Sorry about my blurry input; but thats because I'm new to the subject and I also don't know what to give as helpful information to you.

Yes, we will connect the output of the rows to the DAQ device and then to a PC. Then by the help of software, the output voltages read from the rows (with the sequential driving of columns) will be interpreted. The DAQ has its own A/D converter inside. But in its spesifications I could'nt see anything about settling time. But as far as I know, they use the same DAQ in very dynamic tests like crash tests; so there shouldn't be a problem about settling time.

What I didn't understand is these MOSFET switches or darlington arrays you are talking about, how are they going to accomplish sequential driving of columns? I am a bit confused.

What's more, doesn't the resolution depends on number of sensels/resistances per unit area in our sensor? How does it affect our system or settling times? Or are we talking about different resoultions?

Thank you very much for your patience and interest =)

Ayca

 

Sorry about my blurry input; but thats because I'm new to the subject and I also don't know what to give as helpful information to you.

Hey, I'm not mad! Just trying to get the facts out of you. This can be rather trying with "n00bs", as quite often they don't have a good grasp on the whole "big picture" thing. On the flip side of the coin, those of us trying to help out frequently miss some detail that the poster provided, and that can result in a number of iterations before everything gets sorted out.

Yes, we will connect the output of the rows to the DAQ device and then to a PC. Then by the help of software, the output voltages read from the rows (with the sequential driving of columns) will be interpreted. The DAQ has its own A/D converter inside. But in its spesifications I could'nt see anything about settling time. But as far as I know, they use the same DAQ in very dynamic tests like crash tests; so there shouldn't be a problem about settling time.

OK, do you have a part number and manufacturer name of this DAQ device? It will be a huge help to us to read up on it.

What I didn't understand is these MOSFET switches or darlington arrays you are talking about, how are they going to accomplish sequential driving of columns? I am a bit confused.

OK. Either the Darlington drivers or MOSFETs would be used as switches to turn on or off current to rows and columns. There will be other devices that control the sequencing of which switches get turned on/off at what times.

Darlingtons are made using BJTs, or Bipolar Junction Transistors, or what you know as a common silicon transistor. The problem with using BJTs is that there will be a voltage drop between the emitter and collector when it is switched on, and that voltage drop will not be consistent; it will depend on the current being passed through it and the individual BJT. This will introduce a degree of non-linearity that for your application should be avoided. Otherwise, your calibration procedure could become tedious to the extreme.

MOSFETs have a low Rds(on) that is relatively consistent, and results in a relatively small voltage drop from drain to source. Where a saturated BJT might have 0.7v to 1.2v or so from collector to emitter, a MOSFET such as a 2N7000 might have just a few Ohm's resistance - and this is a small fraction of your piezoresistor transistor's resistance. This means that the linearity of the response will be mostly dependant upon your piezoresistive transducer rather than the driver circuitry.

What's more, doesn't the resolution depends on number of sensels/resistances per unit area in our sensor? How does it affect our system or settling times? Or are we talking about different resoultions?

We're talking different resolutions.

I'm talking about the resolution of the ADC (what you're calling the DAQ).

It's going to take a certain amount of time to charge the ADC input from the resistive divider, and then for the ADC to convert the sample to a near-final result, then re-sample the input before providing the final result. YMMV (Your Mileage May Vary) - there are quite a few different schemes for ADC. None of them are instantaneous. Generally, the faster the ADC with a high bit count, the more expensive it is.

Isolating the digital side from the analog side will be another can of worms - we'll visit that later. In the meantime, need more info on this DAQ as I mentioned above.

 

Thanks for the response! Our DAQ has 16 bit resolution and 100 kHZ/channel. I have actually never thought that the whole solution was going to be this complex : ).
My job here will be to concentrate on the control circuit used for the switching of columns. However as far as I understood the rest of the system also affects the circuit which will be used for switching. Well, life is difficult. : )

 

I've designed a circuit which lights 10 LEDs sequentially by using a 5 stage Johnson counter and a 555. But still how to integrate/customize it to a switching circuit is unknown =s ?

 

Here is the circuit I made, it looks a little bit messy I know For now, it just sequences 17 lights but if I add more Johnson Counters I can expand it to 42 as I needed. Any comments, suggestions?

This ciruit can be useful for my application, right?. Now, I have to think of the switches. I haven't got any response from you for so long and I am getting scared because I really dont know what is the best switch that can be used in my spesific case.

Thanks again in advance..
Ayca

 

Actually, I did start on a version. See the attached; it's quite large, and certainly not complete.

The logic for the horizontal/vertical scan is there; it's simply a smaller version than what you require. IC4 provides 9 outputs, IC5 and IC6 provide 8 each, for a total of 25 columns.

Similarly, IC1 provides 9 outputs, IC2 and IC3 provide 8 each for a total of 25 rows. So, this is roughly 1/4 scale of what you need.

A 50kHz master clock is input to the circuit at A1 (lower left).
At B1, ROW-CLK is a synchronization pulse to indicate that a row scan has been completed, and the next is about to start.
COMPLETED indicates that an entire scan of the array has completed, and the next is about to start.
(Note: there is a missing junction "dot" right below the "E" in completed.)

I started with using 2N7000's for the upper and lower side of the resistors. However, I quickly realized that an array of 48 2N7000's will need a gate driver IC; a 4017 would be woefully inadequate by itself. I'm thinking that a MicroChip TC4468 Logic-Input Quad CMOS Driver is a good candidate for the low-side driver (as wired now, the columns, or IC4-IC6). A high-side driver is still TBD.

It's going to be tough to get all of those components on a board!

 

OK, I don't know what I was thinking about with driving a zillion gates and needing drivers. I was tired when I was last working on the schematic.

Here's another iteration. Only the first and last MOSFET transistors and piezoresistors are shown for each row/column 4017 drivers for clarity's sake.

The key to the whole thing is at location F3. V+ is the voltage source that will feed your array. R37 is a resistor that you'll have to decide on the value, but likely somewhere near the max value you might read across a piezoresistor. VSENSE is the point where you'll connect your ADC.

The Vdd for the IC's will need to be around 15v. This is to help ensure that the gates on the 2n7000's get fully turned on. V+ and R37 should be chosen so that the maximum voltage across any piezoresistor is 10V lower than Vdd; or 5v. So, if R37 is equal to the highest value a piezoresistor might present, and V+ is 10v, then your VSENSE voltage range will be between 0v and 5v.

Is this making sense to you?

 

Here is the circuit I made, it looks a little bit messy I know For now, it just sequences 17 lights but if I add more Johnson Counters I can expand it to 42 as I needed. Any comments, suggestions?

OK, here are some suggestions:
1) Go to Freescale or ONSemi's site, and download the datasheet for the MC14017B. (where almost everyone else used a CD prefix for CMOS 4000 series IC's, Motorola used MC1 instead.) In that datasheet you'll see a method for cascading multiple 4017's using an AND gate between each stage. The first and last stages in the string of counters has unique wiring; the intermediate stage can be cloned for as many stages as you need.

2) The circuit you drew would not work properly, as it is lacking the AND gate logic.

3) Don't be so stingy with your use of the ground symbol You can use it as many times as you like. It helps avoid having a huge spaghetti of ground lines running around.

4) Since the 4017 will have only one output high at a time, you could have used just a single current limiting resistor on the ground side of all of your LEDs - as long as they were all the same rating (Vf @ current)

This ciruit can be useful for my application, right?. Now, I have to think of the switches. I haven't got any response from you for so long and I am getting scared because I really dont know what is the best switch that can be used in my spesific case.

OK, I've drawn it up using 2N7000 MOSFET transistors, which are an industry standard part, and are pretty inexpensive. These are in a TO92 case, and are thru-hole mount. If you can handle SMT/SMD, you could go with 2N7002 or NDS7002A, which have similar specifications but are much smaller.

The price isn't much different. At Mouser:
2N7002 - $6.60 for 100
2N7000 - $8.80 for 100

 

Firstly, thank you very much for giving your valuable time to me! Yes, these designs definetely made sense, you used the same approach in the datasheet to cascade 4107s. and R37 is to adjust the output to the range of the data acquisition device/ADC used, right?

Actually, the DAQ device we use have enough inputs for all the rows, so sequencing only columns will be adequate (i think this circuit also sequences both. ) In other words we can sense all rows simultaneously with the DAQ we have, so no need to sequence them. Will there be any disadvantages of this do you think? But we will need some kind of a synchronization pulse for columns as you did for the rows (ROW_CLK), am I correct?

What's more, thank you very much for the ground advice. I dont know why, but I have some kind of a obsession of connecting all the grounds to the same place

Again, thank you!
Ayca

 

And I have another question too.. It may be stupid I am not sure : ) IS there something for generating clock pulses, or do I have to generate my clock pulse on my own using a IC like 555? Because the output of my clock pulse generator is not perfect, lengths of the pulses are not equal.

 

Firstly, thank you very much for giving your valuable time to me! Yes, these designs definetely made sense, you used the same approach in the datasheet to cascade 4107s. and R37 is to adjust the output to the range of the data acquisition device/ADC used, right?

Yes, the combination of R37 and V+.
There's going to be a happy medium somewhere, where you get decent response time from the ADC, decent accuracy, and are not stressing the piezoresistors.

The input of the ADC will likely be a small sampling capacitor. So, the time it takes to charge that capacitor will be determined by the size of it, and R37. However, to get good range out of the piezoresistors while ensuring low stress, a resistor as large as the maximum value of a piezoresistor could be used.

Actually, the DAQ device we use have enough inputs for all the rows, so sequencing only columns will be adequate (i think this circuit also sequences both. )

I guess it would've helped if I'd read the documentation on it.

In other words we can sense all rows simultaneously with the DAQ we have, so no need to sequence them. Will there be any disadvantages of this do you think? But we will need some kind of a synchronization pulse for columns as you did for the rows (ROW_CLK), am I correct?

Well, yes. Actually then, you could use just the IC4-IC6 portion of the circuit to scan the columns. The counters advance on the rising edge of the clock. If you triggered your ADC on the falling edge of the clock (use an inverter to make an opposite polarity clock if you need it) then the circuit would have time to settle for the ADC to take the sample.

What's more, thank you very much for the ground advice. I dont know why, but I have some kind of a obsession of connecting all the grounds to the same place

Again, thank you!
Ayca

You're welcome

OK, on the grounds - there can be several types of grounds in a circuit. Usually, everything is digital, or everything is analog, or everything is power. But this is one of those circuits where you have digital switching an analog circuit, and you can get bitten due to the noise caused by the digital side. It's like inviting a punk band to play in a library.

Keep your digital grounds separate from your analog grounds. Use 0.1uF capacitors across the supply pins of EACH logic IC. The grounds to the 2N7000 MOSFETs should all be connected together (analog ground), and the grounds to all the logic IC's should be connected together (digital ground). The two grounds (analog and digital) should be connected together in only ONE place.

 

And I have another question too.. It may be stupid I am not sure

Most questions aren't stupid. It's some of the answers you get around here that are

IS there something for generating clock pulses, or do I have to generate my clock pulse on my own using a IC like 555? Because the output of my clock pulse generator is not perfect, lengths of the pulses are not equal.

Well, 555 timers have been wildly popular for many years, and they're pretty good. However, they're not terribly accurate due largely to the RC time used and stability over time and temperature. I don't know how accurate of a timebase you really need.

You can get crystal oscillators in DIP form, and then use a counter or series of counters and/or a series of D-type flip-flops to divide the frequency down.
An XO is a crystal oscillator.
A VCO is a voltage-controlled crystal oscillator; it will be more stable than an XO
A TCXO is a temperature-compensated crystal oscillator; some manufacturers say temperature controlled instead.
An OCXO is an oven-controlled crystal oscillator; after a warm-up time of about a half-hour to an hour, it will be very stable, several orders of magnitude better than you could achieve with a 555 and an RC time network.

 

Hola!
Sorry for my late reply, I was not in my office for 3 days. Thank you again Sgt! Attached, is the simple version of our circuit, the thing is I am not sure I have put the synchonization point right. Is it? I haven't understood the concept of where to put that sync. point yet I think !? Also note that Ive just put 1k resistors instead of our sensors resistors, because I couldnt find something better to put as sensor's piezoresistors in NI Multisim.

I will spend more time on selecting the components next week. But probably a crystal clock will be needed, since I want high frequencies.

Thanks for spending your valuable time!
Tschüss!
Ayca

 

And I have a final question just for my understanding: What would happen if we just connect the output of 4017s dirreclty to the mat without MOSFETs?

 

Copied from a PM...

My internship is lasting in a few days but before I go, I will test my circuit, probably tomorrow.(I am waiting for the shippment) I posted a simple version of the circuit under the thread " sequential switching". I would be really happy if you just give it a look. It is almost the same as you put under the thread, except it only sequences columns.

OK, the circuit looks good so far. You can expand it by "splicing in" more 4017 counters as U3 is wired (each will require an additional AND gate, of course.)
Your array of 1k resistors represents the piezoresistors, correct? If so, all's well and good. You haven't shown the Rsense; perhaps that's in your ADC unit - I don't know.

However, I still didnt get the MOSFET concept, what would have happened if we connected the outputs of the 4017s directly to the mat? I am really curious about this. Why do we need MOSFETs? Still not clear.. Sorry for private messaging, but it was kind of urgent =s

That's OK, you posted this awhile ago and I hadn't responded. Sorry, I've been busy.

The 4017 CMOS ICs have very low current handling capability, and their output is the wrong polarity. The 4017 outputs only one '1', the remainder of the outputs are low.
You need the low side of only the piezoresistor being tested to be connected to signal ground.

The 2N7000 MOSFETS serve as inverters, and also allow isolation of the digital side from the signal ground side.

While you might be able to use something like a 40106 hex inverter, you will not get output as linear as with the 2N7000 MOSFETS, nor would you be able to isolate the digital ground from the signal ground.

CMOS IC's have very little current sink capacity. With Vdd=10V, sinking just 1.3mA could make the output pin's voltage rise to 0.5v, which would very significantly affect the accuracy (linearity) of your readings. In contrast, the 2N7000 has a very low RdsON (maximum=5 Ohms). Even at 75mA, the drain would have less than 0.45v on it - and you're not going to have anywhere near that much current.