LTSpice limitations can bite. The macro models used in simulation don’t sufficiently take into account parasitic and circuit layout effects that can haunt you down the road.
My wideband magnetic loop project uses an remote head end using an op amp – the LMH6629 transimpedance amplifier. You can see that this minimalist design looks good on paper and works well in simulation.
However, actual performance is not stable. In my rush to build this thing, I had not fully understood the LTSpice limitations that have come back to bite me! To date, Jim and I have built two versions of the head end amp. First, a hand crafted Manhattan style board made by Jim, VE6JF. Second, a professionally made circuit board previously described.
In the above graphic, you can see examples of the head end instability from my tests. The top row shows the PCB version at 2 and 10 MHz. At some power supply voltages I get lots of oscillation and distortion. The input is in yellow and output in blue. You can see harmonics of oscillations in the FFT purple trace.
Jim’s hand crafted prototype performance at these frequencies is shown in the bottom row. You can see more stable performance, but amplification drops off dramatically at lower and higher frequencies. Also, if you can take a closer look at the waveforms, you can see they have distortion and are far from proper sine waves.
By varying the supply voltage, I can get these amplifiers to settle down, but performance is disappointing. So, why have things gone wrong?
LTSpice Limitations – Two Reasons for Poor Results
From what I can tell, there are two inter-related reasons for these instabilities. The first is the intrinsic limitations for Spice programs. The second is the need for better construction practices when trying to implement a wideband high gain operational amplifier.
In the real world, performance of an LMH6629 or similar complex amplifiers is strongly effected by circuit board layout and judicious use of bypass capacitors and stray signal mitigation. You will find this especially true when you try to implement the circuit using simple voltage dividers to accomplish single supply operation.
Any amplifier is an oscillator just waiting to start. My head end amplifier design does not have enough bypassing or filtering to mitigate spurious signals. In particular the sensitivity to changes in power supply voltage tell me that the biasing is not stable. I seem to have a lot of common mode noise floating around my lab, and it is getting amplified and fed back into the system biasing.
The intrinsic LTSpice limitations are based on its use of macro models of device performance which do not take into account the micro effects of things like parasitic effects in the circuit design and layout. A proper engineer would probably have ways to extend the models to take these effects into account. I am not a “proper engineer”, just a hobbyist who got out of his depth.
Good learning experience, though. In particular, I have learned the importance of testing on a breadboard regardless of how good the circuit looks in simulation.