Wildly overshooting current feedback amplifier

[Vilius]

Hi,

I have an experimental task involving photomultiplier tube pulse analysis. The major part of the analog circuitry is to drive 16 parallel high-speed S&H amplifiers (significant current required while not compromising signal speed and waveform even a bit), so that should justify why I did not go with some generic op-amp, even though looking back there were much better choices than AD8000...

I have a clean 0.5-1V asymmetric Gaussian form pulse (rise time is 50 ns, fall time is 500 ns) coming from the LTC6268-10 amplifier. That's the input signal for the AD8000 amplifier I am struggling with. My setup for AD8000 currently is very simple: a non-inverting amplifier with a gain of roughly 3.2. However, the output signal becomes critically distorted (yet still fragments of Gaussian form are somewhat recognizable) with a huge voltage overshoot right around the peak of the supposed Gaussian signal, then there is the exponential voltage decay (bottom part of the attached picture). I am indeed observing an approximately correct voltage amplification factor, but the signal form is not acceptable for me. The setup should be as simple as putting 2 resistors in the feedback path, but after hours of testing, I am clueless about what's happening and causing this problem.

Things I have already tried:

1. Switching x1 and x10 passive oscilloscope probes and playing with the probe compensation trimmer capacitor. Also using a standard 10 cm ground clip as well as a short high-speed ground spring
2. Putting various 15 R to 50 R series dampening resistor values right at the AD8000 output, I tried disconnecting the load completely and measuring AD8000 output directly with no load.
3. Putting various fixed several pF capacitors across my feedback path resistors for intended frequency compensation

Nothing has changed the result considerably, only very mild overshoot dampening or pulse duration broadening, but not a fundamental solution. I also attach my circuit diagram right now as well as a small piece of my layout depiction. Please ignore the strange footprints for feedback resistors. At first I had another idea on my mind regarding the feedback for AD8000, but after observing that nothing actually works, I isolated the problem as much as possible and tried to achieve at least minimal/fundamental operation. I have solid decoupling, my supplies are +5V and GND, of course, a solid ground plane on layer 2. Just like I mentioned, I would not choose AD8000 for this task again, but since this is only a first revision device, I really need to make it work at least marginally well, so any help is much appreciated.

[moffy]

Just three questions:
1. How are you probing the signal.
2. What does the 5V look like at the AD8000 supply pin.
3. What does the gnd look like at the AD8000 gnd pin.

[Kokoriantz]

Add a resistor in series with the negative input, it reduces the OLG.

[sourcecharge]

I have solid decoupling, my supplies are +5V and GND, of course, a solid ground plane on layer 2.

"Do not use ground and power planes under any of the pins of
the AD8000. The mounting pads and the ground or power planes
can form a parasitic capacitance at the amplifiers input. Stray
capacitance on the inverting input and the feedback resistor form
a pole, which degrades the phase margin, leading to instability.
Excessive stray capacitance on the output also forms a pole, which
degrades phase margin."

-ad8000 (rev.D) datasheet.

[MT4S301]

1. inverting input net has too physically long trace for a HF amplifier
2. please try the smallest resistor you can handle in the NFB loop for best signal integrity and minimal excess delay
3. also use the smallest cap you can handle for VCC decouple
4. place GND via as close to device terminal as possible and replace top layer ground traces with large groung plane to minimize equivalent inductance
5. put multiple vias for one device terminal instead of forcing multiple devices to share a single grounding via
6. remove GND plane on 2nd copper layer underneath inverting input pin of high freq OPAMPs

I made a simple AD8000 layout in kicad to illustrate the points. All passives are 0402in sized; drill diameter is 0.3mm.

[mtwieg]

So you have a 5V single supply, and the input pulse ranges from 0.5V-1.0V.

Datasheet states that with a 5V single supply, the input common mode voltage range is 1.5-3.6V. So based on that alone you should not expect proper operation.

Even if the input signal were shifted so it fits in the Vin range, the resulting Vout would not fit within the Vout range (1.1-3.9V for a 5V single supply).

So basically you need dual supply rails for the AD8000.

[MariuszD]

On a rectangular waveform, overshoot and ringing can be easly assessed. On a Gaussian waveform, it is difficult to assess.

See how it looks in the datasheet.

[James Anderson]

The root cause of your issue is that the AD8000 is a Current Feedback Amplifier (CFA), not a standard Voltage Feedback Amplifier (VFA):Feedback Resistor ($R_f$) is too small: In CFAs, $R_f$ dictates loop stability. If $R_f$ is too small, loop gain spikes and causes severe ringing/overshoot. Increase $R_f$ to $499\,\Omega - 681\,\Omega$ and adjust $R_g$ accordingly for your $3.2\times$ gain (around $226\,\Omega - 309\,\Omega$).Feedback capacitors are fatal: Never place a capacitor across $R_f$ on a CFA. Unlike VFAs where this helps frequency compensation, on a CFA it destroys phase margin and causes extreme overshoot/oscillation. Remove it immediately.Capacitive loading & Isolation: Driving 16 sample-and-hold inputs presents a heavy capacitive load. You must place a $33\,\Omega - 50\,\Omega$ series isolation resistor as close to the AD8000 output pin as possible.Summary: Remove the feedback capacitor, increase $R_f$ to $\sim 499\,\Omega$, and add a series isolation resistor at the output. Your pulse waveform should clean up right away.

[EE_Audio]

Check the input voltage range for the amp.

[booscrawl]

The root cause of your issue is that the AD8000 is a Current Feedback Amplifier (CFA), not a standard Voltage Feedback Amplifier (VFA):Feedback Resistor ($R_f$) is too small: In CFAs, $R_f$ dictates loop stability. If $R_f$ is too small, loop gain spikes and causes severe ringing/overshoot. Increase $R_f$ to $499\,\Omega - 681\,\Omega$ and adjust $R_g$ accordingly for your $3.2\times$ gain (around $226\,\Omega - 309\,\Omega$).Feedback capacitors are fatal: Never place a capacitor across $R_f$ on a CFA. Unlike VFAs where this helps frequency compensation, on a CFA it destroys phase margin and causes extreme overshoot/oscillation. Remove it immediately.Capacitive loading & Isolation: Driving 16 sample-and-hold inputs presents a heavy capacitive load. You must place a $33\,\Omega - 50\,\Omega$ series isolation resistor as close to the AD8000 output pin as possible.Summary: Remove the feedback capacitor, increase $R_f$ to $\sim 499\,\Omega$, and add a series isolation resistor at the output. Your pulse waveform should clean up right away.

Unformatted unattributed AI slop is rude.