEC inlet is more useful, double fused if you use Schuko mains plugs.

Quote:

Second, for 3) could we say that while a flying bus cable is best avoided, it should be used to get a star configuration (i.e. attach it to a bus board as close as the entry point of the ground signal as possible,with as few modules as possible, and to keep badly interacting modules apart (such as two VCOs that want to sync with one another, or digital modules and analog modules) ?


No, that's nonsense. "Flying buses" should be avoided. Period. Ribbon cable for power is best avoided too.
Talking about star grounds is meaningless without considering the resistances involved. A busbar is really a star because it has very low resistance compared to the cables on it, it's just physically a different shape.


Quote:

Third, for the rest, should we be thinking about what is the right size (or range) of decoupling caps for the particular slew of each module more? If a module has a long slow DC draw (like an ADSR or super-slow LFO) or needs reference stability then some larger cap is appropriate (10 uF? 100uF??) but perhaps such a large cap serves no useful purpose on an audio oscillator (100 nF to 1 uF?) that needs faster power. And something with sharp edges needs a smaller cap again (100 pF to 10 nF?).


This is the wrong way of looking at it. Reservoir (and decoupling) capacitors are not passing signal through them. They are local reservoirs of charge and supply current on demand and are recharged. The resistances of the pcb and inline components determine the charge/discharge rates. i = dq/dt.
MUFF WIGGLER Forum Index -> Music Tech DIY  
Page 1 of 1
Powered by phpBB © phpBB Group