Argthtjtr

According to

https://www.lawlessenglish.com/learn-english/pronunciation/consonants-voiced-unvoiced/

there are number of pairs of consonant sounds whose only difference is that one is voiced and the other isn't. Some of the examples listed are t vs. d, k vs. g, s vs. z and p vs. b.

If this is true, how come I can tell the difference between 'pet' and 'bet' when they are both whispered?

Or do I trick myself into believing there is a difference and only tell the difference by context?

(1) Understanding "phonology"

Phonology deals with sounds as abstract representations in a mental system. We assume that sounds have a value in this psychological system when they can make a difference in meaning. In this case, we call them "phonemes" and indicate them with slashes. Here we know that /t/ vs. /d/, /k/ vs. /g/, /s/ vs. /z/ and /p/ vs. /b/ are phonemes becomes they cause a difference in meaning in minimal pairs such as tip vs. dip, cot vs. got, Sue vs. zoo and pet vs. bet. We then try to find one single feature to describe abstractly the difference of these phonemes. This is a methodological consideration, an attempt at formalizing the mental system. Here, the single most obvious phonetic correlate for the difference of the phonemes you mention is voice - for /t/, /k/, /s/ and /p/ your vocal folds tend to not vibrate, for /d/, /g/, /z/ and /b/ they do tend to do so. In a phonological description, we reduce systemic difference to a single feature difference if possible.

(2) Other phonetic differences

Phonetics looks at the actual, real, measurable properties of sounds that actually exist, as they come out of your mouth. We call these sounds phones and indicated them in squared brackets. So a specific sound [p]₁ will be different from another specific sound [p]₂ and [p]₃ etc. We can then look at the actual phonetics of a specific [t] and a specific [d] and describe their differences in great detail. In fact, there will be a large amount of physical detail to list concretely phonetic differences. This task would be an empirical analysis, typically an attempt at a comprehensive analysis of sound waves. What we might typically, but by no means always find, is that indeed, [t], [k], [s] and [p] are less sonorous, don't have a clear band of formants, are more "spikey", than [d], [g], [z] and [b], i.e. the former are not voiced and the latter are. We might also find that [t], [k] and [p] extend their voicelessness onto the next sound, producing a little puff of air, called aspiration, whereas [d], [g] and [b] do not. We might also see that [t], [k], [s] and [p] are produced with more pullmonic pressure, harder tightening of the lips (sometimes called the fortis - lenis, or strong - weak distinction) etc. than their counterparts [d], [g], [z] and
[b]. The list goes on and on. In a phonetic description, we try to list all phonetic aspects of a sound but a comprehensive list of differences is hardly possible.

When you can tell the difference between two whispered sounds A and B and you conclude that they likely instantiate the phonemes /t/ vs. /d/, you are probably relying on several of these many, many phonetic cues.

(3) Illustration

Consider the image below. It shows a wave diagram and corresponding spectogram for the sequence [pe] and [ba].

In phonological terms, we would say that /p/ and /b/ are phonemes in English because they make a difference in meaning, and describe the difference as voiceless vs. voiced. This difference is systematic, minimal and useful in teaching.

However, in phonetic terms, we would analyze actual differences between [p] and [b]: [p] looks more like a spike than [b]; [p] has less "color" in the spectogram than [b]; [p] occurs with aspiration [h] before the vowel, but [b] occurs without aspiration before the vowel, etc. But where do you start, where do you stop? How do you make the description systematic? How much detail is needed? And what is this description supposed to be good for?

Comparison of p, t, k with b, d, g when occurring in initial position, medial position or final position of a word will yield different results; as a consequence of this position shift the b, d, g group of plosives can either be fully voiced, partly voiced or voiceless. Thus being voiced or voiceless is not of much importance here.

So what makes us differentiate between these two groups?

There's another category besides voicing, and that is the force of the articulation. The voiceless plosives p, t, k are, therefore ,called fortis(meaning strong), and the b, d, g group of plosives are called lenis(meaning weak). It is said that p, t, k are pronounced with more force than b, d, g; subsequently in the initial position aspiration happens for p, t, k, but not b, d, g so this is the distinguishing factor for the initial position occurrence.

In the final position b, d, g have little voicing, and the plosion following p, t, k is very weak so it's harder to distinguish between them, but nevertheless they have a difference, and that is the shortening effect of p, t, k on the preceding vowels. The medial position has the characteristics of either the initial or final position.

This is loosely based on English Phonetics and Phonology by Peter Roach; I recommend it if you're interested to read further.EDIT: Chapter 4: Voicing and Consonants.