High speed op amp circuit problem

[jeremyn]

increasing gain of ad8000 amplifier

Hi,

I'm trying to build a non-inverting gain circuit using one of several possible op-amps (Analog AD8000, National LMH6702, LMH6703).

I am following the manufacturers recommended design, but having the same problem with all 3 chips. My output is dropping sharply and reversing phase at about 45MHz.

I know this is a pretty specific question, but maybe someone here has dealt with the same problem...I am totally new to this and don't have any ideas.

I've attached a schematic of my circuit and a Bode plot to anyone who might be able to help.

Thanks!

-Jeremy

 

[jeremyn]

Re: Op-amp

I'll try to upload the bode plot

 

[Jone]

Re: Op-amp

Study frequency behaviour and stability of opamps and you'll understand what has happened. Amplifiers are only stable up to a finite frequency. They are therefore usually frequency compensated internally or externally. I guess these opamps are internally compensated. Roughly speaking, this means that the gain will approach zero at some frequency where the opamp otherwise could have been unstable.

 

[matrix99]

Re: Op-amp

U are saying that phase reverses at 45M , because AD8000 is a high speed opamp for video apll's may be ur ckt loads are improper and u r getting a fast phse deteriotations at min freqs. Its despends on how much load ur using and cap &R of probe u r using for measurements. See ur set up probes that may have any strong loading effect .

 

[Jone]

Re: Op-amp

Now knowing that these are high speed opamps they should give more than 45 MHz. I agree with matrix99, it is likely that this problem is due to your lab equipment because all three devices behave the same. It is probably the loading that causes this. Make sure your probes are compensated. You might also check the input loading; simply measure that the signal is ok at the input (keep the input connected to the opamp while doing this).

 

[jeremyn]

Re: Op-amp

The input signal is (mostly) ok at 45 MHz. It is a bit distorted, but only very slightly. However, if you continue on to 60 MHz, the input signal distorts and reverses phase similar to the output at 45 MHz.

Again, I am totally new to electronics...I have never done any electronics work before... What do you mean when you say to make sure my probes are compensated?

Thanks for the help,

-Jeremy

 

[arsenal]

it doesnt make sense since ur closed loop bandwidth should be much higher than 650Mhz as mentioned in the spec. so pls check the loading of the probe again and can u tell us what is ur scope and the spec of the probe u a using?

 

[jeremyn]

The output is just a 50 ohm BNC cable attached to an SMA connector that is soldered onto the board. It goes straight into the scope.

What do you mean by check the loading of the probe?

Added after 41 minutes:

Here's the bode plot

 

[jeremyn]

I'm just bumping this up to try to get some help. Maybe someone can give me a reference for a place to look up WHAT IS loading and how does compensation work?

 

[azuther]

You said the input to the oscilloscope comes off the board, through an SMA connector, through a 50-ohm BNC cable and directly to the oscilloscope-is that right? Considering you are measuring a high frequency signal, you should definitely be using an Oscilloscope probe. You said you're very new to electronics, so if you don't know already, oscilloscope probes are essentially specialized wires with high inline resistance and adjustable capacitance which needs to be matched to the scope's input (parasitic) capacitance to get accurate readings at high frequencies. Bottom line-if you're using just a straight coax cable to get the signal to the oscilloscope, try a true oscilloscope probe.

 

[jeremyn]

Azuther,

Thanks for the advice. In fact, I didn't know that! I suppose this is what the other posters meant when they told me to "compensate my probe."

So, I got a 10x oscilloscope probe. Is that what you mean? My scope manual instructed me how to output a square wave and make adjustments to the probe to make sure it looked square. I didn't have to make any adjustments...it looked ok.

Anyway, using this probe to test my circuit didnt help. I had the same problem. Am I doing what you suggested or have I misinterpreted something?

Thanks,
--Jeremy

Added after 1 hours 22 minutes:

So I talked to the op-amp chip manufacturers again, and they suggested that it might be an impedance matching problem.

I am using a 55 ohm resistor at the output. This goes into the 50 ohm BNC cable. I WAS using a 1 M-ohm scope impedance, but now I changed it to a 50 ohm impedance. This DOES get rid of the node at 45 MHz.

However, using a 50 ohm scope impedance, I don't get the full output. It's drawn down by a factor of 2 or 3 since its impedance is no longer much greater than the other output components.

I plugged in a 10x scope probe to try that out. Using this (combined with the 50 ohm scope impedance) seems to work. I don't understand this. Isn't the impedance of the 10x probe pretty high? If we're supposed to match impedances, why is this seeming to work?

On the other hand, using the 10x probe, the phase shift changes a lot. It's already 90 degrees by the time I get out to 45 MHz. Any explanation for that?

Thanks for helping me with this. As you might have been able to tell, this is my first real project involving electronics...at all...so the hand-holding is much appreciated.


--jeremy

 

[melc]

Jer,

You have an RC filter on the OA output, made by the 30 ohms in series with the probe capacitance (usually 10 pF). Using a 50ohm input impedance the frequency range must increase. Using the 1M or 10M input impedance the output frequency limit will decrease. With care, take the 30 ohm out and see what's happening.

NY ?

 

[jeremyn]

I am now using a 50 ohm on the output to match the 50 ohm coax that runs to the scope (which is also now set at 50 ohms).

Do you want me to remove the 50 ohm resistor at the output and replace it with a 0 ohm? Then the impedance won't match the coax cable, which seemed to be my problem to begin with. So, I don't think I understand what you mean.

--jer

Added after 8 minutes:

Here's an updated bode plot showing where I'm at right now. My gain is now flat, so that is good. However, the phase difference between the input and the output is rolling off lower than I would expect.

The engineer at National told me that it has to do with the time delay in the cable between the input and the output. I tried to jot down notes as fast as I could but he kind of lost me. Anyone have any idea what he's talking about?

-jeremy

 

[azuther]

I looked at your most recent bode plot. I believe what you are seeing should be coined "time delay" rather than "phase shift", although the two are basically interchangeable. Looking at the bode plot and doing some estimated calculations of the shift at 10MHz (-15 deg) and at 100MHz (-145MHz), in both instances the output appears approximately 4 picoseconds after the input. If we were dealing with true phase shift, I don't believe the phase shift value would change with changing frequency until the inductance of resistor leads comes into play at extremely high frequencies (on that note, try to keep resistor lead lengths as short as possible, use surface mount components if possible). Anyway, I believe it is a 4ps time delay between the input and output causing the confusion, not phase shift which would be due to inductances and capacitances. This in most likely an unavoidable aspect of the op-amp (the input signal has a lot of complex circuitry to work it's way through before it reaches the ouput). At any rate, your output should be an amplified version of your input, simply delayed/phase shifted. Try repeating the experiment and watch the time delay between the input/output rather than the phase difference.

Added after 13 minutes:

If you have equipment capable of going to frequencies above the 100MHz point shown on your most recent bode plot, try increasing frequencies higher and higher and see if the signals eventually come back into phase with each other. If they will I would expect it to happen at approximately 250MHz. This would verify you are dealing with a time delay of approximately 4ps. (At 250MHz a 4ps time delay would appear as a 360 degree phase shift, making the two signals appear to be identical and in phase)

 

[arsenal]

remove the cable and load resistor, and try another probe without a much higher resistance and lower capacitance.
Its the loading that gives the sharp rolling off and the phase error.

arsenal