Most, but not all, hardware-based convolvers - for example, modellers that use IRs for cab sims - can only process IRs of 20 milliseconds in length. A few double that to about 40 milliseconds. When you load a longer IR into such a device, it necessarily gets shortened at some point in the loading process. The best way to reduce the length is by windowing the IR (gradually reduce the level to zero at the limit the processor can convolve). A less desirable technique is to truncate it.
If you are using an IR to simulate a guitar cab, it can sound like the cab only if there are no room reflections in it (and a number of additional conditions are met). There is nothing to be gained by using more than 20 milliseconds, assuming that any leading silence has been removed. The first 20 milliseconds of the IR of a guitar cab contain all the audible details of the cab. Bass cabs are another matter and merit their own discussion for several reasons.
Most IRs of guitar cabs contain room reflections in the first 20 milliseconds of the IR. When you shorten such an IR, what remains does not sound in any way like the room in which the IR was taken. Room reflections in this early time window create additional tonal colorations that are not present in the sound of the cab alone.
Acoustic spaces - even very small ones - produce reflections that take seconds to die out. An IR that is half a second or shorter in length cannot produce a convincing illusion of any acoustic space. I posted a demo here some years ago that definitively bore that out.
From the above, a few conclusions are obvious:
1. To convincingly simulate a guitar cab, a short (~20ms) IR with no room reflections is all you need.
2. To simulate an acoustic environment ("room") a long (more than a second) IR is an absolute necessity.
3. Given that most commercially available cab IRs have room reflections in the first 20 milliseconds, adding length cannot possibly make them sound more like the cab.
Cab IRs of 200ms and 500ms definitely sound different from the same IR shortened to 20ms. They do not sound more like the cab from which they were taken, however, nor can they possibly do a good job of capturing the room in which they were acquired.
Q: So is there a point to have/use/make a 500ms guitar cab impulse since the room information will not be adequately captured? Is it useful for a mic placed 20 feet or more from the cab to simulate distance?
A: Mic distance is important, but not to "simulate distance." The response of the cab is completely different with a mic placed close to the grille than with the mic placed at a realistic playing distance. Practically speaking, two meters is far enough away for guitar cabs. The response of the cab won't change at greater distances than that.
The "time of flight" (the time it takes for sound to travel from the cab to the mic) is always removed from cab IRs. Ergo, greater mic distance does not require a longer IR.
Q: What would explain the sound difference between the 20ms and the 200ms IR?
A: Much greater energy from room reflections. The IR won't sound like the room, but the reflections will be audible. IRs between about 30ms and 1000ms are of little value. If you could capture a longer guitar cab IR with no room reflections - none of the IR sellers can - it wouldn't sound different from the same IR shortened to 20ms.
Q: what is the suggested length for bass cab IRs?
A: There's no easy answer to that. The length of an IR determines how much frequency detail can be represented. At the low end, a cone speaker in a cab is a highpass filter. A 20-ms IR can capture highpass behaviour that would affect the sound of a speaker used for guitar, but it might (I stress might) not capture highpass behaviour that would affect the sound of a bass guitar. Many bass speakers cut off at 80Hz or higher and would be well-represented in a 20-ms IR. It is possible that some bass cabs cut off lower - the fundamental of low E on a bass is 41Hz - and would require a longer IR to capture their highpass response. How great an audible effect the added length would actually have is an open question. It's not a subject I've explored in detail yet.
Every system with an input and an output that processes signals has a characteristic relationship between stimulus and response. In the frequency domain, that relationship is called the transfer function. It consists of two components: amplitude and phase. Amplitude response is often erroneously referred to as "frequency response" but is in reality only one part of a system's frequency response. The other, equally important, part is phase. In the time domain, the relationship is called the impulse response.
It is important to understand that both frequency- and time-domain expressions of the stimulus/response relationship contain exactly the same information expressed in different forms (time and frequency domains contain the exact same information. An analog EQ contains "time information"). They are related through the Fourier or Laplace Transform. As implemented in Digital Signal Processing, a filter can either be an Infinite Impulse Response ("IIR") or a Finite Impulse Response ("FIR") type. IIR filters are commonly used to implement DSP-based representations of simple analog EQ filters. FIR filters can represent the response of an electronic device, a speaker, an acoustic space, or a combination of systems in sequence. FIR filters have a number of useful properties, one of which is the ability to capture extremely complex responses of the sort that loudspeakers and acoustic spaces have.
When used to represent guitar cab responses, the length of an IR (in time, not number of samples) and the conditions of acquisition are critical to its sonic accuracy. In order to have a chance of sounding like the cab it came from - always my goal when I acquire IRs, but not the goal of everyone who takes them - an IR must be long enough to contain all the audible detail in a cab's response, be taken at a position relative to the cab that a player's ears could occupy, and be free of room reflections. The last requirement is very seldom met, and the result is that, when you play through most IRs, you get two sets of room cues: those from the room in which the IR was taken, and those from the room you're playing in.
Equalizer filters have impulse responses, therefore an analog speaker sim is just as much an "IR" as an FIR filter. However, many (but not all) guitar cabs have very detailed and audible time/frequency response behaviours that cannot be duplicated with available analog EQ (or a manageable number of IIR filters) but can be quite accurately captured in an impulse response of length =/> 20 milliseconds.
Guitar cabs produce "time" information, so a "good" IR - IOW, one that accurately captures the sound of the cab - will contain the same "time" (interchangeable with frequency, remember?) information as the cab.
Q: IR's are just very detailed EQ's.
Correct, but that is because "detailed EQs" - e.g., narrowband filters - contain "time" information and can have relatively long impulse responses.