I agree that it should definitely be a bench unit. The build quality of a handheld unit is much more important (it's more likely to be dropped) and harder. It would probably be hard to even match the build quality of cheap $5 multimeters, let alone a Fluke.
I'd caution against trying to design a complete electronics lab at once, start with just a DMM. At least build a prototype before deciding on building a function generator/power supply/oscilloscope/etc.
I don't see the advantage of modular instruments beyond stackable cases. There were some series of modular instruments in the past, eg. Tektronix TM500, 7000 series, Hameg HM8000 (this one in still sold). The motivation was usually:
- Common power supply. Transformers, filter caps and heatsinks for pass transistors are large and expensive. These days, you can buy a switcher for a few dollars, although I'd probably use a linear supply for a precision bench DMM for low-noise.
- Common display. A CRT is also large and expensive. A simple character LCD is fairly cheap. Graphical displays are more expensive and larger.
- Internal connections between instruments. I think this is more important in automated test setups than your average DIY lab. I tend to use my instruments for all kinds of jobs, not in some sort of fixed test setup.
- Expensive microprocessors. An 8-bit, 8-pin micro is about $2 (eg. ATtiny13). An 8-bit, 28-pin micro (ATmega48) is $3. I'm sure there are equivalent PIC/MSP430 devices. I don't see a reason not to put one of these in every module.
I don't think a good DMM would make a good low-speed scope or logic analyzer:
- A multimeter is high-precision and usually single channel
- A low-speed scope (beyond basic data logger) would have to sample at least 200kS/s or so (for up to 20kHz audio range), but doesn't need much accuracy (1%, 8-bit ADC usually)
- A logic analyzer needs at least 8 channels, just 1-bit, and again a fairly high sample rate, even for relatively slow signals like I2C.
You'd just be designing three completely different front-ends, the only common parts would probably be case, a uC, a display and some buttons.
Re Arduino, I don't think there's much point in using an off-the-shelf Arduino board, since the complexity of designing a basic Freeduino circuit into the board is trivial compared to the analog part. This is not like putting a basic relay shield on an Arduino and hooking it up to a lamp. I'd also want to isolate the front-end from the digital part, which probably wouldn't fit very well as Arduino shield (needs separate power supply).
Schematics, here are some service manuals of brand-name bench DMM's (with theory of operations and schematics):
HP/Agilent 3478A (5.5 digit, introduced in early eighties)HP/Agilent 34401A (6.5 digit, introduced in early nineties or so, probably more custom parts)Keithley 196 (6.5 digit, introduced in late eighties, few custom parts)Fluke 8840A (5.5 digit, introduced in early nineties)These are high-precision, low-noise and high impedance designs, which means component selection and PCB layout is probably very critical. They usually feature an input impedance of more than 10 gigaohm at the lower DC voltage ranges. A 1 year accuracy specification for a good 4.5 digit multimeter is 0.03%, for a good 5.5 digit meter 0.006%, and for a good 6.5 digit meter 0.003%. Plus you need to calibrate it to something with better specifications after you've built it. I'm not an expert in high-precision design, but it seems really hard to match this to me. I'd stick to mediocre accuracy/resolution for a bench meter (say 4.5 digits), and focus on price and features compared to commercial units (plus the fun).
Someone mentioned simultaneous voltage and current measurement. Since current measurement is just voltage measurement plus a shunt, I'd make it in a full dual channel unit. The extra design effort is probably almost nothing. The extra component costs are probably also marginal, and it would be very useful (I often use multiple DMM's at the same time).
Alson