Parallel Series stuff again ;/

I think this does almost what you want, for < $25.00 I haven't figured out, in this circuit, how to do series parallel where you have two series strings of 2 parts and you want to put the strings in parallel.

For four components, this circuit uses 4 relays (4PDT), and one switch for each relay. (daigramed below)

If you want to add phase reversal, you need 4 more relays (DPDT) and 4 switches. (diagramed below)

If you want to be able to turn components off, you need

4 more relays (and the component must be switched to parallel). These component on/off relays are not shown in the diagrams below - the relay points simply go in series with the component leads.

If you want to place 2 or 3 parts in series and put the combo in parallel with the other part(s), you need 1 more relay.

Parallel bus ----o------o------o------o-------- ==> output

Parallel bus +---o------o------o------o-------- | serial bus +---o o---o o---o o---o o--- etc ==> output 1 2 3 4 etc

A 4PDT relay for each part (component). When the relay is in the parallel position, the part's leads are connected to the parallel bus pair, and the contact is closed on the serial bus for that position. When in the serial position, the part's leads are connected to the contacts on the serial bus. Example:

Parallel --------+-------------------------------{{--- ==> out | o o----------------+ Ry1-1 ---part--- Ry1-2 | Serial +---+---o o---+---serial-- | ---{{--- ==> out | | | | | | | | | +---o o | | Ry1-3 ---+-------+--- Ry1-4 | | | Parallel +----------------------------------+---{{---

4PDT relay

Ry1-1 is connected to RY1-3 and to the bottom parallel bus rail. No other RYn-1 point is connected to the bottom parallel bus rail.

For phase reversal:

---+---+ +---+----- | | | | | o o | | ---Part--- | | o o | | | | | | +------------+ | | +------------+

For series parallel:

Parallelbus --------------+--------- ==>Output | | o----+ ParallelBus +------------ | --- | | o | Serial Bus +-------------- | o--------+ ==>output

Ry9

4 4PDT and 5 DPDT relays and 9 switches Item# DCTX-2460 $4.50 * 1 = 4.40 24V dc wall wart RLY-464 $1.20 * 9 = 10.80 24V DPDT relay RLY-466 $1.00 * 4 = 4.00 24V 4PDT relay SSW-37 $0.25 * 13 = 4.00 DPDT slide switch ===== Total $23.20

(Switches are 4 for $1.00)

The switches can be replaced by your automation circuitry. You mentioned that the switching voltages and currents are quite low, so you'll need a transistor driver for each relay when you automate. You'll need to figure out the logic of what relay needs to operate when and sort out the signals to drive the transistors. The pesky parallel of two series strings bugs me, but it's 2:30 A.M. and my brain is in shut down mode. If you figure that out, please post it!

Ed

the way I see it for each device a 4PDT and DPDT relay can be used together to selest forwards/reverse/isolated/shorted+isolated

or a DPDT and 2PST to select forwards/reverse/isolated (isolated is neither input connected to the device) or a DPDT amd SPST to select forwards/reverse/open (here open is one of the inputs not connected to the device) or SPDT and SPDT to select forwards/reverse/shorted (here shorted is both outputs connected to one input)

a further 4PDT can select series/parallel (or you can use a DPDT here if you don't mind the phase of one of the devices reversing when you switch from series to parallel)

anyway 5 relays can do the job. (and thus 5 input signals)

tyring to manufacture switches can be a really tricky...

Bye. Jasen

I think the main thing you guys are missing is that I need to build this for n components and not just for 2. You can see that even for n = 3 it is a much more complex circuit than for n = 2 since one has all kinds of series and parallel combinations. Since I will be using around n = 6 or 7 it would require many many relays(probably atleast 50+ if not more). The graphic was just an example for n = 1 and n = 2... just to get the concept across.

Although I might be able to accomplish this with the 4PDT relays since ehsjr has pointed out that they can be found for ~1$(everywhere I've looked they cost atleast 4$ ;/). The problem with the quad relay is that I need it to be latching and I'm not sure if its a good idea to have it where I control 4 switches with one input(what I mean is that it seems for relays that one always has to flip all the switches inside the relay together and one can't control each switch individually). This seems that it would require a lot of thought to all the actual combinations I would end up using(and for n = 6 it is huge) but I might be able to get it done just using 4PDT relays.

Another thing is I cannot use solid state or mechanical switches because I have to have no degredation of the signal(except during switching it is ok) and I cannot have them manually controlled(since that defeats the whole purpose of the project... I already have a "prototype" that uses dip switches to do most of the combinations).

So the reasons that I feel I would need to come up with my own method is simply that it seems to be cost effective(in the long run atleast) and allows the circuit to be easily built and directly to the point of what I'm trying to do... and ofcourse that I need a latching type of switch since I don't want to waste power to hold the state when I don't have to(hence my method can be used with no power except when one wants to switch.... Which isn't very often... could be hours or even days without needing to switch).

Ofcourse this isn't to say that it still can't be done with relays and also be cost effective(actually I think the relays tend to be quite expensive though compared to my switches since it involes just basicaly punching out the paths then stacking them together... except for the motion control which might make it more expensive(Still working on how to control the switch electromechanically)). I'll look into it some more and see what I can do...

Thanks for the help, AD

Can this get all those paths that I need(such as all the "poles" connected together) and if so can it be gotten for ~$1?

My idea was simple to somehow make the layers and then I could easily stack them and glue them together then I could worry about the contacts later(which I have an idea how to do but I'm not quite sure if it will be reliable).

One could make the conductive paths by punching them out on copper sheets(I'd have to make a punch which will be somewhat difficult to do to get it accurate I think, but if my switches are large then the accuracy decreases... once I get a prototype then ofcourse I can worry about how to make it smallter later).. I could use the etching method to generate a lot of the paths but I don't have a laster printer to do it... I'd still have to place the paths on the wafers with the proper rotation... I might be able to do this easy by including a little notch or something to align them up with the wafer.

Once I am able to make the actual switch(which I think would actually be pretty easy once I get a good method down) then I'd have to figure out how to electromechanically control it. I'm not sure how hard that will be but I have several ideas that may or may not work.

Ofcourse there will be many precision problems and I'll have to work them but end the end it might be worth it since it seems to be the only option I have that covers all my requirements(atleast at this point in time).

Thanks, AD

have you considered a 3 pole 7 way rotary switch ?

you could easily make the wafer part on a pcb, many multimeters have quite complex rotary switches of this type if you can see inside one youl get an idea of construction.

Colin =^.^=

What do you mean the use for the switching setup? It is simply to select one of the many possible series and parallel combination circuits for n components. (n being fixed ofcourse).

That is all you guys need to know as that is pretty much exactly the project... ofcourse there are more details but they will just get in the way and for some reason it already seems like a confusing topic... not sure why its so difficult to understand what I want to do.

To explain it again,

Given N electrical components how can one design a circuit that will allow you to realize the components in any wiring(one that make sense of course... any way to combine them in combinations of series and parallels(since thats the only two ways) to get a new circuit).

What I mean by a combination is simple a certain, but arbitrary, wiring out of one of the many possible ones. e.g., for N = 3 one can have the first component in series with the two who are in parallel and all in phase, i.e.

-->

--> --- ---- -->

or all can be in parallel

--
-->

You might want to look at the design of the mechanical switches they use to use at the telephone office. They both rotated and moved vertically to do the telephone circuit switching. If you describe the use for the switching setup somebody may already how it has been solved.

even your 28-way switch will be needed in large numbers to handle all series/parallel/polarity combinations for 7 items.

I don't see that your 28-way switch offers a great improvement over using relays, you may find that it's difficult to makke reliable contacts, and driving them isn't going to be simple.

they could easily cost more to manufacture than the equivalent circuit builtt using relays.

ISTM that your 28-way switch could be replaced by 6 SPST relays. and apropriate driving hardware.

by spaciing the contacts unevely it may be possible to use a simpler wiring pattern and fewer positions, there are only 15 ways the switch can connect the 4 terminals so gives a lower limit.

a solenoid operated ratchet is what they used in phone exchanges. it might be fast enough.

probably the simplest way to do it for N devices is using 2N N+1 throw switches (eg single pole slide or rotory switches) one on each end of each device, Unless you can find some old telephone exchange equipment (could be hard - most of this stuff was recycled 10-20 years ago) _and_ it proves to be suitable you'll need to to build these switches and the servos to operate it yourself.

| | -o- = connection -+- = no connection | | a b c d e f gh i j k l m n o

| | | | | | | | | | | | | | 1-o---+--+---+--+---+--+---+--+---+--+---+--+---+-- (+) immagine the o's | | | | | | | | | | | | | | are connections 2-+---o--o---+--+---+--+---+--+---+--+---+--+---+-- beween the lines | | | | | | | | | | | | | | and can be moved 3-+---+--+---o--o---+--+---+--+---+--+---+--+---+-- up and down the | | | | | | | | | | | | | | columns to alter 4-+---+--+---+--+---0--0---+--+---+--+---+--+---+-- the way the | | | | | | | | | | | | | | devices are 5-+---+--+---+--+---+--+---o--o---+--+---+--+---+-- interconnected | | | | | | | | | | | | | | 6-+---+--+---+--+---+--+---+--+---o--o---+--+---+-- | | | | | | | | | | | | | | 7-+---+--+---+--+---+--+---+--+---+--+---o--o---+-- | | | | | | | | | | | | | | 8-+---+--+---+--+---+--+---+--+---+--+---+--+---o-- (-) | | | | | | | | | | | | | | `' `' `' `' `' `' `'

to reverse the polarity of switch a code to position 2 and switch b to position 1 to put and in paralled j goes to 4 and i to 6

for all devices in parallel

a b c e d f g h i k l m n o 1 8 1 8 1 8 1 8 1 8 1 8 1 8 etc... with this setup the number of actuators needed is 'only' 2n... it is be possible to do it with some of switch positions missed I'm just not sure what the optimal reduction is. and I don't think there are many that can be ommitted. i think the maximum reduction is N(N-1)/2 ) switch positions If you need a way to islate the devices from the circuit without short-circuiting them add another row but don't wire it accross. the above (with 7 devices) could be done using 3 relays per column (SPST+DPDT+4PDT) for a total of 14 of each type, total 42. (you were't far out with your prediction of 50 relays) latching relays (which stay switched with no power applied) will cost extra, but should offer high reliability ... At retail it's probably about (ballpark) $150 worth of relays, wholesale or in bulk the price will drop. Bye. Jasen

yep, but they would probably be very easy to make and only a fraction of the number of relays that would be needed(atleast with my initial estimate).

Well, If say the relays cost 1$+ and I need 4* as many for for each switch and I can make the switches relatively easy(which I think I can else I wouldn't even bother going down that path) then I think it is a very good improvement. After all, I doubt I could make any decent relay that would do the job at half the cost I could buy them... but surely this switch has a chance. I have some ideas for the contacts though that will make them pretty reliable. The biggest issue is making them motorized and of alignment. I'm still working on those ideas to see what I can do but if I can overcome them then I think its viable... ultimately I think this switch handels what atleast 4+ SPST relays can do easily(and thats a low estimate). It maybe that one could wire some relay combinations up in an a certain way to get the same effect but I doubt it can be done about the same cost as using the rotary method.

I doubt it... except for the motorized part. Think about it a little and you will see just how easy and cheap it can be made... after all, I don't need a tank of a switch. I could be wrong though but I have a feeling it would be must cheaper than going the relay method. If one says that they need atleast 4SPST relays to accomplish this(which is the min I think in the general case) then thats 4$ for as compared to one switch which might cost

0.5-1$... or even 2$ if, say, the motorized part costs a lot. It could be that to get any reliablity would increase the cost though. I could be totally wrong but until I make my first switch(which I hope to get to work on soon) I will not know how difficult it will be.

Its possible, I personally can't see how but thats still approximately

6$(atleast) for compared to one switch which, I believe, shouldn't cost more than $2... Ofcourse I'm excluding the labor costs and some other possible issues but I just can't see how it could be more than a few dollars(except for the motorized issues... if that can be done cheap then I think the switch is a real possiblity).

Yeah, I thought about that too. I'm not sure what the optimal case is though. It was just the best thing I could come up with that easily solved the problem. I'm going to have to look at it in more detail though to see what if any optimizations can be made. I'm sure that there would be many times in the matrix where some switching paths will not be used. It might turn out that I would need far less number of relays that the rough estimate of n^2*m where m is the number required for one "node".

I have several ideas to electromechanically control the switch... none of which might work. Also, it isn't necessary that I use a rotary method but I could also use some type of push-pull method. The switch would be designed on the same basic principles though. I'm not sure if it would be easier or not to use that but I will probably do some tests to see.

Well, I feel that it would be a good learning experience. Maybe the outcome won't be to my liking but atleast I'll learn something.

I'm not sure I understand your matrix. if the +'s at the nodes are open connections then I can't see how we can get any circuit because there is always open nodes everywhere. Although I think I get your drift I'm just having some problem understanding your method ;) If it does work and gives all combinations I think it would be much easier to implement and cheaper and actually not require any relays. I would require some motorized slides and 2n total like you said though.

| | V

---------*------*---------

where

| | V

is just a level that will push on a movable wire like this:

| | V

---------* *--------- --------

so one could easily open the "line" by "sliding" to the right position.

I guess, not sure if that makes to much sense though and I'm not sure if it goes with what you mentioned above(and it might be backwhere where you want to make a connection instead of break it... but its pretty easy to fix.

I'll work on it some more. I've been a little under the weather lately and its been kinda hard to think ;/

Thanks for your help, AD

--
Really?

Do you have a solution which you can post for any number of circuit
elements?

I ment the straight lines to be continuous conductors that cross but don't contact except where indicated by the "o"

they don't have to be straight sliders they could be 8-way rotory switches

It might be possible to arrange so that two motors could control a number of these switches bay using rotora switches all on the same axle and the axle having a bulge with teeth that engage the centre of the switch to turn it.

it could be tricky switching more that four or five in less than 0.5 seconds though, so that might not be a good approach,

a single solenoid could be used to advance the switch around the 8 positions it'd need careful design to do it in 0.5 seconds but is certainly possible, (I ran an electromecanical counter at 3 times the required rate) an 'index' contact would need to be added so the system could reset and get the switches into a known configuration.

Bye. Jasen

If you used three dimensions instead of just two, you probably could get a lot more circuits and save some space too! ;)

Right. And it can get really ugly ... for example, say R1 has to be in parallel with C1 and C2, and that combo has to be in series with R2 - and that combo has to be in parallel with another similar setup of R3, C3,C4 and R4 and so forth. The problem is there are an ever increasing number of ways the components can be combined as their number increases - something involving N! springs to mind.

Originally, I thought he wanted this for 4 components. Turns out, I missed what he wanted by N-4 :-)

Ed

ok, I gotcha.

Can you explain to me how to get say

--------------G->--------------| | | | |---B->---| |---D->---| | | | | | | |

--- A-> -+-+ +------+ +-+- F-> --- | | | | |---C->---| |---E->---|

I can't see how to get that with just 'o' being making the horizontal conductor contact the vertical one(i.e. a 4way).

It seems to be that one can't get "sub-parallel" circuits from your matrix with just the 'o' I described? Am I missing something?

AD

If you mean arbitrary number then ofcourse not... as one doesn't exist. My method works for any arbitrary bux fixed number in such a way that it is very easy to do and is cheap(assuming one has the proper components). The relay method is also somewhat easy to do but can get expensive really quick(which my method might get expensive too if I can't make the components cheap enough). i.e. These methods are direct in the sense that they are straight forward and scale directly with an increase in the number of components. There might be other methods that uses some extravagant circuits to get the same result be are more complicated to build. I doubt those methods would reduce the cost a half an order of a magnitude than the easier methods though. (Ofcourse if they do then there extra complexity might be worth it).

Any easy method that requires N "devices" for N components is simply to use ADC's and DAC's. One can then simply create the desired combination through software. This is not an option though.

this bus though only allows for one parallel arrangement? If you look at the other post I mention the same problem. i.e., how does it deal with "sub-parallel" and "sub-series" circuits? that is, series and parallel circuits within series and parallel circuits?

Thanks, AD

yeah, well, my N is about 6 but can go from 4 to 8 for all pratical purposes. It would be nice if I only had to deal with N = 4 ;)

AD

I thought about that but my 3rd dimension was just stacking(or folding) the

2-D circuit to save space... I'm not sure if I could come up with a true 3D circuit that will do what I want (I guess its possible though). It would be hard to figure out how to make the switches too ;)

AD

a b c d e f gh i j k l m n o

| | | | | | | | | | | | | | 1-o---+--+---+--+---+--+---+--+---+--+---+--+---+-- (+) | | | | | | | | | | | | | | 2-+---o--o---+--o---+--+---+--+---+--+---+--o---+-- | | | | | | | | | | | | | | 3-+---+--+---o--+---o--o---+--o---+--+---+--+---+-- | | | | | | | | | | | | | | 4-+---+--+---+--+---+--+---o--+---o--0---+--+---0-- | | | | | | | | | | | | | | 5-+---+--+---+--+---+--+---+--+---+--+---+--+---+-- | | | | | | | | | | | | | | 6-+---+--+---+--+---+--+---+--+---+--+---+--+---+-- | | | | | | | | | | | | | | 7-+---+--+---+--+---+--+---+--+---+--+---+--+---+-- | | | | | | | | | | | | | | 8-+---+--+---+--+---+--+---+--+---+--+---0--+---+-- (-) | | | | | | | | | | | | | | `' `' `' `' `' `' `'

A from row 1 to row 2 B and C from row 2 to row 3 D and e from row 3 to row 4 F from row 2 to row 4 F from row 4 to row 8 The rows represent the nodes (junctions) in the diagram. 1 and 8 I have chose to be in power connections.

Bye. Jasen

Cool. I believe it will allow for all combinations? If so then it is much easier to implement then my original idea by far. I'm not 100% it works for all combinations but I do believe it can. The rows seem to "control" the "parallel'ness" and the columns the "serial'ness" I'm going to go ahead and try and prove it can get all the combinations and if I can do it then I'll probably use your idea as it seems much much easier in the sense that I don't really have to manufacture much as compared to having to make my own switches(which I might do for fun one of these days but I'd rather get my project done).

Thanks for the help, AD