Need Help with design issue

[tommurphy73]

Hi,

I have and old design product which outputs a frequency up to 100khz. This output frequency is meausred by an external instrument such as a counter or oscilloscope. The output buffer used in this application is a 74lcx244. The issue is the end customers will sometimes sort the outputs to ground and we get some returns with damaged lcx244 devices.

We have a new design which has similar outputs and in order to improve the fault tolerance we decided to go with open drain outputs. We used a 74lcx760 which is the equivalent of the lcx244 with open drain outputs. We put a 3k3 pull up resistor on all the outputs. The issue we are finding with this new device is that the seems to be a lot of noise and ringing on the output.

Anyone have any experience with something similar or can recommend a device which will give us fault tolerant outputs or something like hot swap capability.

Best Regards
Tom

 

[Nigel Goodwin]

The thing is your original design actually 'drove' the cable, your new design 'releases' the cable and relies on the 3.3K pullup to charge the capacitance of the cable - this means it doesn't go high very quickly, dependent on the length and capacitance of the cable.

Perhaps a transistor buffer after the old version might be a better choice?.

 

[tommurphy73]

I agree. We experimented with smaller value pull-up resistor values down as far as 650ohms and were able to get a much faster rise time. However this increased the ringing when the signal switched to ground through the open drain device. There was also a lot of noise on the waveform which we cannot explain.

What configuration of transistor would you use after the old buffer chip to give fault tolerance and short circuit protection


Rgds
Tom

 

[Roff]

I would design a driver with 50 ohm output resistance. It will drive cables with minimum ringing, and will be short-circuit proof. What supply voltage(s) are available to power the driver?

 

[tommurphy73]

Hello,

Thank you for taking the time to look at this for me. It is much appreciated.


We have either 3.3v or 5v supplies available to us. The 5V supply is an input voltage and has a tolerance which allows it to go up to 6.5V.

Another interesting thing which is related to the output impedance is that when measuring the frequency using an oscilloscope the signal is much noisier when using the X10 setting on the probe and much less noisy when measuring on the X1 setting. I believe that the input impedence of the scope is 10meg on X10 and 1Meg on X1

Best Regards
Tom

 

[Roff]

What is the desired amplitude and rise and fall times of your output signal? I'm assuming the input to the output driver will be a TTL signal.

 

[tommurphy73]

The maximum frequency out will be 100khz and the minimum will be 1khz. The rise and fall time is not too critical but at 100khz I would like it to be less than 5% of the pulse width. The signal level should be compatible with 3.3v cmos if possible.

Tom

 

[Roff]

Does anyone know if all commercially available 3.3V CMOS devices have 5V input compatibility?

 

[Willbe]

the signal is much noisier when using the X10 setting on the probe and much less noisy when measuring on the X1 setting.

So the source impedance of the "noise generator" is high, more like a "noise current source"? That is interesting.

The problem seems to be maintaining low output impedance but having a current limit, so the V-I curve would be a rectangle.

How fast can linear regulators (LMXXX) be controlled by their ground leg? Some are advertised as "bulletproof."

How high a series current sense resistor [and then a diff amp] could you use?

 

[Roff]

The problem seems to be maintaining low output impedance but having a current limit, so the V-I curve would be a rectangle.

How fast can linear regulators (LMXXX) be controlled by their ground leg? Some are advertised as "bulletproof."

How high a series current sense resistor [and then a diff amp] could you use?

A good low-impedance driver with a 50 ohm series resistor could be shorted to GND and still only dissipate a half watt or less.

 

[Roff]

Tom, are you prepared to redesign your PC board to make this change? How much room do you have?

 

[tommurphy73]

We are currently in the middle of a re-design so a good time to implement some changes but room is a bit of an issue as the board is fairly small and we are also trying to keep the cost down as much as possible.

Rgds
Tom

 

[mneary]

Try the original design, with a series resistor to limit the current within the rating of the driver.

 

[Roff]

Did your original design use a single 74LCX244 as the output driver, or did you have multiple buffers in parallel?

 

[tommurphy73]

The original has two lcx244 in parallel to generate multiple outputs

Rgds
Tom

 

[Roff]

Here is the simplest circuit I could come up with. The capacitor slows the transition times to reduce ringing at the end of an unterminated 50 ohm coax. You should test it with the load on the end of a 50 ohm coaxial cable. To drive 3.3V logic, you will have to leave out the 50 ohm resistive termination on the end of the cable. I would add a 50pF dummy load for the test.

 

[tommurphy73]

We did a test today and using a series resistor of 67ohms on each output of the LCX244 driver but no load capacitor and the signal was so much better than the LCX760. There was a small amount of ringing on each transition which was only obvious when using the oscilloscope in 100meg bandwidth mode and X10 probe. When limiting the bandwidth to 20Meg or using a X1 probe there was no ringing at all. There were no other spikes or other undesirable pulses which we were previously getting. I did not try any pull down capacitor but would expect it to filter out any remaining ringing on the signal.

Thank you for all the help. The pcb design will be updated on Monday and prototypes will be available in two weeks.

Great Forum

Regards
Tom

 

[Roff]

We did a test today and using a series resistor of 67ohms on each output of the LCX244 driver but no load capacitor and the signal was so much better than the LCX760. There was a small amount of ringing on each transition which was only obvious when using the oscilloscope in 100meg bandwidth mode and X10 probe. When limiting the bandwidth to 20Meg or using a X1 probe there was no ringing at all. There were no other spikes or other undesirable pulses which we were previously getting. I did not try any pull down capacitor but would expect it to filter out any remaining ringing on the signal.

Thank you for all the help. The pcb design will be updated on Monday and prototypes will be available in two weeks.

Great Forum
Regards
Tom

Why 67 ohms? Did you parallel outputs? And did you test the output at the end of a sizeable length of coax?

 

[Hero999]

Here is the simplest circuit I could come up with. The capacitor slows the transition times to reduce ringing at the end of an unterminated 50 ohm coax. You should test it with the load on the end of a 50 ohm coaxial cable. To drive 3.3V logic, you will have to leave out the 50 ohm resistive termination on the end of the cable. I would add a 50pF dummy load for the test.

Why have you used 43Ω instead of 51Ω?

 

[tommurphy73]

I used a 67 ohm resistor as it was the closest size to what I wanted that was available at work. Based on the maximum current drive capability of the LCX244 if I put a 50R resistor between the output and GND the current will be greater than the spec from the data sheet with a 3.3V output.