Maximum and minimum quantizers don't always work

[Katie Boundary]

I recently attempted an XviD encode in which I-frames were locked to both a maximum and minimum quantizer of 2, and P-frames were locked to 4.

The encoder accepted only some of my suggestions.

Is this a known bug? Is there a plan to fix it?

[blob2500]

If Adaptive Quantization/VAQ is activated, q-max goes slightly outside the set range (example: 3 -> 5; 5 -> 7). I think it's normal.

Xvid window status has an issue(?) and behaves as if VAQ is not activated and reports max values exactly as set.

In all cases, q-min/max values for I-frames can only be set for 2-pass encoding.

[Katie Boundary]

Originally Posted by blob2500

If Adaptive Quantization/VAQ is activated

I don't use adaptive quantization. I got tired of always having to look up whether lumi masking or variance masking was the "correct" one to use.

[blob2500]

Without adaptive quantization:

1 pass (quantization or "CBR"): only the limits set for P and B frames are respected. This is not a bug, it is a behavior expected by Xvid.

2 pass: limits are respected for all types of I/P/B frames.

With version 1.2.x (certainly with 1.2.2 that I used for many years of 2-pass encoding) it also happened to me that occasionally in the encoding of a film one and only one I-frame still arrived at Q=1, even though I had set the I-qmin to 2. Typically it was also the first frame of the video. This yes, I think was a small bug.

With versions 1.3.4 + it never happened to me again.

Note also that B-frame Quantizer ratio and Quantizer offset have priority over the limits imposed in the quantizer tab. Vice versa, in 2-pass mode, between I-boost (= I/P ratio) and I qmin, the latter prevails.
Example, 1 pass/Constant quantizer encoding: if you set everything to "1" except "quantizer offset" which you set to 3, in the encoded stream you will have B frames with Q = 4.