i'm very glad to receive some feedback from you. Although i want to comment some of the points you addressed, i would like to start with a short explanation of how the circuit i posted works.
The most important concern for me was the 1MOhm input range. So everything is built arround this. The input stage is done with IC1, which is configured as an inverting amplifier with a gain of G=1/30. This way, all DSO range (+-40V) is supported without modifying the input impedance. Obviously there are many other solutions, but they generally involve the use of manual switches, reles or solid state switches (optocoupled or similar). None of them are suitable here (size, cost, current consumption). Main disadvantege is the small signals are also divided by 30.
Capacitors C5 & C7 are needed to propagate fast signals. Input impedance is so high that even the OA input capacitance would act as a low pass filter. They conform a capacitive divider which propagates fast signals and have to be properly matched with the resistive divider and hence the 647pF (non standard value) value on C5. Whitout these capacities input stage simply won't work for "high frequencies". The advantages over the original circuit are mainly two: there is only a capacitor to tune (C5) and input impedance is 1Mohm for the whole input range.
What is it important for the AO at this point? I considered these factors: offset (static + drift), input current, CMRR, noise (V, I) and GBP (Gain Bandwidth Product) and supply range. Doing some math with the AD8601 values (big numbers - refered to DSO INPUT):
CH_A offset (due to input current): 0.2 pA * 1 MOhm = 0.2 uV
CH_A offset (due to input offset): 80 uV * (967K/33K) = 2.42 mV
There are a couple of components not yet explained in the input stage: IC2 & C1. These two conform the AC/DC input selection. The idea is to have C1 shorted for the DC measurements and open for AC. I think is important to have an AC measurement mode, useful when you have to measure the ripple of a signal, a small signal with a big DC offset, etc. In the original circuit this was done by modifiying non-inverting input in IC5 amplifier with a filtered PWM signal.
The presented circuit won't work. Inverting input is not referenced to any point and the input current will charge C1 either to VCC or GND in case of an AC measurement. And the leakage of IC2 is orders of magnitude greater than IC1 input current, so this is a big no-no. In order to have a working circuit, this capacitor has to be connected between IC1 output and the input of the LTC.
After the input stage comes the amplifier. The input signal has been divided and offseted (also inverted) by IC1 to modify its range to 0..3 V for a -40..40 V input. Now it's time to amplify it a little in order to sample it. At this point, input impedance of the second amplifier has no importance as it's sourced by IC1. The important factors now are offset, input noise, GBP and gains.
First of all, the LTC. I've also considered the PGA113. I have worked with it in a couple of designs and performs very good for the price. But i partially discarded it for a couple of reasons. Offset, drift, noise and multiplexer are better than fine, but GWP is only 8MHz. That means 3dB attenuation @ 380 kHz (page 5 of datasheet) with a 10 kOhm load and G=100. And that is a lot of attenuation. It's true that G=100 would be the 10 mV/div scale, but it is something to account for. Also, VREF input is not a high impedance one so an extra opamp is required to set bias level. In contrast, LTC6912 has 33MHz GWP and with an integrated midsupply (VREF). Obviously, the input impedance is lower, but that is not a point here as the first amplifier is needed in any case. And for the other input characteristics, noise is similar and Vos is higher for the LTC but small enough (125 uV typ.) to not be a concern.
The offset adjust you mention initially puzzled me. The original scope doesn't have an AC measurement mode, so modifying reference to have a pseudo AC mode is a good aproximation... but better than an AC mode itself? I don't think so. In any case, its easy to modify the non-inverting input of the first amplifier to do that (as the original scope do). Using Vref in PGA113 for the same function it's not possible unless you have an input buffer to convert the DSO input range to 0..3 V.
To use this part, close attention will be needed to properly de-couple the power supplies, filter the PWM out from U10 pin PB11 for a clean DC offset voltage (may also require a Buffer Op Amp), and layout and shielding to minimize noise. The inputs should have a 10,000 Ohm or higher resistor in series to limit overload. A 100 kOhm HV resistor would be better (protects to 1000V input) but it may affect the max bandwidth a bit.
If you plan to make PGA inputs the DSO input with a 10K resistor then two problems arise: input impedance is undefined and PGA inputs will function well within supply ranges but input clamp diodes will do its work beyond 0..VCC. If the resistor is 100K then another one is added: input current is 1.5 nA, so input offset will be displaced 0,15 mV. It's very likely that i don't get the whole picture well, so please clarify connections.
Two channel analog input its worth when the two channels are sampled at the same time. Although PGA117 switches very fast (200 ns), it must be commanded to do so, which takes enough time itself to hinder its use in higher (100 ksps) sampling rates. And AFAIK the uC has to ADCs.Another possibility is to use the TI PGA117, which has 10 multiplexed inputs. This would add the ability to use it as a 2 channel analog input with an additional 8 channels of logic. Unfortunately this would require something like a 10 pin 1 mm header and a cut out in the case to access the connector for the 8 Logic Inputs. I suspect that the uC would require additional memory (external) as well. Just a thought.
I don't really know what people want. But if i forget the DSO switch on and the battery gets depleted (3 hour) resulting in a dead battery, then i will be certainly upset. Leaving apart that a single "usual" mistake has costed me 10+ Bucks & 15 days (delivery).Auto Power Off and DC Input. I think most people would not want a simple Auto Power Off, unless it could be adjusted for time and disabled. This may cost more than it is worth. With a good charging circuit, battery life should be good, and when connected to external USB power, it is not an issue. Small USB power supplies are readily available for a few dollars on e-bay and elsewhere. DC input would be difficult to implement, especially with some Overload Protection for the DSO Nano.