How can I convert 3.3V IO to 5.0V without compromising signal integrity?

[jbbjshlws]

Hello, I have a circuit that i will be running the IO lines a reasonable distance, and i would like to amplify the lines from the current 3.3Volt to 5.0volt. I'm sure this is probably a simple task and (hopefully) their is a single IC someone could direct me to. I thought that maybe a Darlington transistor array could work, but I would like to verify this and know the pros and cons. Obviously latency is an issue.

Any help is greatly appreciated.

Josh.

 

[berkeman]

74ACT244 would do the trick.

 

[jbbjshlws]

Thank you for you reply, i will be testing this in a couple of weeks when the parts arrive, and i will let you know the result

 

[jbbjshlws]

just a quick one,
would
http://jaycar.com.au/products_uploaded/ZC4867.pdf
or
http://jaycar.com.au/products_uploaded/sdls144.pdf

do the trick?, they are a local supplier of electronics here in Australia, and it would be easier for me to get the parts here

 

[berkeman]

just a quick one,
would
http://jaycar.com.au/products_uploaded/ZC4867.pdf
or
http://jaycar.com.au/products_uploaded/sdls144.pdf

do the trick?, they are a local supplier of electronics here in Australia, and it would be easier for me to get the parts here

Yes, the 74S series would work, since Vih=2V min is compatible with 3.3V outputs.

BTW, if you are driving lines very far, you will need to treat them as transmission lines, and at least back-terminate them. With 74S logic, you may have enough current output to forward-terminate the lines, but I haven't done that calc in a while, so I don't know offhand. You also need to use a reasonable cable for the transmission lines, like Cat-5 twisted pair cables or something. Each twisted pair would have the signal and a ground wire.

 

[jbbjshlws]

32-40 meters would be the transmission distance i was going to use twisted pair, but their is 10 lines all up. Ill work something out and let you know how i go, cheers

 

[berkeman]

32-40 meters would be the transmission distance i was going to use twisted pair, but their is 10 lines all up. Ill work something out and let you know how i go, cheers

That's pretty far to be sending TTL signals. A better choice would be to use a line driver/receiver pair for the haul down the twisted pair to the far end, then convert it back to 3.3V or 5V logic signals. You should be using a forward termination for going that far (like, terminate in the 100 Ohm characteristic impedance Zo of the Cat-5 twisted pair).

A line driver/receiver pair like ones from Texas Instruments:

http://focus.ti.com/docs/prod/folders/print/sn75lbc179.html

Cat-5 comes with 4 pairs per sheath, I think, so you can do your 10 signals in 3 Cat-5 cables with 10 pairs of line driver/receivers.

 

[jbbjshlws]

the only problem i would have is i am using this signal and daisy chaining it to several TLC5940's (LED PWM drivers). 4 groups of 3 in fact, (RGB), the furthest one is at most with current design is about 40 meters, but i am looking at changing the design to make it 10 meters. if i used a 74S series chip at each cluster of TLC5940's would that do the trick?

 

[jbbjshlws]

I have changed the circuit, the transmittion distance is now 5 meters, a bit more realistic.

I have tried to wire up the buffer ic, but with no luck.
The Buffer that i think would be the correct one is the 74HC240 Octal Buffer Line Driver IC asit has a higher fan out, and i am driving 12 IC's off of it.

I think that the 1A0 - 2A3 lines connect to the output of the device, and the corresponding 1Y0 - 2Y3 lines connect to the driving lines. GND to the device, the VCC into the Buffer chip would need to be the 5.5 volt from an external source, and the 1OE and 2OE trigger lines should connect to Clock Select.

is this correct? or am i doing something silly
Also are Buffers Bi-Directional?
--Josh

 

[jbbjshlws]

I am still stuck on this issue
any help would be greatly appreciated

 

[berkeman]

Can you post a schematic of your drivers, transmission lines and receivers, including distances? What is your data rate, and what termination scheme are you using now?

 

[jbbjshlws]

Hello Berkeman,
I am driving 12 TLC5940's LED Drivers from a propeller Parallax Micro controller.

The micro controller outputs IO voltages at 3.3 volts, and over the 4 meter distance from the controller to the last chip (all are connected in parallel along the 4 meter distance). there is a .3 volt drop (to 3.0 volts) and this is the minimum working spec of the TLC5940 chips.

I believe the TLC5940's are indecisive as to whether the line is high or low as i think over the transmitting distance the resolution is not defined enough for the chips to distinguish the difference. i believe Schmidt triggered buffers would get around this as the outputting lines would be ramped up to 5.0 volts. A 74LS240 non inverting tri-state octal buffer i think would be the right chip.

the current draw on the vcc line with all of the chips combined is 500mA @ 3.3 volts. with 5 volt lines this should become less, as the resistance should stay constant.

The termination is through 5k resistors pulled to vcc. the data rate is at maximum 30mHz,
and all of the cables are shielded.


Tell me if you think this is the best way to achieve this. i cannot make the distance of the wire shorter, and the conductors are currently the largest given the restrictions.

Any help is greatly appreciated.
Joshua

 

[berkeman]

The micro controller outputs IO voltages at 3.3 volts, and over the 4 meter distance from the controller to the last chip (all are connected in parallel along the 4 meter distance). there is a .3 volt drop (to 3.0 volts) and this is the minimum working spec of the TLC5940 chips.

What voltage is the Vcc for the TLC5940 chips? What is their Vih specification? What is the resistance of your twisted pair wire to give you the 0.3V drop in just 4m?

the current draw on the vcc line with all of the chips combined is 500mA @ 3.3 volts. with 5 volt lines this should become less, as the resistance should stay constant.

The 500mA gives how much voltage drop in the power wires out to 4m? So you have the option of running the TLC5940 LED drivers at 3.3V or 5V?

The termination is through 5k resistors pulled to vcc. the data rate is at maximum 30mHz,
and all of the cables are shielded.

5kOhm is not a termination. You would either forward terminate with a resistor pair at the end of the line, or back terminate at the transmitter. The termination values will be on the order of the characteristic impedance of the twisted pair transmission line.

What do you mean by 30mHz. Is that 30MHz (30 MegaHertz) or 30mHz (30 milliHertz)? Or maybe you mean 30Hz or 30kHz?

 

[jbbjshlws]

The VCC for the TLC5940's is between 3 - 5.5 volts,

i have used ribbon cable, not twisted pair, twisted pair was not feasible in this situation, there was limited space restrictions and the additional 2nd core (making 16 wires total for communication lines) took up to much area.

500mA is the current draw when the voltage drop is 0.3Volts over the 4 meters when the chips are in operation.

The speed of the transmission wires are at 30 MegaHertz, sorry for the confusion.

Hope this helps

 

[berkeman]

The VCC for the TLC5940's is between 3 - 5.5 volts,

i have used ribbon cable, not twisted pair, twisted pair was not feasible in this situation, there was limited space restrictions and the additional 2nd core (making 16 wires total for communication lines) took up to much area.

500mA is the current draw when the voltage drop is 0.3Volts over the 4 meters when the chips are in operation.

The speed of the transmission wires are at 30 MegaHertz, sorry for the confusion.

Hope this helps

Well, I don't think you are going to get reliable communication at 30MHz over ribbon cable, especially for a multi-drop, unterminated line. There are a number of reasons, including:

** Crosstalk

** For multi-drop transmission lines (TLs), you would usually use a full forward termination at the end of the line, and ensure that each drop had light enough loading on the TL so as to not cause any intermediate reflections. The 240 series of line drivers has enough drive to support a forward termination. Ribbon cable has a Zo around 90 Ohms, IIRC. It's best to split the forward termination, so you have a symmetric termination with respect to the Vih and Vil of the receivers.

** A back termination generally cannot be used with multi-drop TLs, because the first reflection coming back from the open end causes ringing to be seen by the intermediate nodes.

Even if you do use forward terminations in the correct value of Zo, the crosstalk issue is going to be a problem.

Can you describe the application more, and say why you are not using standard Cat-5 type twisted pairs to carry your 30MHz data stream? How are the data encoded in this data stream?

 

[jbbjshlws]

At the moment, i have ribbon cable connecting 12 TLC5940 LED PWM Drivershttp://focus.ti.com/lit/ds/symlink/tlc5940.pdf" . They are nearly working except they occasionally flicker. I believe this is because there is not enough contrast between the High and Low data (currently running at 3.3 volts) and depending on how much current the chips are drawing (relative to the amount of modulation). I think that if the data was running at a higher voltage, the 0.3 volt drop would cause less of a flicker if any because the 0.3 volt drop would then be negligible over a 5.5 volt total voltage on the vcc line.

I hope this better explains my project.
I have attached a video to show the layout, i don't think that the information explained in the video is still relevant.
http://www.atticsolutions.com.au/joshuasstuff/dancefloorissues.mp4

 

[berkeman]

At the moment, i have ribbon cable connecting 12 TLC5940 LED PWM Drivershttp://focus.ti.com/lit/ds/symlink/tlc5940.pdf" . They are nearly working except they occasionally flicker. I believe this is because there is not enough contrast between the High and Low data (currently running at 3.3 volts) and depending on how much current the chips are drawing (relative to the amount of modulation). I think that if the data was running at a higher voltage, the 0.3 volt drop would cause less of a flicker if any because the 0.3 volt drop would then be negligible over a 5.5 volt total voltage on the vcc line.

I hope this better explains my project.
I have attached a video to show the layout, i don't think that the information explained in the video is still relevant.
http://www.atticsolutions.com.au/joshuasstuff/dancefloorissues.mp4

Are you daisy-chaining the serial TLC5940 devices (as shown on page 12 of the datasheet), or are you multi-dropping them with separate chip selects? Why are you running at the max datarate of 30MHz? Do you really need to do that?

I wasn't able to open the *.mp4 file easily -- I must not have the right player installed on my PC. Do you have any *.jpg files you can link to or upload?

You should also calculate the ground shifts that you will get with the delta-I current changes as the LEDs turn on and off. Are the various grounds all just referenced to the ribbon cable, or is there a separate metal chassis ground that is used to tie the ground returns harder together to mitigate the ground bounce effect on signal integrity?

 

[jbbjshlws]

All of the TLC5940 chips have all input lines in parallel, except for the Serial Data in, the SIN goes through each chip.
The ground is the same ground in the ribbon cable.
I am not running at the max data rate at the moment. it is much lower than that, but if later it is needed i figure its better to over engineer a problem. and not introduce more..

I am not sure how to calculate the delta- I current changes.

http://www.atticsolutions.com.au/joshuasstuff/

there is a couple of pictures in there if that helps

 

[berkeman]

Yes, the pictures help. Thanks. You really do need to take several considerations into mind with this kind of project. You expressed an interest to get the engineering part of it right, so you really need to deal with at least these three key issues:

-1- You need to consider whether the data lines can be considered a lumped system (the length is much less than the bandwidth of the transmitted data, both the data bandwidth, and harmonic content on the edges if you are not slew-rate-limiting the waveform), or if you need to treat it as a transmission line. If you run that at 30MHz with sharp XX240 drivers, for sure you will need to treat it as a true transmission line, with impedance control, multi-drop issues, and a valid termination scheme. Even if you run it at 1kHz datarate, the sharp edges on the XX240 driver outputs will generate reflections and ringing on the signals as seen by the receivers.

-2- With a ribbon cable, you really do need to consider the effects of crosstalk on signal integrity. At the very least, the ribbon cable should have the signals ordered: Ground-Signal-Ground-Signal-Ground-etc. That may reduce crosstalk enough, but you will need to look closely at it in the real system.

-3- Putting 0-500mA down the cable will generate ground offsets and ground bounce (when the current draw changes as LEDs are turned on and off). Since you are using a direct-connected datapath (not transformer isolated), the ground reference for the transmitter and the receivers will be different, based on the voltage drop coming back up the ground wire(s) from the current-consuming nodes. This ground shift eats into your DC signal margin, and causes issues with AC signal integrity via ground bounce.

Have you studied transmission lines enough to understand why I am concerned about these things? I work with data transmission line issues every day, and for a system to be reliable, you cannot be cavalier about the issues I've highlited in this thread. For a hobby project for a dance lighting system, where you also have the flexibility to run well below the max datarate supported by the basic components, you may be able to make it work. But for a real product that you design in the future, you will want to understand and manage the TL considerations, or the product will not be reliable in the field.

Ack, I just remembered something else that you mentioned -- I think you are daisy-chaining the serial data line. That's okay, but you will want to back-terminate each line at the driver (the serial output of the LED drive chip). That will keep the receive end from seeing ringing on the line.