Chapter 632: Gao Zhendong who is chasing after me
"According to my experience and some experimental calculations, for a call to be satisfactory, the signal-to-noise ratio should be greater than 26dB. To achieve this effect, combined with the dynamic range of the voice signal of up to 40dB, we can calculate that if uniform quantization and linear coding are used, the code length will reach 11 bits to meet the requirement!"

11 bits may not seem like too many, but in a time-division multiplexing system, this is unacceptable because the allocated time slots are short, and it is difficult to transmit such a long code.

Moreover, for the codec circuit, an 11-bit code length means 2048 levels! This is really difficult for the current circuit.

At this point, everyone who thought it was simple just now realized the seriousness of the problem. The theory seemed simple, but once it was put into practice, they found that there was a reason why they had no idea in this area.

The theory is simple, but the practical requirements are very complicated.

"Huh? This is troublesome..." Someone bit the pen.

To divide the data into 2048 levels, the requirements for AD and DA circuits with this kind of precision are a bit too high for now. Even decades later, there was a time when 10-bit AD could be considered a high-precision AD device, and 8-bit AD was commonly used.

Now, even the comrades from the Military Communications Institute, who did not consider time division multiplexing for the time being and therefore did not have high requirements for code length, could not laugh anymore. Don’t talk about whether the code length is long, first talk about whether you can come up with this code!

After thinking for a long time, he still couldn't figure it out: "Chief Engineer Gao, how do we solve this problem?"

Gao Zhendong smiled and said, "Just now I said that under the condition of uniform quantization, the situation is like this..."

One comrade understood a little: "Chief Engineer Gao, do you mean that non-uniform quantization can solve the problem?"

Gao Zhendong nodded.

But how to do this non-uniform quantization? Or in other words, what are the benefits of non-uniform quantization? How this benefit comes about, I don’t know at all. Without knowing this, I naturally can’t think about how to operate this method.

Gao Zhendong said: "In fact, the signal-to-noise ratio is also segmented in the signal. For small signals, the signal-to-noise ratio will be low under the same noise, but for large signals with the same noise, the signal-to-noise ratio will be much larger. You may wish to think about it in this direction."

"Signal-to-noise ratio... the same noise... yes, whether it is white noise or quantization noise, it basically does not deviate too much within a certain range... non-uniform quantization... Chief Engineer Gao, what you mean is that for the small signal segment, give it more levels, and for the large signal segment, give it fewer levels. Although this is not uniform, according to your theoretical calculation just now, the signal-to-noise ratio of the quantization noise in the small signal segment will increase. The large signal is not affected much, so just ignore it!"

A comrade from the Military Communications Institute thought for a long time and seemed to have figured it out. He expressed his ideas excitedly.

His words were very difficult to understand, but most comrades still understood them.

Before Gao Zhendong could say anything, the other comrades started shouting.

"Yes, we will take care of the section where the situation is worse."

"This is considered caring for the elderly, the weak, the sick and the disabled. It seems that doing good deeds really makes sense."

"This is a great idea! I really can't think of it."

"..."

Gao Zhendong nodded and said, "Yes, that's roughly what I mean."

The comrade thought about it and was a little dumbfounded. This would be too difficult for the AD/DA circuit. With uneven sampling and quantization encoding, the AD/DA circuit has to determine which segment the signal is in and then switch to different quantization weights. This circuit sounds like it is not much simpler than an 11-bit AD/DA.

Moreover, this kind of non-uniformity must not be simply divided into two sections, but divided into many sections. Just thinking about the AD circuit makes my scalp tingle!
Seeing everyone worrying about this rather difficult AD/DA circuit, Gao Zhendong smiled.

"Comrades, let's change our conversion ideas. Let's process the signal first, and then let the digital-to-analog circuit do the conversion."

"Deal with it? How?"

"Before the signal enters the AD/DA circuit, it first enters a signal preprocessing circuit. This preprocessing circuit is on the AD side. We call it a compression circuit. Its function is to change the signal. Small signals are amplified, and large signals remain unchanged. Its input-output characteristic curve is an upward convex curve. In this way, combined with ordinary AD, the signal is naturally non-uniformly quantized."

Gao Zhendong drew a picture on the blackboard. With the lines and Gao Zhendong's explanation, everyone understood what it was.

Brilliant idea!
"Great! It frees AD from the work of non-uniformity and leaves the non-uniform pre-processing to the compression circuit, so the complexity of the circuit does not increase too much!"

"This is just like the technique of non-uniform quantization. It breaks out of the original framework and takes a different approach to cleverly achieve the goal!"

"Then I guess the voice reproduction must enter the expansion circuit before entering the DA..."

"What are you thinking? You should enter the expansion circuit after entering the DA. Otherwise, it is still a digital signal. How can you expand it?"

"Oh, yes, yes, I was stupid..."

"..." As everyone discussed, they were impressed by Gao Zhendong's brilliant ideas. No wonder Chief Engineer Gao asked "Is it simple?" It turned out that the real troubles were in these places. However, he solved these troubles very beautifully and at a low cost.

Seeing that they all understood it, Gao Zhendong said, "You can use the formula and rules we just used to calculate the signal-to-noise ratio to calculate how many bits of non-uniform quantization are needed to achieve the effect of 11-bit uniform quantization."

Upon hearing this, all the comrades grabbed paper and pen and began their own calculations according to the formula and calculation process that Gao Zhendong used on the blackboard to calculate uniform quantization.

The calculation is very slow. There are connections and differences between the two. Some comrades who have just come into contact with this aspect have not even found the direction, and they have to slowly explore the substitution and transformation between formulas.

Finally, a comrade shouted out.

"8! Fuck! 8 is enough!"

Although he was a mature and steady old comrade, he still couldn't help but swear. This improvement was really too shocking.

The other comrades were shocked, not because he calculated quickly, in fact he was already slow enough, but because the performance improvement was too exaggerated.

8位和11位之间的区别，可不是8和11之间的区别那么点儿，是8倍的区别！

This means that the 2048 levels that originally had to be processed have now been reduced to 256 levels, which is directly reduced by an order of magnitude!

In this way, whether it is AD or DA, it will be much easier to do.

"Chief Engineer Gao, did I calculate correctly?" This comrade himself was a little unconvinced about his own results. Did he calculate wrongly? He asked Gao Zhendong with a little trepidation.

Gao Zhendong smiled and said, "I calculated it correctly, that's the data! 8 bits are indeed enough for non-uniform quantization."

This is the charm of mathematics. This processing method is actually a mathematical means rather than a physical means. In fact, the entire digital signal processing system is basically the charm of mathematics.

The comrades of the Military Communications Institute are full of confidence! This time they dare to guarantee that they will go back and make the next generation of radio stations completely digital!
The most difficult thing to break through is the principle. Once the principle is broken through, you can find ways to implement it.

But since Gao Zhendong has said so much, he will definitely not let them waste too much time. This is just the principle of encoding. As for how to encode it specifically, they don’t know yet.

Gao Zhendong said: "Comrades, please turn to page XX, A law characteristic 13 fold line code."

This is actually the non-uniform quantization coding method that my country later adopted on PCM equipment, and Gao Zhendong developed it early.

Its compression curve is divided into two sections, both of which are logarithmic functions with A as a constant. This constant A determines the shape and properties of the curve, hence the name A-law.

my country’s A=87.6.

Since a strictly compliant compression curve is difficult to achieve in a circuit, a 13-segment broken line is used to approximate this curve.

Everyone was happy when they heard this. Chief Engineer Gao was a good man who did his best to help others. He even selected the compression curve. They were sure that this compression curve was definitely the most suitable one within a certain period of time.

The knowledge of A-law compression coding is not abstract, but it is a little longer. Gao Zhendong speaks very quickly and the comrades understand it very deeply. They dare to pat their chests and say that with this thing, it will be very easy to design a voice encoding and decoding circuit when they go back.

But they obviously underestimated Gao Zhendong's determination to chase and feed them.

"Comrades, next, let's discuss what I personally think is the most suitable PCM codec circuit at present - the successive feedback encoder and the weighted network decoder!"

The comrades were so happy that they couldn't stop smiling. Hey, look, when we come to Mr. Gao's place, not only are we full, but we can also bring takeout!

This obviously went beyond what Gao Zhendong said at the beginning, that is, to gather everyone together to teach the principles, and the implementation has almost been done. Although it is impossible for Chief Engineer Gao to come up with the specific circuit, but looking at the encoder block diagram in his information, this is not like pouring food directly into the stomach through a tube, at least the food is ready and served on the table.

It seems that Chief Engineer Gao was afraid that his comrades would have difficulty digesting the ideas and waste time, so he simply did it himself!

It took a long time to study these two things because they were very practical. Gao Zhendong broke down all the contents of the two circuits one by one and taught them carefully to his comrades.

His idea was simple. Since he had this knowledge, he would teach it all to his comrades at once, reducing the possibility of them taking detours and saving time as much as possible.

For Gao Zhendong and his ideas, timing is everything.

(End of this chapter)