Speed Comparisons--MainConcept vs CCE Basic

[wwaag]

I've acquired both the MainConcept and new CCE Basic encoders. Just finished some speed trials and thought I would post the results. For all trials, I used the same video clip--a 2 min DV file I captured from an ADVC-100 using Scenalyzer and saving in a Type 2 format. I used 2 methods--a direct file encode and frameserve using Vdub. The clip was 121 sec in duration. For CBR encodes, I used 6000Kbs. For VBR encodes, I used a mean of 6000 with 2000 min and 8000 max. All audio was encoded at MP2, 224kbs. Here are the results

Direct Encode
MainConcept
1-pass-VBR--117 sec--speed 1.03
1-pass-CBR--112 sec--speed 1.08
CCE
1-pass-CBR--90 sec--speed 1.34
2-pass-VBR--174 sec--speed .70
Frameserve through Vdub
MainConcept
1-pass-VBR--156 sec--speed .78
1-pass-CBR--153 sec--speed .79
CCE
1-pass-CBR--117 sec--speed 1.03
2-pass-VBR--225 sec--speed .54

The only direct comparison is really the 1-pass CBR since MainConcept does not offer 2 pass-VBR and CCE does not offer 1-pass VBR. For CBR 1-pass, CCE took 90 sec vs 112 sec for MainConcept. Overall, I would say that CCE is definitely faster. Any of the speeds greater than 1 indicates better than real-time encoding--bottom line--both are very fast.

Since quality is such a subjective matter, I haven't done any detailed comparisons of the resulting encodes under these various conditions. There are just too many variables. However, I do offer a couple of personal observations. I finished an encode of Star Wars (from laserdisc) using Tempgen. With filtering and only CBR, it took 12 hours--looked very good to me--almost indistinguishable from the laserdisc. Using CCE 2-pass only for a selected segment (without filtering), the resulting encode was almost indistinguishable from the Tempgen encode--basically same quality--lot less time.

For a video clip from a Hi8 source, I compared MainConcept (1 pass VBR)and CCE (2-pass VBR). Both looked very good. However, CCE was a lot better on certain scenes which included shots of carpeting. In MainConcept, the carpet appeared to "move", whereas the CCE encode was very smooth and detailed with no apparent movement. So for home video work, my choice is clearly CCE.

Hope these observations are of some use.

[Lamont Cranston]

I have also run tests comparing CCE Basic mpeg 2 2 pass vbr with MainConcept 1 pass VBR. The test clips were designed to stress the encoder. In all cases the Q (quantanization) levels, as measured with Bitrate, of CCE were consistantly lower than Mainconcept under the same conditions. Usually the lower Q levels indicate a better encoder but that does not always appear the same visually. Visually CCE produced a slightly better looking video than Mainconcept. As with all these type of tests the results are subjective.

- LC

[digitalmaster]

Thanks for the information. I think TMPGEnc better get moving with their speed issue or they will start to lose market share.

[wwaag]

I have also run tests comparing CCE Basic mpeg 2 2 pass vbr with MainConcept 1 pass VBR. - LC

Since you've used MainConcept quite a bit from reading your posts, are there any settings that you would recommend for 1 pass VBR? Unlike CCE, there are tons of options. I've only played around with a few (like line noise filtering, and setting the intrablock DC precision to 10 bits). I've found that none of these changes dramatically alter encode time--or picture quality to my old eyes. Most of my encoding using MainConcept is for satellite captures, which is usually pretty good video from the start, so I usually resort to using the "standard" settings, except for the average bit rate to control the size of the resulting file. There are published "templates" for Tempgen, but none for MainConcept as far as I know. Any suggestions or recommendation you might have based on your experiences would be greatly appreciated. Thanks.

[digitalmaster]

According to some other post, MainConcept employees said the Encoder is a very new program. It has just started catching on in the past couple of months.

[fingernailX]

wwaag:

I too have looked for this info. All I could find was at: http://www.mainconcept.com/forum/viewtopic.php?t=54 scroll down to bottom of page

There are some 'optimized' profiles that come with Vegas Video and the Adobe Premiere 6.5 versions of the encoder but I didn't find a 'magic bullet' in reviewing the text of those.

If you find any other info please share.

[SaSi]

Here are the settings I use with MCE for 1 pass VBR. (I reached these settings after prolonged testing between Tmpgenc (my previous encoder), MCE and trial CCE BE).

Detailed Settings:
(FOR PAL Interlaced Video)
Frame size 720 x 576,
FrameRate 25
Field Encoding = Bottom Field First
Deinterlacing = None
Search Method = 11
Search Range = 5
Enable Line Filtering = unchecked

Advanced Settings
Video Settings:
Aspect Ratio = 16:9 for widescreen movies, 4:3 for 4:3 episodes and movies
GOP Structure:
I Frames = 15
P Frames = 5 (This decreases the number of P frames in favour of B frames)
Closed GOP every = 5 GOPS
Bitrate Type: Variable Bitrate
Rate Control Mode = Mode 1
Maximum = 6000
Average = 4800 (This makes an up to 2 hours movie <4Gb so that the audio AC3 can also fit. Have used down to 3200 for very long movies with equaly excellent picture).
Minimum = 200
Constant Quality = 10

Advanced Video Settings
Profile ID = Main Profile
Noise Sensitivity = 12
Motion Search Mode = 11
User Quantization Matrices = Checked
Do Half Pixel = Checked
Set Motion Search areas = Enabled and 3 h x 3 v

In terms of Quant. Matrix, I use the following for

INTRA BLOCKS NON_INTRA_BLOCKS

8 16 16 22 26 27 29 34 16 18 20 22 24 26 28 30
16 16 16 26 27 29 34 37 18 20 22 24 26 28 30 32
16 16 18 27 29 34 34 38 20 22 24 26 28 30 32 34
22 26 27 31 36 37 38 40 22 24 26 28 30 32 34 36
26 27 29 36 39 38 40 48 24 26 28 30 32 34 36 38
27 29 34 37 38 40 48 58 26 28 30 32 34 36 38 40
29 34 37 38 40 48 58 69 28 30 32 34 36 38 40 42
34 37 38 40 48 58 69 79 30 32 34 36 38 40 42 44

The matrices above are adoptations from the KWAG matrices for Tmpgenc. KWAG also modifies the GOP structure heavily in an undocumented way for Tmpgenc. I find it is not required to do so.

The above settings produce excelent results and encoding in record speeds. My P4/2.57GHz encodes at x1.7 almost an original mpeg stream. This last thing is what makes MCE excelent for me. I rip the Mpeg video stream out of the VOB file (a 4 minute process), feed it to MCE and after 1.5 hours I have a re-encoded movie with excelent picture quality.

[SaSi]

After viewing the matrix tables, I realized I didn't leave enough space between them. Use the first 8 numbers for the Intra block matrix and the next 8 numbers on each row for the non-intra block matrix. Sorry about the messing up.

[wwaag]

SaSi,

Thanks for your quick response. Could you send me an e-mail with one of your *.ces files as an attachment. This would certainly be the easiest way (at least for me--I'm a lousy typist). Thanks.

[digitalmaster]

I agree MainConcept produces great quality vids using KWAGs matrix from www.kvcd.net. The reason for using that matrix is that it helps compress the mpeg file more without losing quality.

[SaSi]

wwaag,

for the benefit of the thread, I copy the profile (Mainconcept profile) in this posting.

MCE stores it's templates at
C:\Documents And Settings\{Login Name}\Application Data\MCMPEGEnc

using files with an extension of ".mef"

Cut the part of the posting below the ----- and paste it into notepad saving the file as DVD_TWEAK.mef

Then start MCE and use this profile.
--------------------
[MAIN]
StreamType=1
MpegType=4
VideoMode=1
AudioMode=2
Width=720
Height=576
SampleRate=48000
LineFiltering=1
Deinterlacing=0
ID=mep-profile
Name=DVD KWAG
[basic]
settingsVersion=2
online=0
perfLevel=16
motionSearchType=23
mpegType=4
deinterlacing_mode=0
relaxed_standards=0
[video]
videoPulldownFlag=0
rateControlMode=1
aspectRatio=3
frameRateCode=3
iFramesInGOP=15
ipFrameDistance=5
bitrateFlag=0
mquantValue=10
constBitrate=6000000
maxBitrate=6000000
avgBitrate=4800000
minBitrate=200000
timeCode0=0
dropFrameTC=0
constrParms=0
closedGOPs=0
[audio]
audioLayer=2
audioMode=2
deEmphasis=0
audioBitrate=11
privateBit=0
copyrightBit=0
originalBit=0
errorProtection=1
psychModel=2
[multiplex]
programStreamType=4
vbrMultiplex=1
padVCDAudio=0
writeProgramEndCode=1
muxSVCDScanOffsets=0
alignSequenceHeaders=1
usePrivateStream2=1
useComputedBitrate=0
frameTimestamps=2
muxPulldownFlag=3
userMuxRate=25200
sectorSize=2048
packetsPerPack=1
videoBufferSize=232
audioBufferSize=4
audio1BufferSize=4
sectorDelay=0
videoDelay=180
audioDelay=180
audio1Delay=180
alwaysWriteSysHeader=0
maxFileSize=0
muxStartTime=0
muxStopTime=0
resetClocks=0
setBrokenLink=0
writeEndCodes=0
[advVideo]
profile=4
level=8
chromaFormat=1
colorPrimaries=5
transferCharacteristics=5
matrixCoefficients=5
videoFormat=1
intraDCPrec=1
vbvBufferSize=112
forceVBVDelay=1
displayHorizontalSize=720
displayVerticalSize=576
progressiveSequence=0
progressiveFrame=0
fieldPictures=0
topFieldFirst=0
repeatFirstField=0
intraVLCFormatI=1
intraVLCFormatP=1
intraVLCFormatB=1
framePredDCTI=1
framePredDCTP=1
framePredDCTB=1
quantizationScaleI=1
quantizationScaleP=1
quantizationScaleB=1
alternateScanI=1
alternateScanP=1
alternateScanB=1
writeSequenceDisplayExt=0
writeSequenceEveryGOP=1
writeSequenceEndCode=1
embedSVCDUserBlocks=0
writePictureDisplayExt=0
frameCentreHorizontalOffset=0
frameCentreVerticalOffset=0
qtNoiseSensitivity=12
featureFlags=0
reactionParameter=0
initialAverageActivity=0
initialIComplexity=0
initialPComplexity=0
initialBComplexity=0
initialIFullness=0
initialPFullness=0
initialBFullness=0
minFramePercentage=25
padFramePercentage=0
autoVectorLengths=1
horzPelMovement=3
vertPelMovement=3
forwHorzFCodeP=2
forwVertFCodeP=2
forwHorzSearchP=15
forwVertSearchP=15
forwHorzFCodeB1=1
forwVertFCodeB1=1
forwHorzSearchB1=3
forwVertSearchB1=3
backHorzFCodeB1=2
backVertFCodeB1=2
backHorzSearchB1=12
backVertSearchB1=12
forwHorzFCodeB2=1
forwVertFCodeB2=1
forwHorzSearchB2=6
forwVertSearchB2=6
backHorzFCodeB2=2
backVertFCodeB2=2
backHorzSearchB2=9
backVertSearchB2=9
forwHorzFCodeB3=2
forwVertFCodeB3=2
forwHorzSearchB3=9
forwVertSearchB3=9
backHorzFCodeB3=1
backVertFCodeB3=1
backHorzSearchB3=6
backVertSearchB3=6
forwHorzFCodeB4=2
forwVertFCodeB4=2
forwHorzSearchB4=12
forwVertSearchB4=12
backHorzFCodeB4=1
backVertFCodeB4=1
backHorzSearchB4=3
backVertSearchB4=3
forwHorzFCodeB5=0
forwVertFCodeB5=0
forwHorzSearchB5=0
forwVertSearchB5=0
backHorzFCodeB5=0
backVertFCodeB5=0
backHorzSearchB5=0
backVertSearchB5=0
forwHorzFCodeB6=0
forwVertFCodeB6=0
forwHorzSearchB6=0
forwVertSearchB6=0
backHorzFCodeB6=0
backVertFCodeB6=0
backHorzSearchB6=0
backVertSearchB6=0
forwHorzFCodeB7=0
forwVertFCodeB7=0
forwHorzSearchB7=0
forwVertSearchB7=0
backHorzFCodeB7=0
backVertFCodeB7=0
backHorzSearchB7=0
backVertSearchB7=0
ignore_frame_interval=0
user_iquant=1
intra_q[0]=8
intra_q[1]=16
intra_q[2]=16
intra_q[3]=22
intra_q[4]=26
intra_q[5]=27
intra_q[6]=29
intra_q[7]=34
intra_q[8]=17
intra_q[9]=16
intra_q[10]=16
intra_q[11]=26
intra_q[12]=27
intra_q[13]=29
intra_q[14]=34
intra_q[15]=37
intra_q[16]=16
intra_q[17]=16
intra_q[18]=18
intra_q[19]=27
intra_q[20]=29
intra_q[21]=34
intra_q[22]=34
intra_q[23]=38
intra_q[24]=22
intra_q[25]=26
intra_q[26]=27
intra_q[27]=31
intra_q[28]=36
intra_q[29]=37
intra_q[30]=38
intra_q[31]=40
intra_q[32]=26
intra_q[33]=27
intra_q[34]=29
intra_q[35]=36
intra_q[36]=39
intra_q[37]=38
intra_q[38]=40
intra_q[39]=48
intra_q[40]=27
intra_q[41]=29
intra_q[42]=34
intra_q[43]=37
intra_q[44]=38
intra_q[45]=40
intra_q[46]=48
intra_q[47]=58
intra_q[48]=29
intra_q[49]=34
intra_q[50]=37
intra_q[51]=38
intra_q[52]=40
intra_q[53]=48
intra_q[54]=58
intra_q[55]=69
intra_q[56]=34
intra_q[57]=37
intra_q[58]=38
intra_q[59]=40
intra_q[60]=48
intra_q[61]=58
intra_q[62]=69
intra_q[63]=79
user_niquant=1
inter_q[0]=16
inter_q[1]=18
inter_q[2]=20
inter_q[3]=22
inter_q[4]=24
inter_q[5]=26
inter_q[6]=28
inter_q[7]=30
inter_q[8]=18
inter_q[9]=20
inter_q[10]=22
inter_q[11]=24
inter_q[12]=26
inter_q[13]=28
inter_q[14]=30
inter_q[15]=32
inter_q[16]=20
inter_q[17]=22
inter_q[18]=24
inter_q[19]=26
inter_q[20]=28
inter_q[21]=30
inter_q[22]=32
inter_q[23]=34
inter_q[24]=22
inter_q[25]=24
inter_q[26]=26
inter_q[27]=28
inter_q[28]=30
inter_q[29]=32
inter_q[30]=34
inter_q[31]=36
inter_q[32]=24
inter_q[33]=26
inter_q[34]=28
inter_q[35]=30
inter_q[36]=32
inter_q[37]=34
inter_q[38]=36
inter_q[39]=38
inter_q[40]=26
inter_q[41]=28
inter_q[42]=30
inter_q[43]=32
inter_q[44]=34
inter_q[45]=36
inter_q[46]=38
inter_q[47]=40
inter_q[48]=28
inter_q[49]=30
inter_q[50]=32
inter_q[51]=34
inter_q[52]=36
inter_q[53]=38
inter_q[54]=40
inter_q[55]=42
inter_q[56]=30
inter_q[57]=32
inter_q[58]=34
inter_q[59]=36
inter_q[60]=38
inter_q[61]=40
inter_q[62]=42
inter_q[63]=44