Probably we can never know. The pronunciation of Chinese has changed drastically but the writing system has remained relatively constant. But it's fortunate that sound changes are systematic. What we can do is to start from Middle Chinese and construct the phonology of Old Chinese while systematically fitting the Classical/Pre-Classical rhymed texts, the pictophonetic system, the phonology of modern Chinese dialects, as well as the general patterns of Sino-Tibetan languages.

How good is the pictophonetic system is really hard to say. They are not done perfectly; some pronunciations were the same, while others were close but different in Old Chinese. The difference can lie in the onset or the rhyme, or both. For example, the voiced, unvoiced unaspirated, unvoiced aspirated onsets in the same group or close groups can be used in one pictophonetic pair.

Some of the characters that have different pronunciations in Standard Mandarin have the same pronunciation in the common tongue of Middle Chinese, while others were still pronounced differently in MC but showed clearer patterns than in Standard Mandarin. In general, native speakers now don't understand the pictophonetic system and will have the same doubts as you have unless we study 音韻學. As for dialectal differences, southern dialects in general keep more ancient features and thus fit the pictophonetic characters better. Speakers from the south may have easier time understanding how the pictophonetic system works.

For reference, I post the pictophonetic patterns of your examples below. Even for native ears they sound different. When kids are learning characters at school, the pictophonetic patterns are usually not instructed. The map of pronunciation to glyph is taught character by character. But we are indeed aware of the pictophonetic concept. When encountering an unlearned character, we pronounce the component or a similar character containing that component, which is of course often wrong.

Voiced vs. Unvoiced

婆 and 波 are the easiest pair. In MC, they only differ in the voicing of the onset: 婆 [b], 波 [p]. At that time tones 1 and 2 are not differentiated. They are the same tone of 平聲. The modern mapping rule is called 平分陰陽, that is 平聲 with a voiced onset turns to 陽平 tone 2, and 平聲 with an unvoiced onset turns to 陰平 tone 1. The dialects that haven't gone through this change thus keeping the voiced vs. unvoiced are Wu in general and some subdialects of Xiang.

Unaspirated vs. aspirated

The difference in the pair 青 and 精 is aspiration. 青 is aspirated while 精 is unaspirated.

照三歸端

童 and 徸 have very different onset in MC. But evidence shows that 照三歸端 from MC back to OC. The onset of 童 is 定 which belongs to the group of 端. The onset of 徸 is 昌 which belongs to the group of 章（照三）. In OC, the two groups are pronounced very closely, but not the same. Another difference is that 定 is voiced, while 昌 is aspirated and unvoiced.

偷 and 輸 are the most difficult pair among your examples. Their onsets are also different in MC. But they also follow 照三歸端 from MC to OC. The onset of 偷 is 透 which belongs to the group of 端. The onset of 輸 is 書 which belongs to the group of 章（照三）. So they belong to two groups that were pronounced very closely in OC. But 透 is aspirated and unvoiced, while 書 is unvoiced.

偷 and 輸 also have different rhymes.The views of categorization of 上古韻部 are more diverse. 偷 is under rhyme 侯 and 輸 under 虞. They belong to the same or close 上古韻部 depending on different scholarly views.

Probably we can never know. The pronunciation of Chinese has changed drastically but the writing system has remained relatively constant. But it's fortunate that sound changes are systematic. What we can do is to start from Middle Chinese and construct the phonology of Old Chinese while systematically fitting the Classical/Pre-Classical texts (rhymed, 重文、異文, pronunciation notes), the pictophonetic system, the phonology of modern Chinese dialects, transliteration material from phonogram languages, as well as the general patterns of Sino-Tibetan languages.

How good is the pictophonetic system is really hard to say. They are not done perfectly; some pronunciations were the same, while others were close but different in Old Chinese. The difference can lie in the onset or the rhyme, or both. For example, the voiced, unvoiced unaspirated, unvoiced aspirated onsets in the same group or close groups can be used in one pictophonetic pair.

Some of the characters that have different pronunciations in Standard Mandarin have the same pronunciation in the common tongue of Middle Chinese, while others were still pronounced differently in MC but showed clearer patterns than in Standard Mandarin. In general, native speakers now don't understand the pictophonetic system and will have the same doubts as you have unless we study 音韻學. As for dialectal differences, southern dialects in general keep more ancient features and thus fit the pictophonetic characters better. Speakers from the south may have easier time understanding how the pictophonetic system works.

For reference, I post the pictophonetic patterns of your examples below. Even for native ears they sound different. When kids are learning characters at school, the pictophonetic patterns are usually not instructed. The map of pronunciation to glyph is taught character by character. But we are indeed aware of the pictophonetic concept. When encountering an unlearned character, we pronounce the component or a similar character containing that component, which is of course often wrong.

Voiced vs. Unvoiced

婆 and 波 are the easiest pair. In MC, they only differ in the voicing of the onset: 婆 [b], 波 [p]. At that time tones 1 and 2 are not differentiated. They are the same tone of 平聲. The modern mapping rule is called 平分陰陽, that is 平聲 with a voiced onset turns to 陽平 tone 2, and 平聲 with an unvoiced onset turns to 陰平 tone 1. The dialects that haven't gone through this change thus keeping the voiced vs. unvoiced are Wu in general and some subdialects of Xiang.

Unaspirated vs. aspirated

The difference in the pair 青 and 精 is aspiration. 青 is aspirated while 精 is unaspirated.

照三歸端

童 and 徸 have very different onset in MC. But evidence shows that 照三歸端 from MC back to OC. The onset of 童 is 定 which belongs to the group of 端. The onset of 徸 is 昌 which belongs to the group of 章（照三）. In OC, the two groups are pronounced very closely, but not the same. Another difference is that 定 is voiced, while 昌 is aspirated and unvoiced.

偷 and 輸 are the most difficult pair among your examples. Their onsets are also different in MC. But they also follow 照三歸端 from MC to OC. The onset of 偷 is 透 which belongs to the group of 端. The onset of 輸 is 書 which belongs to the group of 章（照三）. So they belong to two groups that were pronounced very closely in OC. But 透 is aspirated and unvoiced, while 書 is unvoiced.

偷 and 輸 also have different rhymes.The views of categorization of 上古韻部 are more diverse. 偷 is under rhyme 侯 and 輸 under 虞. They belong to the same or close 上古韻部 depending on different scholarly views.

Probably we can never know. What we can do is to start from Middle Chinese and construct the phonology of Old Chinese while systematically fitting the Classical/Pre-Classical rhymed texts, the pictophonetic system, the phonology of modern Chinese dialects, as well as the general patterns of Sino-Tibetan languages.

How good is the pictophonetic system is really hard to say. They are not done perfectly; some pronunciations were the same, while others were close but different in Old Chinese. The difference can lie in the onset or the rhyme, or both. For example, the voiced, unvoiced unaspirated, unvoiced aspirated onsets in the same group or close groups can be used in one pictophonetic pair.

Some of the characters that have different pronunciations in Standard Mandarin have the same pronunciation in the common tongue of Middle Chinese, while others were still pronounced differently in MC but showed clearer patterns than in Standard Mandarin.

  As for dialectal differences, southern dialects in general keep more ancient features and thus fit the pictophonetic characters better.

For reference, I post the pictophonetic patterns of your examples below. Even for native ears they sound different. When kids are learning characters at school, the pictophonetic patterns are usually not instructed. The map of pronunciation to glyph is taught character by character. But we are indeed aware of the pictophonetic concept. When encountering an unlearned character, we pronounce the component or a similar character containing that component, which is of course often wrong.

Voiced vs. Unvoiced

婆 and 波 are the easiest pair. In MC, they only differ in the voicing of the onset: 婆 [b], 波 [p]. At that time tones 1 and 2 are not differentiated. They are the same tone of 平聲. The modern mapping rule is called 平分陰陽, that is 平聲 with a voiced onset turns to 陽平 tone 2, and 平聲 with an unvoiced onset turns to 陰平 tone 1. The dialects that haven't gone through this change thus keeping the voiced vs. unvoiced are Wu in general and some subdialects of Xiang.

Unaspirated vs. aspirated

The difference in the pair 青 and 精 is aspiration. 青 is aspirated while 精 is unaspirated.

照三歸端

童 and 徸 have very different onset in MC. But evidence shows that 照三歸端. The onset of 童 is 定 which belongs to the group of 端. The onset of 徸 is 昌 which belongs to the group of 章（照三）. In OC, the two groups are pronounced very closely, but the same. Another difference is that 定 is voiced, while 昌 is aspirated and unvoiced.

偷 and 輸 are the most difficult pair among your examples. Their onsets are also different in MC. But they also follow 照三歸端 from OC to MC. The onset of 偷 is 透 which belongs to the group of 端. The onset of 輸 is 書 which belongs to the group of 章（照三）. So they belong to two groups that were pronounced very closely in OC. But 透 is aspirated and unvoiced, while 書 is unvoiced.

偷 and 輸 also have different rhymes.The views of categorization of 上古韻部 are more diverse. 偷 is under rhyme 侯 and 輸 under 虞. They belong to the same or close 上古韻部 depending on different scholarly views.

Probably we can never know. The pronunciation of Chinese has changed drastically but the writing system has remained relatively constant. But it's fortunate that sound changes are systematic. What we can do is to start from Middle Chinese and construct the phonology of Old Chinese while systematically fitting the Classical/Pre-Classical rhymed texts, the pictophonetic system, the phonology of modern Chinese dialects, as well as the general patterns of Sino-Tibetan languages.

How good is the pictophonetic system is really hard to say. They are not done perfectly; some pronunciations were the same, while others were close but different in Old Chinese. The difference can lie in the onset or the rhyme, or both. For example, the voiced, unvoiced unaspirated, unvoiced aspirated onsets in the same group or close groups can be used in one pictophonetic pair.

Some of the characters that have different pronunciations in Standard Mandarin have the same pronunciation in the common tongue of Middle Chinese, while others were still pronounced differently in MC but showed clearer patterns than in Standard Mandarin. In general, native speakers now don't understand the pictophonetic system and will have the same doubts as you have unless we study 音韻學. As for dialectal differences, southern dialects in general keep more ancient features and thus fit the pictophonetic characters better. Speakers from the south may have easier time understanding how the pictophonetic system works.

For reference, I post the pictophonetic patterns of your examples below. Even for native ears they sound different. When kids are learning characters at school, the pictophonetic patterns are usually not instructed. The map of pronunciation to glyph is taught character by character. But we are indeed aware of the pictophonetic concept. When encountering an unlearned character, we pronounce the component or a similar character containing that component, which is of course often wrong.

Voiced vs. Unvoiced

婆 and 波 are the easiest pair. In MC, they only differ in the voicing of the onset: 婆 [b], 波 [p]. At that time tones 1 and 2 are not differentiated. They are the same tone of 平聲. The modern mapping rule is called 平分陰陽, that is 平聲 with a voiced onset turns to 陽平 tone 2, and 平聲 with an unvoiced onset turns to 陰平 tone 1. The dialects that haven't gone through this change thus keeping the voiced vs. unvoiced are Wu in general and some subdialects of Xiang.

Unaspirated vs. aspirated

The difference in the pair 青 and 精 is aspiration. 青 is aspirated while 精 is unaspirated.

照三歸端

童 and 徸 have very different onset in MC. But evidence shows that 照三歸端 from MC back to OC. The onset of 童 is 定 which belongs to the group of 端. The onset of 徸 is 昌 which belongs to the group of 章（照三）. In OC, the two groups are pronounced very closely, but not the same. Another difference is that 定 is voiced, while 昌 is aspirated and unvoiced.

偷 and 輸 are the most difficult pair among your examples. Their onsets are also different in MC. But they also follow 照三歸端 from MC to OC. The onset of 偷 is 透 which belongs to the group of 端. The onset of 輸 is 書 which belongs to the group of 章（照三）. So they belong to two groups that were pronounced very closely in OC. But 透 is aspirated and unvoiced, while 書 is unvoiced.

偷 and 輸 also have different rhymes.The views of categorization of 上古韻部 are more diverse. 偷 is under rhyme 侯 and 輸 under 虞. They belong to the same or close 上古韻部 depending on different scholarly views.