OK, so be careful with your output voltages if you are running the OP-Amps on a positive and negative supply, the output can go negative and above the input threshold of the transceiver. If I remember right, that part has an input voltage of 3.3V but I don't remember what the max is, same with a negative voltage, you don't want to exceed the ratings of the part. The outputs of OP-Amps can go close to the power voltage rails and that's why I mentioned it in my other post. You can use a single voltage part but you need to look at the specs to ensure they will swing to the levels needed by the Xcver. Both of those OP-Amps you mentioned have short circuit protection on the output so you should be safe in that respect also. I suspect you are right in that the uC controller doesn't have any ability to configure it's IO pins other than Input ot Output, but you might look to be sure. That said, you might just try a single transistor driver circuit and depending on if you need to invert the input or not and see if that will work. The output of that 741 looks really strange and I don't know if you had any load on it or not, but it just looks like a large capacitance on the output? It looks to start to go up and then starts a ramp, like something is really loading it down? I'm hoping someone else can chime in on this, I'm running out of ideas... :-/

 

You might want to just use an optoisolator for simplicity.

 

You might want to just use an optoisolator for simplicity.

Really like the idea of this. Will have to to track one down.

I've finally found a spare moment to draw out my circuit. I've only shown the input of the transceiver, hopefully that's all that's necessary to grasp what's needed to drive the circuit. If anyone can gleam any useful observations based on the info displayed here, I'm all ears. Again, the uC worked with a 741 op-amp but with too low a slew rate, meanwhile everything crashes with a 318.

 

If it's a digital signal to the transceiver, what about using something like a Schmitt trigger to 'clean up' the signal?

 

Can you provide the part number of the Transceiver rather than just the input circuit shown? Looking at the datasheet for the MAX3483 … … … that Sparky 1 posted the link to, I don't see a reason why the uC wouldn't drive that part. If that's not the part, please post the one you're using if you have that info. Not saying that I can help on this, but maybe someone else can see something right off?

 

Hey everyone,

I've been making some progress on my 318 buffer, although now I'm dealing with some ~5MHz power supply noise.

Here's my circuit in more detail than I posted before:



Here's a shot of the noise on the scope (bottom signal) after I soldered it all together on a piece of protoboard:



Some my component choices just came down to trial and error (I'm no engineer). I found the signal was cleaner without the 5pF cap on the negative feedback line, for example. Adding a 10K resistor to the negative feedback line also increased the gain above 1 for some reason. I added 1uF tantalum caps on the power supply because, well, someone told me to, though I didn't find they made any difference. I could definitely use some guidance on how to more wisely choose (read: calculate) my components.

Some other notes: the signal was very clean on the breadboard - this noise only showed up after I soldered it all together on the protoboard. Also, this scope shot is from the buffer-out NOT connected to the transceiver, but to a 10K resistor instead. I want to make sure the signal is good before I start trying to get it work with the transceiver.

Any pointers would be greatly appreciated.

Also...

Can you provide the part number of the Transceiver rather than just the input circuit shown? Looking at the datasheet for the MAX3483 … … … that Sparky 1 posted the link to, I don't see a reason why the uC wouldn't drive that part. If that's not the part, please post the one you're using if you have that info. Not saying that I can help on this, but maybe someone else can see something right off?

Unfortunately I can't. I've been told it's some fancy proprietary NDA protected kind of thing that I'd get in trouble if I shared any info on.

If it's a digital signal to the transceiver, what about using something like a Schmitt trigger to 'clean up' the signal?

This sounds like a good idea. I'll give it a try if I can't get the op-amp buffer to work.

 

Couple of points. First, the 1uF cap on the -5V power is reversed, the positive lead needs to go to ground, not the -5V. I also would try a .01uF on those power leads to since they will help at the higher frequencies much better. You mentioned the noise was higher on the proto board. Is this a solid clad type of perf board by chance? If you have large areas of copper like that, be sure you have a good connection to the return path, (Ground) on the power supply and make sure anything tied to GND is solid all through your circuits. Ground noise is a possibility here since you didn't have it before. Also, where is and how long is your scope probe ground lead? Try to keep that as short as you can and connect it to ground as close to the circuit part you are measuring, (Input, output, etc.) as you can to reduce the loop for the return and keep noise to a minimum. I suspect that on your breadboard, you had a better ground return and thus less noise or you have an open ground on the proto board, that much difference like an open ground issue, just check your wiring closely to see. As you found out, the tantalum caps didn't really help on your supply rails, so I wouldn't worry about that any more. That 10K you added in the feedback line should be a gain of 1 if you had the same 10K on the input, but any difference in the values could give you a gain of more or less than 1. Rf/Rn = Gain. As I mentioned, make sure your ground between the power supply, Transceiver, OP-Amp, and uC are all good solid connections to reduce any noise issues. Let us know if any of this helps some. Good Luck.

 

Couple of points. First, the 1uF cap on the -5V power is reversed, the positive lead needs to go to ground, not the -5V. I also would try a .01uF on those power leads to since they will help at the higher frequencies much better. You mentioned the noise was higher on the proto board. Is this a solid clad type of perf board by chance? If you have large areas of copper like that, be sure you have a good connection to the return path, (Ground) on the power supply and make sure anything tied to GND is solid all through your circuits. Ground noise is a possibility here since you didn't have it before. Also, where is and how long is your scope probe ground lead? Try to keep that as short as you can and connect it to ground as close to the circuit part you are measuring, (Input, output, etc.) as you can to reduce the loop for the return and keep noise to a minimum. I suspect that on your breadboard, you had a better ground return and thus less noise or you have an open ground on the proto board, that much difference like an open ground issue, just check your wiring closely to see. As you found out, the tantalum caps didn't really help on your supply rails, so I wouldn't worry about that any more. That 10K you added in the feedback line should be a gain of 1 if you had the same 10K on the input, but any difference in the values could give you a gain of more or less than 1. Rf/Rn = Gain. As I mentioned, make sure your ground between the power supply, Transceiver, OP-Amp, and uC are all good solid connections to reduce any noise issues. Let us know if any of this helps some. Good Luck.

Really appreciate these thoughtful responses. It looks like my problem is with the perfboard, at least that's my judgement based on the following observation: if I touch one of the resistors with my thumb, the noise goes away and the circuit works perfectly fine. I've check my grounding everywhere many times now and can't find any other faults, have even replaced some of the components, same problem persists.

Has anyone run into this problem with perfboards? Is there a favoured type for this sort of thing?

Here's a picture of the type of board I'm using now...

 

So your breadboard looks OK but the long lead wires could be the reason for the noise. Think of them as antennas, you can broadcast with an antenna but they also receive signals, in this case, receiving noise. By touching the circuit, you are a much larger "Ground" or "capacitor" to reduce the noise. At least in my opinion. That said, since the voltages of your signals are only 3.3V you need to remember that in the days of 5V logic, a 300mV or 400mV level of noise on a signal line was well below the switching margins of the devices. If you lower that logic level to 3.3V, that same noise value is into the switching levels of these devices, (OK, I didn't look up the switching levels so I could be and most likely am wrong, but the concept is the same.) so the noise levels need to be reduced even more. True that noise levels would be reduced some because the signal levels are lower, but you can still get more noise into things and cause issues, as you may be seeing. As for the Perf board you're using goes, I think you may still need better grounding or much shorter lead wires. Is there a way you can mount this circuit on the same board as the Transceiver or make the connections to and from this board no more than maybe 3" max? You mentioned a coax cable feeding the transceiver from the uC if I remember correctly, is the shield of that cable tied to ground on both ends or just 1 end. It should be at both if you are using the same power for all your IC's and that may help some. From the picture, it "looks" like the coax is coming in on the left and has wires exiting the shrink tubing? You could also try twisting leads like the power and ground, and the signal leads with a ground wire if you can't shorten them. That would at least keep a closer return path for the signals and help keep noise lower maybe. Last question is regarding the OP-Amp, is it a dual part or single? If it is dual, you may want to connect the spare as a voltage follower and tie the + to ground or +3.3V so it doesn't float. That really shouldn't hurt, but if it's causing some of that noise, that may stop it too. Good Luck again. Let us know how it goes.

 

For the record: I worked on this on an off for a while but I finally got it the other day. I added a 1M resistor from the input to ground, that took care of all the noise. Here's the final schematic:


Thanks a lot for the input, everyone. This has been a HIGHLY educational little project.

 

I'm guessing that maybe your input to this circuit from the uC was the issue and that you may have needed to configure the IO pin of the uC as a driven signal to a high or low and not allow a hiZ state if that was an option. Glad it worked out for you. I learn something new all the time and that's the reason I'm not an Expert...

 

SO... I spoke too soon. The noise is BACK. Thing is, it completely disappears when I put my finger on the negative input of the op-amp. Any pointers on how to recreate that effect in the circuit itself?

I've tried adding small caps from the negative input to ground (5pF), no real effect. I've also tried adding a 1M resistor from the negative input to ground. Still nothing.

No explanation why the noise was gone when I tested it last week.

Mysteries abound!

 

Hummm, Lucky You is seems. Well you might try and remove that feedback cap (100pF) and see if the noise you are getting is being coupled back to the input and maybe causing that issue? Is the noise repetitive at some regular level or totally random and how much, as in level? I'm not sure that cap is needed, but you might see if removing it helps. I'm running out of ideas here... :-/ Good luck and let us know what happens.

 

Here is the thing about noise. There is always a source and a coupling mechanism.
Let us consider some possibilities:

1. The coax shield is grounded at both ends to grounds with different potentials
2. The noise is entering the parts via the +Vcc,, -Vcc or ground leads
3. Large value resistors are a little appreciated source of noise, especially at elevated temperatures. See this --> https://en.wikipedia.org/wiki/Johnson–Nyquist_noise
4. LM318 is a cranky part with stability problems. Unity gain is in the middle of the 40M band @ 7+ MHz.

Identify the source and the coupling mechanism and it will lead you to the proper solution. As you have no doubt figured out, "hnaging components" is a hit or miss proposition. Guys I know have been fired for taking that approach.

BTW - I'm curious about the alleged 50 Ω pin on the processor. Is that on the die or because somebody added a series resistor.

 

Really appreciate the replies.

Got it working again. Gave up on the LM318 though, as I probably should have done a while ago. Now using an AD8055, works MUCH better. Using 0.1, .01, and .0022 (I couldn't find any .001's) uF caps on the power rails. No resistors or caps on the inputs or feedback loops. Here's my circuit now:



VERY basic. The thing, I've still got some ringing on the tailing edge of the output signal. It looks like this (my output signal is the bottom line, input is the middle line):



As always, any advice would be greatly appreciated.

Also...

BTW - I'm curious about the alleged 50 Ω pin on the processor. Is that on the die or because somebody added a series resistor.

50 Ω is the impedance on the CPU pin based on the datasheet. Maybe I've got the language/concepts twisted around here?

And...

As you have no doubt figured out, "hnaging components" is a hit or miss proposition.

Couldn't agree more. Certainly planning on developing my theory skills over the next few months for this very reason!

 

So I was looking at this and thinking what might be wrong and causing that ringing on the output. I was sure it's not noise and so I pinged one of the EE's I have worked with in the past and, well, I got a ton of info I'll try to pass on here. So here goes. First off, don't use an OP-AMP in digital circuits. OP-Amps are linear devices and Digital is, well digital. Second, he suggested using something like a 74HC series part like the SN74HC14 or it's non inverting equivalent Schmitt Trigger. If you can't find the non-inverting part, connect 2 in the package and invert the inverted signal, net effect no inverter. They would be so fast the propagation delay for your signal would be nothing. You can tie the unused inputs to Ground just to prevent any noise issues. If you want to use an IC, use a comparator like a LM239A which I believe is a quad package which would require a couple of external resistors for hysteresis which will prevent output ringing and oscillations which is what you are seeing with the OP-Amp. Depending on what you might find, I might be able to figure some values for the external resistors if you go the comparator route, the HC part doesn't require anything external other than power and ground and your signals. Let us know what you end up doing.

 

In posts #7, 22 & 24 (maybe others as well) he got advice of using components used in the digital realm. Why he focused on op amps was not explained.

 

VERY basic. The thing, I've still got some ringing on the tailing edge of the output signal. It looks like this (my output signal is the bottom line, input is the middle line):

Although you have a nice set of bypass caps on the rails, the physical location of them is also very important. How far are they from the chip?