I have moved this from the details page because it became too large.

I expect the amplifier to draw a milliampere or more. This is not comparable to the DS32KHZ and a different solution is needed because the charge pump will not work at all.

The first question that must be answered is : how much power can the test point provide ? I don't think that there is a standardised design, my cheap scope provides 2V at 1KHz but it doesn't need to be precise because all it has to do is provide clean edges.

This lack of standardisation implies that the DC/DC design must be flexible and accomodate different frequencies, voltages and current ratings.

There are 3 usual methods to step-up :

• Charge pumps: already tried, accomodates low currents, simple, cheap but limited.
• Single-coil step-up converters: some circuits can work down from very low voltage, like a used AA battery, and increase the voltage to any desired level. There is one problem though: they draw more current as the input drops. The quiescent current is not negligible (some new chips get better though) but basicly, the big drawback is that it hammers the source when it gets weaker... leading to a catastrophic drop.
  Integrated solutions might also have startup issues: the oscillation must start VERY fast because there are only 500µs per cycle to gather energy. The circuits I know have slow ramp-up time, like 10ms... This totally shatters any claim of high efficiency.
• Resonant, double-coil converters: So-called "Joule thief", these circuits are usually composed of a BJT, a transformer and a diode. They have meaningful advantages and drawbacks:
  - Power output depends on all the parameters and parts, including available input energy (output power will decrease with the input)
  - cheap, easy to source, though the transformer is often tricky (single-coil versions are usually preferred)
  - can generate relatively high voltages (but beware of the breakup voltage of the hashing transistor !)
  - with a 3rd winding, can simultaneously generate negative voltages (great for analog amplifers!)
  - startup is immediate (must be tuned and measured though)
  - The transistor can die if there is no load connected to the output (been there, done that)
  - Efficiency is not as good as a perfectly controlled IC (50% to 70% from my decade-old experiments)

From this, it appears that a "modified Joule thief" is the most promising solution. The ability to automatically adjust the input current and the output voltage are essential because they will vary all the time ! It can accomodate your brand and make of oscilloscope.

The step-up DC/DC is one thing but it will only work about half of the time. During the peak hours it will try to convert as much energy as possible. This energy is stored in a big capacitor before being used by a linear regulator.

This is where the design borrows from other established systems, in particular the "active PFC", power factor correction circuits that have to deal with the same kind of problem, when the rectified power from the mains (50Hz sine in Europe, 100Hz half-sines) does not provide energy during a significant fraction of the cycle.

The solution is to step-up as much as possible and store in a large capacitor that will provide the necessary energy during the remaining 500µs.

Some desired properties of the capacitor are:

• Low ESR (series resistance) and high quality, thus it's more expensive than a typical capacitor, or it will fail from internal heating (well, that's what I remember from past projects, we're speaking about some mA not Amperes here). Ringing can also be an issue.
• High voltage tolerant: the step-up can generate high voltage and the capacitor will switch continuously between charging and discharging
• Capacitance: it must be computed but 47µF was not enough in the case of the DS3KHZ. With the step-up, a large ripple is not a problem because the capacitor can be charged at 30V if needed.

So there is a trade-off between capacitance, ripple, voltage etc...

Note: the probe must be differential (plus a common pin) because some current (with nasty oscillations) will go through the ground-connected shield.

A split-system becomes necessary:

• one module (power supply) hooks directly on the scope's BNC socket to reduce the power current-induced interferences.
• 2 wires come out with +/- along the cable of the scope probe. The tip is equiped with the differential amplifier.

So that's why I went on a quest to find BNC connectors...