Question on LMH6629 datasheet page 28

Hello Samir Cherian.

Thank you for your helpful quick reply.

You are right about overshoots in the transient response of the circuit. I adjusted C1 and it produces very good results (no peaking) . I thought ~20MHz was the "practical" bandwidth of the circuit since that was the point where the amplitude stopped its flatness. Beyond that point the transient response of the circuit was pretty bad.

After C1 adjustment though, the circuit was overall good up until for signals that are 80+MHz fast. Below is a transient response for 150MHz signal. Seems this time there is about 2ns difference at when they both peak (input signal leads). Also for the first few ps, the circuit looks unstable. Do you have any thought on what's possible causing this?

New AC signal Analysis,

Thank you

Hello Festus,

  The time-difference between the signals is because of the limited bandwidth and open loop gain of the amplifier. With regards to the initial non linearity, i am not entirely sure what could be causing that. Couple of thoughts are:

1. Simulation tool trying to find the correct bias point.

2. Insufficient phase-margin in the circuit causing transient issues. This is probably less likely given the flatness of your amplitude response.

A good check for stabiilty/phase margin is to apply a single square pulse with very fast rise/fall time and check the circuits response to it. If there is a lot of overshoot/ringing then this could indicate insufficient phase margin. Can you please check that as well. If the response comes back clean then I think you are okay and can disregard the initiial non-linearity in the sinusoidal response.

-Samir

Samir,

With a 1ps rise/fall time150MHz input signal, there aren't lots of overshoots/undershoots except a tiny overshoot at the beginning (see the graph below). However this affect the output Vpp (decreases) but overall the circuit is stable around 150MHz.. By the way, would this also be applicable (using a fast square input signal) when stability/phase margin at small frequencies though?

Hello,
When analyzing an amplifiers response to a pulse, we do not need a repetitive signal. What I do is to provide the amplifier with a fast (1ps) positive going single pulse and study the response for overshoots. I then provide it with a separate negative going single pulse and study the overshoot. The high freuqnecy content is contained in the fast edges. We are studying the response to these fast edges. I would redo the analysis with the above setup. Hope this was clear.
One more thing to remember is that BW and rise/fall time are inter-related by the equation BW = 0.35/Rise time. So if you input a very fast signal (1ps) into an amplifier with 150MHz BW, the output of the amplifier should have a rise time of 2.33ns.
-Samir

Hi Samir,

Just to confirm what you are saying, I need pretty much a 0 to 1 transition and 1 to 0 transition in 1 period and study the output signal for overshoot/undershoot? See the image below: (I used 1ps 150MHz input signal, the rise time of the output signal(10% - 90%)  turned out to be 2.16ns

I cannot study any overshoot/ringing behavior based on the last results. The reason is that the negative going edge occurs before the complete amplifier response to the positive edge has occurred. Reduce the frequency of your input to 5MHz and rerun the simulation. Because of the low frequency input we will be able to study how the signal settles.

These are the results I am getting when I change the input freq to 5MHz. Does this mean our circuit is not stable at low frequencies?

It doesnt look like there is too much overshoot/undershoot at the rising & falling edge so you should be fine. Generally < 10% overshoot indicates sufficient phase margin. So, in your case your output moves around 1.25V, so you ideally want < 125mV of overshoot or undershoot at the edges.

I have few questions about this though. 1) Ideally I want my output to move around 1.8V which I can get with a sinusoidal input. The amplitude decreases, however, when I use a square input like in that example above. What might be causing this? 2) Can the same conclusion, that there aren't too much overshoot/undershoot, be drawn for higher frequency input signals (like 150MHz input signal) ? 3) And finally, to end this discussion, would you recommend to go ahead and use , for my application - 150MHz -200MHz current to voltage converter, LMH6629...circuit on page 28 of datasheet (with adjusted C1) and think will perform well assuming a good pcb design ?

1) The reason you see an amplitude reduction is because the amplifier does not have the sufficient bandwidth. Remember that the rising edge of 1ps input has a very high frequency component that is filtered out by the amplifier circuit.
2) The analysis that was done will hold for high frequencies.
3) You can use the circuit shown on the datasheet, however with board parasitics and possible variation in photodiode capacitance you may want to fine tune the circuit further in the lab.
-Samir

Samir,

Thank you for your patience with me and your continuous support on the issue. I have a small question that goes like this: If you perform AC signal analysis of the circuit and it shows a flat response over a range of some frequencies [0-150MHz] (see Figure 1 below) and then, you perform a transient analysis on the same circuit using an input sinusoidal signal with a 100MHz frequency and the response shows clipped amplitude (see figure 2 below) , doesn't that itself somehow contradictory since the AC signal analysis previously stated that your amplitude will not be affected ? 

Figure A:

Figure B:

Thank you,

S. Festus-

Hello Festus, If you recollect from DSP class, the spectrum of a fast going pulse/edge contains lots of high frequency components, so even if you apply a 150MHz "Square Wave" its spectrum/FFT will contain a 150MHz component and in addition a lot of high-frequency components(due to the sharp rising/falling edge of the square wave). It is these higher frequency components that the amplifier is attenuating since it only has a BW of 150MHz. If you applied a 150MHz sine wave you would see a faithful representation at the output.

Oh I see. I think I understand why it happened like that. But, let's say you run a test and this happens:

In the image below (fig 1), I have 1ps risetime 130MHz square wave signal as input. I am expecting, based on BW = 0.35/ Rs , to get an output signal with a rise time close to 2.69ns theoretically. My actual output rise time is around  2.6ns which is pretty good on that end, but still the signal looks "triangular shaped" which is way off from the original square wave input we had. Does that suggest insufficient bandwidth of the circuit? or Does that mean I am pushing the bandwidth envelope too far? because If I lower the frequency to 50MHz but keep the same rise time (1ps), the output resembles more the original square wave - a little distorted but still looks better than the triangular shaped one , but also with slower rise time (3.7ns) (fig2)

What would you consider the practical bandwidth of this circuit (The AC signal analysis shows small attenuation at 130MHz)?

Figure 1:

Figure 2:

 

1. For a 150MHz sine wave the time period is 6.7ns. This means half a square wave is 3.35ns.
2. If the BW of the amplifier is 150MHz, then based on the (Rise Time=.35/BW) math, the rise time is 2.33ns. Remember this is only 10% to 90% of the final value. The 10% on each side will take a lot longer than 2.33ns. So before the signal ever reaches the final values the next edge occurs and the direction of swing will reverse, making the resultant output look more triangular.
I would very strongly suggest checking out the (FFT) spectrum for a sqaure wave. It is important to understand that a 150MHz Square wave and 150MHz sine wave have very different dpeactral characteristics.