Power System for Driving a Matrix of Elements (Resistors, Diodes, etc)

Thread Starter

srv565

Joined Apr 10, 2019
4
Hello,

I was hoping to get some guidance or even direction to sources of solution for this project. Before mocking up a circuit diagram I wanted to know if the concept I was thinking is even physical. I'm not an electric circuit engineer, so forgive for any blaring mistakes.

In short, I'm trying to drive a matrix of elements (like an LED display) with a limited number of power buses N. It's impractical to drive N elements with N buses, i.e. 1000 to 1,000,000 elements as that would require same amount of buses. The elements are somewhat like a diode however they take very little current (less than 1 microamp) and are mainly voltage driven. There are only two ports for each element.

I have an N amount of power buses that can provide -V to V Volts 24 bit (from DAC). I was thinking of laying out rows and columns with these N buses and creating a matrix of voltage nodes. The device would go between these voltage nodes.

Something very similar to the below (the right diagram);


The CS# would be V+ buses and SW# would be V- buses for example.

I've looked high and low but most solutions are tailored for LED displays (like the above picture). The requirement is to power all elements simultaneously with voltages much higher than that of LEDs, i.e. 20 to 40 V. Switches or multiplexing would not work as sequential activation of each element will cause a few problems. I've thought of using a capacitor to hold the charge of each element during sequencing, like active matrix, but this also causes problems.

The way I am approaching this problem is that we know exactly what voltages we want at the nodes, we just need to solve for the N buses. I was thinking of using the impedance matrix from power flow theory to solve for the voltage buses. If I had the correct impedance matrix, would I be able to solve for the correct voltage buses assuming I know what the current is?

Is there some sort of dot matrix driver that's meant to power systems, not LEDs? I'm hoping instead of using this method there may already be a solution out there that's in the form of an IC.
 

Sensacell

Joined Jun 19, 2012
2,520
All of these matrix solutions are dynamic, time-multiplexed ideas.

Your requirement is vague and mysterious, you will get better help with more disclosure.
What are these "elements"??
 

AnalogKid

Joined Aug 1, 2013
8,229
Since the element voltage is so high, zillions of standard parts designed for LED systems will not work. but the concepts still hold. Switching 40 V at uA current levels is not particularly difficult, it just takes parts rated for the voltage. You would need PNP or p-channel transistors for one axis, and NPN of n-channel ones for the other axis. These kinds of parts can be bought ganged together in groups to reduce the assembly effort. For example, the ULN2803 can take eight 3.3 V or 5 V signals and has eight 50 V NPN output transistors in one IC. There is a complimentary PNP part.

A potential problem with the very low "element" current is that it is near the leakage current of some transistors in the off state.

How fast does the switching matrix need to be? What are the minimum times any element will be on or off?

ak
 

MisterBill2

Joined Jan 23, 2018
4,579
The concept of driving an array of devices is not at all new, and it has been done with LEDs and the circuits have been published. It would need a row of transistors to source the current and a complementary column of transistors to sink the current. That is not close to a new concept, and not original with me. It is possible, depending on the control logic, to power the elements either one at a time or all at once. Of course at that point both power supply and control logic can get complicated.
 

Thread Starter

srv565

Joined Apr 10, 2019
4
Thank you all for the suggestions!

The elements can be considered as devices that are voltage based, like varactor diodes.

One very important point that I did not mention is that the voltage at each node is a unique value. If we were to separate the voltages into an (ij) matrix (i for row, j for column), then the voltage at (00) would be at a different value than at (55) for example. This isn't a binary problem (or I would've used the many designs out there for binary matrix systems, a value of either off or on).
The voltage at each node also ranges from 0 to 40 V (or - 20 to 20 V) but 0 - 20 V (- 10 to 10 V) is also doable.

The matrix would have a preferred refresh rate of 60 Hz, 30 Hz, or even 5 to 10 Hz is also fine. All elements would be on at all times with changing voltages. For example, we can imagine a unique offset cos (pulse, tri, or some waveform) being applied to each node.

The most important point is to maintain a unique voltage at each node without turning anything off. Is it possible to power all elements at once each with a different voltage potential across the nodes?
 

MisterBill2

Joined Jan 23, 2018
4,579
Thank you all for the suggestions!

The elements can be considered as devices that are voltage based, like varactor diodes.

One very important point that I did not mention is that the voltage at each node is a unique value. If we were to separate the voltages into an (ij) matrix (i for row, j for column), then the voltage at (00) would be at a different value than at (55) for example. This isn't a binary problem (or I would've used the many designs out there for binary matrix systems, a value of either off or on).
The voltage at each node also ranges from 0 to 40 V (or - 20 to 20 V) but 0 - 20 V (- 10 to 10 V) is also doable.

The matrix would have a preferred refresh rate of 60 Hz, 30 Hz, or even 5 to 10 Hz is also fine. All elements would be on at all times with changing voltages. For example, we can imagine a unique offset cos (pulse, tri, or some waveform) being applied to each node.

The most important point is to maintain a unique voltage at each node without turning anything off. Is it possible to power all elements at once each with a different voltage potential across the nodes?
This addition to the challenge, that each of the nodes would have a different value, alters the game BIG TIME. Once a voltage is set for a specific row, only the column voltages can change to set the voltage for the specific node, making the voltages for other nodes very limited in possible values. Of course, if only one node at a time is energized then it is just a matter of assigning a voltage code for each point, along with each address, and having a fast enough D/A converter. So now I am wondering if this is intended as part of a viewing screen.
 

Thread Starter

srv565

Joined Apr 10, 2019
4
In that case, would I have to use the well established switch multiplexing systems to manage this matrix while connecting each element to some capacitor to maintain it's charge?

I was thinking the impedance matrix would give me the correct values for the N bus voltages such that each matrix voltage node would have the desired unique values. Is my understanding of implementing the impedance matrix incorrect then?
 

MisterBill2

Joined Jan 23, 2018
4,579
Now I am rather confused, close to totally confused. When I use the term "impedance matrix" it is an array of numbers, each with a magnitude and an angle. In software terms it would probably be called an array.
In the LED example in post #1 only one LED at a time would be driven, but the rate of being energized make it look like constant illumination to our human eyes.
So a numerical example of the values similar to what you are seeking may help clarify it quite a bit. even just a 4 by 4 matrix could clarify it, I think.
 

mvas

Joined Jun 19, 2017
538
Can you put a simple Sample-And-Hold Op-Amp at each node in the matrix?

Step #1 - Disable all Rows & Columns
Step #2 - Output the appropriate DAC Analog Voltage for the next Node on a single "Node Voltage" wire - connected to all node inputs
Step #3 - Enable one Row & one Column wire to enable the selected Node OpAmp, to LATCH the Analog Voltage for that specific node
Repeat for all nodes

You would need a Single DAC ...
a) A single ( + ) Bus Power wire
b) A single ( - ) Bus Power wire
c) A single DAC Analog Voltage wire
d) X # of Column Select wires
e) Y # of Row Select wires
Total Wires = x + y + 3

If "voltage load time" is an issue then ...
Use one DAC per column and then address / latch the entire row simultaneously, repeat for each row.
a) A single ( + ) Bus Power wire
b) A single ( - ) Bus Power wire
c) X # of Columns of DAC Analog Voltage wires
e) Y # of Row Select wires
Total Wires = x + y + 2

If "voltage load time" is not an issue ...
Instead of using Row & Column wires, you can use 2-Wire Serial communications
You would need ...
a) A single ( + ) Bus Power wire
b) A single ( - ) Bus Power wire
c) A single DAC Analog Voltage wire
e) A Serial Data wire
d) A Serial Clock wire
Total Wires = 5
 
Last edited:

MisterBill2

Joined Jan 23, 2018
4,579
The arrangement described in post #9 could indeed deliver the stated performance but it would be a huge jump in complexity. A sample and hold node at each cross-point becomes huge very rapidly. So it is not reasonable that it would be the final choice.

Clearly it is important to understand the actual goal before a method of obtaining that goal can be reached.
 

mvas

Joined Jun 19, 2017
538
Adding a single Sample & Hold chip next to each "Device" and then
connecting the Row & Column wires to the S&H chip instead of the "Device" is somehow a "huge jump in complexity" ?
Clearly, it is not a huge jump in complexity.

Other ideas ( assuming 8 nodes per row )
A single 1:8-Channel Analog Sample & Hold IC can supply the unique (+) Bus volts for all 8 Nodes in one (1) entire row.
A single 1:16-Channel Analog Sample & Hold IC can supply the unique (+) Bus volts for all 8 Nodes in two (2) entire rows.
A single 1:32-Channel Analog Sample & Hold IC can supply the unique (+) Bus volts for all 8 Nodes in four (4) entire rows.
 
Last edited:

MisterBill2

Joined Jan 23, 2018
4,579
Adding a single Sample & Hold chip next to each "Device" and then
connecting the Row & Column wires to the S&H chip instead of the "Device" is somehow a "huge jump in complexity" ?
Clearly, it is not a huge jump in complexity.

Other ideas ( assuming 8 nodes per row )
A single 1:8-Channel Analog Sample & Hold IC can supply the unique (+) Bus volts for all 8 Nodes in one (1) entire row.
A single 1:16-Channel Analog Sample & Hold IC can supply the unique (+) Bus volts for all 8 Nodes in two (2) entire rows.
A single 1:32-Channel Analog Sample & Hold IC can supply the unique (+) Bus volts for all 8 Nodes in four (4) entire rows.
A sample and hold would be required for each and every cross-point, not just at the input. Thus an 8 by 8 matrix would need 64 of them, and using a multiplexed IC would totally defeat the whole purpose of having the matrix arrangement. Besides that, the stated voltage spread appears to be from -24 volts up to +24 volts, so now the choice of an IC sample and hold device gets a lot fewer options.
And we still have no hint as to the purpose of this mystery system.
 

mvas

Joined Jun 19, 2017
538
No, you would not need 64 individual IC chips for an 8 x 8 matrix of 64 nodes ...
If you used my suggested 32-Channel S&H IC chip,
then you would need only two (2) IC chips for 64 Nodes.

One physical IC chip can provide the unique voltage required by 32 nodes.

I have moved past the original "Row x Col Matrix" idea in message #1, and I think the TS has also.

I suggest "Binary Addressing" of the S&H Nodes ...
Binary Addressing supports many more nodes vs X-Y Matrix addressing with fewer control wires.
8 Row + 8 Column is 16 control wires = only 64 nodes.
8 Binary Addressing wires = 256 Nodes ... Half the control wires but 4 times the nodes ... that is less complex
 
Last edited:

MisterBill2

Joined Jan 23, 2018
4,579
I refer back to post #2, in that the actual goal is a complete mystery, and thus all of the suggestions are guesses based on no real information about an actual system to accomplish some purpose.All we have been told is that there is to be an array of "things" that are voltage dependent, and that the voltage to these things is to be individually set to different values. So the whole thing is still a mystery.
 
Top