Reduce MAX4477 amplifier noise

[csholmq]

Hi Peter. Great project and very interesting to read about!

I was looking at the amplifier schematic and tried to find some suggestions to improve SNR, but I wound up with as many questions as suggestions I'm afraid. Here's my trail of thoughts;

All in all, we want to keep the noise levels down. So we need to know what type of signal we have and what type of noise. From the MAX4477 datasheet; "The amplifier’s input-referred noise-voltage density is dominated by flicker noise at lower frequencies, and by thermal noise at higher frequencies." From your blog posts, it looks like the pulse width seemed to average around 2 us. Or ~200 kHz in bandwidth. In other words, I would characterise the signal as high frequency in relation to the op amp.

The datasheet then states; "Because the thermal noise contribution is affected by the parallel combination of the feedback resistive network (R_F || R_G, Figure 1), these resistors should be reduced in cases where the system bandwidth is large and thermal noise is dominant. This noise contribution factor decreases, however, with increasing gain settings." Swapping R_F for 3k3R and R_G for 100R could possibly halve the noise density [nV/sqrt(Hz)} while maintaining the gain.

Lastly, looking at the MAX4477 datasheet it appears that each driver can do 30 dB (1000X) up to 200 kHz. Now, if we look at this Linear article it's possible to couple two amplifiers in parallel (doubling the gain, so only 500X per amplifier is needed), thus decreasing the uncorrelated noise by sqrt(2). It also gives us some gain margin which is good for avoiding distortion.

I haven't looked at the LMP7721 yet, but if I find something there I'll make a new post.