I can think of several options:
- The hardest one (because it is still WIP) is to remove the PAL ULA and get a NTSC ULA. NTSC ULA's existed for some clones intended for the American market. Today they are very rare chips.
However, there are some projects that aim to design a direct replacement ULA using programable logic. These projects may support NTSC output with no effort. Besides, you would need to change the 4.43 MHz crystal with a 3.57 MHz one, and possibly some passives that belong to the LM1889 circuit.
This option, once such replacement ULAs become available, has the advantage of having everything inside the case, with no extra equipment, so you end up with a true NTSC system, with 60 Hz vertical refresh. However, the main inconvenient is that games and demos expect vertical interrupt to occur every 20ms, or 69888 CPU clock cycles. Some of them rely on these timings to perform various multicolor effects, or achieve flicker-free sprite animation.
As of today, the only ULA replacement readily available is from Retroleum, and it is called Nebula. According to their FAQ they don't have plans for NTSC support, though.
- It's a pity you don't have a SCART input in your TVs, as this tend to be very flexible with video timings, and as there is no modulation, but pure RGB, you actually get colour. In case you might have a TV with SCART inputs, the Spectra interface is able to generate a clean RGB signal from a Spectrum 48K, no matter if it has the composite mod applied or not.
- The remaining options are actually variations of the same idea: to convert the PAL composite video signal into a format understable by equipment available in America. AFAIK, American TVs understand composite video, YPrPb and HDMI, with some of them understanding VGA as well.
These devices perform a full frame upscaling. That is: there is a memory that serves as frame buffer, actually, with enough capacity to hold two frames. The premise is that while the image from the composite video input is being stored in one frame buffer, the other one is being read at the required speed for the output image to be rendered.
The advantage with this approach is that no matter the format or timings of the input image may have, the output image will have timings that will match your equipment. The disadvantage is that vertical refresh frequency won't be the same (specially if the input is PAL, which operates at 50 fields per second), so both frame producer and frame consumer won't be able to be in sync and sometimes, a frame that is being read by the frame consumer to produce the output will be rewritten in the middle of the operation by the frame producer, causing some tearing that may be visible if the screen features full screen animation such as full screen scrolling.
There is this device. It is small, and is able to convert a composite video signal, either NTSC or PAL, into VGA, with the user having several choices about what resolution and refresh the outputted image will have.
Here you can see that I am feeding the converter with a PAL signal from a PAL C64, while the output signal is rated for 75 Hz.
And this is the menu from which you can change the output resolution to better match your monitor's native resolution, if possible.
Analogously, there are converters from composite video signal to YPbPr or HDMI. All of these are upscalers, and hence, will suffer from the tearing effect described above.