|
You
can't
reach us
BY E-MAIL?
| |
We can't lay eggs but we can
tell if an egg is rotten...
And we might know the reason why.
The
headlines:
- New Ligos encoder has been tested
- Is MediaCleaner Pro worth it?
- New kid on the block, Canopus SoftMPEG encoder
The plans:
- Hauppage as Digital Video
Recorder
- Hardware MPEG-2 encoder/decoder for $199
- Three encoders from different point of view - a surpise...
When? ASAP |
Are
you feeling this page is boring? You are right!
This is why we decided to post some entertaining content.
THE (back)FACE of the YEAR
|
| There are a lot of stories on the market about
the encoder solutions. The basic problem is that the buyers wouldn't purchase an overly
expensive solution. In this case the decision is not so easy. You should buy a safe
solution - some expensive turn-key system - which we are sure can create excellent
quality, but we are assuming that the reader of this article is not the member of Oil
Millioners Club, therefore we chose encoders is in the $0-$10K range We chose some
of the cheap encoders and ran some tests on them to dreveal the truth. Recently we
tested realtime encoders too, although we know that due to the prices the offline encoders
are more popular.Some encoder manufacturers or developers were not ready to collaborate
us. Either they were affraid of the result or when they tried to convert our test file
they got a worse result than they expected and disagreed to publish the facts. Because we
have do not have enough resources, we can't buy all of the encoders released in the word,
and therefore we can't publish the abilities of some products.. If you would like to see
any other encoders here let us know the manufacturer's URL!
OR
You can make your own test and compare the result with the our ones.
 |
Intermezzo, about DivX, to say MPEG-4
Our problem with the MPEG-4 is
- Can't be played on normal TV monitor
- Desktop player hasn't been released
- The quality is not acceptable in some cases |
|
|
|
What
kind of encoders we tested?
- offline, software based
- offline, hardware based
- realtime, hardware based |
|
|
What
is "offline encoding"?
If you have an AVI file (or another kind of picture sequence) stored
on your harddisk, and this will be converted into MPEG-2 stream, this procedure is called
offline encoding. The advantage of this method is that the conversion is fully digital so
no D/A then A/D conversion exists which elliminates the quality loss caused by the output
and input circuits, such as PAL coder/decoder, digitizers, etc.. The disadvantage is that
sometimes (depending on the encoder) this way is very time consuming. |
|
|
The
real-time encoders
They have a big advantage such as that the encoding is not so time
consuming like offline encoding. But (as you can see below), some of the offline encoders
are very fast, and if the new CPUs will be released on the market those will be even
faster than the real-time solutions. The hardware encoders have a disadvantage. They have
fixed encoding algorythm, while the software doesn't. Yes, the new generation of encoder
chips have diownloadable microcode (Panasonic in Amber, C-CUBE in Apollo, RT-2000,
DC1000), but those codes do not contain the whole encoding program. I believe that the
future (realtime) hardware encoders will be just some simple "YUV" based capture
board and the encoder engine itself will be a software. This is more flexibile in relation
to the update. |
|
|
The source.
Because we have had a lot of experiences with or because we have a
lot of experience with MPEG encoding we know that there are critical things which can
cause bad results: Very fast, dynamic movings, a vast amount of small particulars,
unexpected scene changes. We chose a very complicated 23 seconds long clip as source
AVI file that meets those demands. The original source was recorded by a DV camcorder
which is a typical device nowadays. By the way, we discowered an interesting thing:
depending on the content the different encoders did accomplish the tasks with different
results, to be more precise one encoder was able to encode the trees in good quality but
the fast movings caused problems, another one created excellent quality at fast movements
but crashed at the swimming pool - and so on. (Is not optimal choice... - again, I'm not
sure what you mean here. You could say There is no optimum choice. A better way to say it
might be, There is no perfect solution.) |
|
|
Considerations.
Quality of the result. We didn't store any information of price, reliability and the
manageability of user-interface. |
|
|
How
did we run the test?
We prepared a 23 seconds long AVI file including critical scenes.You
can download this file from here:TEST.AVI
(80 Mbyte). To playback this file you need a dv codec, thus download one from the Canopus's
ftp site. |
 |
 |
 |
| Fast dynamic moving |
A lot of small particulars |
Trees, edges, small objects |
The MPEG encoder parameters were:
|
| Standard |
PAL |
| Resolution |
720*576 |
| Profile and level |
MP@ML |
| Frame rate |
25 |
| Bitrate |
around 5 Mbit avg, VBR if applicable |
| GOP structure |
IBBPBBPBBPBB |
After the encoding we loaded the result into Bitrate Viewer software which
tells everything about the basic information. Of course we watched the result on a
Sony high-end TV monitor also. We gave grades from 1 to 10 depending on the result
quality. This was a subjective rating, but we did our best to still be objective. There is
a very important aspect of the quantization scale that you can see on the Bitrate Viewer's
graph. The higher number means worse quality, but you need to know that sometimes is not
accurate to compare the numbers if one encoder uses a linear table the
other uses a non-linear one. This comparison can result in a mistake, not to mention
the alternate or zig-zag quantization tables. In fact, the best way to evaluate is to
download the mpeg clips. If the encoder has had some filtering feature we switched it off.
The computer we used is a dual PIII 500 Mhz with NT 4.0 srvpack 5.
|
|
Concepts
There are several concepts that must be clarrified for
some visitors. Thus we decided to open a page where you can
find some brief explanations of the written MPEG related concepts, such as bitrate, VBR,
CBR, CQ, quantization scale, etc. So if you find that you are a bit confused by our
terminology, please visit our explanations
page. |
|
|
The
encoders |
|
| Some of visitors asked us to
try-out the Wis MPEG software. We did and found so is not worth to deal
about it. Reasons? Creates just progressive stream, CQ is not equal with VBR (maybe they
don't know it), the quality is not so good at all and even they ask money for this
software. Maybe they downloaded some basic encoder software source-code from some
university's URL then try to make some product. Honestly, We don't understand this
company... Of course later if we can try-out
another encoders, the result will be included here. In case of you doubt in the
test, you may download the clips and watch them by your eyes. As mentioned earlier you can
download the original source AVI file also. If you find so the downloading is too time
consuming, we have a CD with these clips what you can order on our shopping site. |
|
The
results
Click on the encoder's name to look the result created by BV! A comment for DVD targeted users: the bitrate is not
allowed to exceed the 9.8 Mbit. Do not forget this point of view! The encoders under 7
points creates too much well visible artifacts, so those are not recommended for hard
sources. With very noiseless, detail-poor and clean sources there are no visible
differences between the encoders of 7-9 points. The order from top to bottom is important! |
By the way...We have tested some encoders
on a DUAL AMD ATHLON 1700+ MP system. Tsunami was done with the test avi
file in 42 seconds, Cinemacraft in 18 seconds. Don't ask more, we will not reply... |
| Position |
Encoder |
Time
(MM:SS) |
Quality |
Mpeg file |
Comment |
| 1. |
Tsunami Encoder version 2.02 |
05:10 |
|||||||| |
9 |
tdmpeg120.mpv |
Perhaps Hori-san has read our opinion... or just realized that the earlier version
has not created good enough quality. We don't want to talk more, the points we gave tell
eveything.Just only one thing, wish it worked faster... |
| 2. |
Cinemacraft Encoder SP 2.0 |
01:49 |
||||||||| |
9 |
anotherx.mpv |
Very
sharp and almost perfect picture. Still is the best encoder. Too expensive. But for pro
user this is the "final solution". It is our dream. Btw. in one pass VBR
encoding the time taken was around 0:30, but we found that the three pass encoding results
in better quality |
| 3. |
Futuretel NS 325 |
RT |
|||||||| |
8 |
ftel.mpv |
This
is a good example of the work of talented engineers. How is it possibile that this company
was able to build so excellent an encoder with a C-CUBE DVXpert chipset? Why can't other
companies do it? Well, we find that this board encodes very good quality mpeg, in spite of
the source being a bit noisy. Because the company does not have any DV playback device, we
were forced to copy the test avi file to Beta tape then send that to them. This copy added
some analogue noise to the original video, but the encoder solved the problem greatly -
without using any filtering possibility. The only thing we found wrong with the color
fidelity was that the result is a bit red - but we suppose that this is just a
setting problem. We are sure that with good quality sources the NS 325 creates very good
quality stream. Our last information is this company went bankrupt, this is a real
pity. |
| 4. |
MPEG Conversion
Studio 1.1 |
04:03 |
|||||||| |
8 |
mpgcstd.mpv |
VBR is
not implemented. Very expensive product. Is not worth it to talk about it. |
| 5. |
ExpertDVD |
18:47 |
|||||||| |
8 |
expdvd.mpv |
The
picture is a little smoothed. The VBR is not a true VBR, but a "Not Constant
Bitrate" CBR. Interesting...Terribily slow. |
| 6. |
LSX Encoder 3.5 |
02:36 |
||||||| |
8 |
lg35.mpv |
This
is what we were looking for. Now the interlaced source is supported, and even the
rescaling too. The VBR technique is very effective. The reason why we did not give 9
points is that there is some problem if the source is DV. We detected that the contrast of
the result does not match the source. We made a grayscale movie (DV compressed) what
contains luminance from 0 to 255, and this encoder created a result containing approx.
8-235. The explanation was that the encoder expects fixed luma scale from 0-255. This is
true, because we made an uncompressed RGB file as well with the same content, and the
result was perfect. The mentioned effect causes strange (blockies) effects at the darker
scenes, btw. Some of the competitor products can solve this problem though. Anyway, if the
source is not DV (but we really don't know what else it would be nowadays) this encoder
produces excellent quality. It is fast, very powerfull and cheap but we miss the scene
change detection. Ligos has a very effictive low pass and median filter, it is a pity that
when choosing the interlaced encoding those are removed, although they are very useful and
badly needed sometimes.Two problems are left. The first one is, it was very
problematic to get 5 Mbit averaged stream, if we used the simple VBR. Finally we set
around 3.5 Mbit to get the expected avg. bitrate.The optimized or simple CQ is not good,
because the the avg. bitrate of result and the file size is unpredictable. This is normal,
btw. Another problem is, also with the simple VBR setting, as you can see on the graph, at
the first second the bitrate is too high for no reason. In spite of mentioned problems the
new Ligos encoder is very promising. If they solve them, maybe they will get the first
position on this list, because the encoder engine itself seems to be very good. Btw, as
far as we know this software supports the new SSE2 instruction set (p4), and it is
possibile that the encoder works significally faster with this CPU. We wil try it very
soon, and let you know the result. |
| 7. |
MPEG Power Professional V2.0h |
04:50 |
||||||| |
7 |
heuris.mpv |
A true
middle class encoder.Each part of test clip created in the same quality. Nothing extra.. |
| 8. |
Cinemacraft Encoder Lite |
00:27 |
||||||| |
7 |
cctlite.mpv |
The
fastest encoder, and even very cheap, too. VBR is not supported. This version supports
just the 4 ,6 and 8 Mbit. A background info: if you give a 480*480 (NTSC) resolution AVI
file to this encoder, it will create a stream with the same resolution and the bitrate
will be 2.6666 MBit which is usable for SVCD, although is not a perfect solution, but
works. |
| 9. |
Canopus SoftMPEG encoder |
03:31 |
||||||| |
7 |
softmpeg.mpv |
It is
bundled with the Canopus's NLE systems such as EzDV, Raptor, DVStorm, etc. We think (or
believe/hope) this encoder is not done yet. The VBR technique is not as good as possibile.
The reason is that it works like a CQ encoder. When we choose the VBR the resulting file
is a semi-VBR stream in which the quantization scale never goes down under 9-10 value.
This value is the cognition level by the way, thus we suspect that the encoder
developer(s) knew this well. Maybe this is why they fixed it in the encoder at VBR. At
11-12 quantization scale the compression artifacts are getting visible (by good eyes). Our
conclusion is that for the average user this encoder gives acceptable quality.Interesting
btw., that this encoder has a special feature. It changes the sub-GOP depending on the
scene complexity. This means that sometimes the usual IBBP sequence changed to PPPP or to
IPPP. This technique results more accurate pictures in some cases, let's say the picture
very nice. Of course the color fidelity, etc. is perfect and as the other Canopus products
its stability is rock-solid. If the VBR technique would be better certainly we could give
at least 8 points. Let's see the future versions. |
| 10. |
MPEG
Profiler |
RT |
||||||| |
7 |
vitecyuv.mpv |
Well,
not this is what we expected. We are very sad that this encoder has not met our
expectations. We used the component (YUV) input to get the best result, but we were not
too contented. First, the color fidelity is bad, although it is adjustable but it is not
so easy to set the right values, however it is possibile - on cost of patience, because
choosing the YUV input the pass-tru control video is B&W, so record, check, adjust,
record again, check, adjust...and so on.Crazy! But what is really bad, the quality is not
as good as possibile. There are even some line pops too, the picture is a bit unstable, it
was really hard to get the encoder to create the average bitrate at around 5 Mbit,
btw.This encoder could be the opposite example of the realization of C-CUBE chip -
refering to Futuretel encoder. We do not recommend this encoder for professional purposes. |
| 11. |
Apollo DVD-er |
RT |
||||||| |
7 |
apollo.mpv |
Hhmmm,
we expected more from this encoder. You may see that the VBR concept is not too perfect.
We suppose that the Pinnacle DC1000 could create similar quality because they the use
C-CUBE chipset. Btw. if we switched on the filters we got better result, but the picture
became more smoothed. |
| 12. |
AMBER DV2MPEG converter |
00:41 |
||||||| |
7 |
amber.mpv |
The
picture is very sharp, the color fidelity is excellent, a typical Canopus quality in this
relation. Very fast encoder, uses the Amber board as encoding engine.This encoder
could get 10 points in reliablity and usability. With very clean sources this encoder
creates really professional quality, btw. We were a bit unfair testing this encoder with
our test file, btw. What we realized at higher bitrates using normal source files Amber
creates really excellent quality. As long as we agree that nowadays the average bitrate of
a DVD-VIDEO is around 6-6.5 Mbit/sec, then with this conditions this encoder could be the
best solution. Really better than Cinemacraft or Tsunami. This is why we suggest it. We'll come back to that later... |
| 13. |
RealMagic DVR |
RT |
||||||| |
7 |
dvr.mpv |
We
could give one point as maximum to this encoder in relation to installation. We made a
heroic struggle while installing. It was not easy. The documentation and
the settings panel have a lot of bad information, not to mention that it is crazy to set
the PAL format. It seems the Sigma Designs's developer team does not know the
Visiontech's KFIR chip which implements just the 704 as horizontal resolution, the user
can choose the 720 which will not be correct. When you choose smaller resolution, the
interpolation adds a lot of vertical artifacts. It is very ugly sometimes.The realtime
preview is great anyway...if works. Recently we tested the W2K version one. We can't say
revolutionar changes. |
| 14. |
Digital VideoCreator II
(Fast DVD.Master in Europe) |
RT |
||||||| |
7 |
dvdmaster.mpv |
That
is an unique product.It is true that it is not ranked too high on our list.The reason
is the same stupidty of developers just like of the Pinnancle's DV500. When you set the
VBR mode the software just asks for only one parameter. The peak and the average bitrate
will be the same, so the software does not utilizes the possibility of the used C-Cube
DVxplore chipset, although that can do more in this regard. Practically the VBR is not
effective at all. Anyway, when we tried to make an SVCD compatible stream, the result was
surprising good. We expected worse, btw. We believe that it is worth it to deal with this
encoder, because hopefully the manufacturer will modify this small problem mentioned
above. If so, this stuff will get more points from us. Tthis is the best buy currently. |
| 15. |
RT
2000 |
2:21 |
|||||| |
6 |
rt2000.avi |
Though
this product is not designed just for MPEG creation, this is an NLE solution basicly, a
lot of visitors asked us to say something about it. It was not too easy to test it,
because we spent a lot of time to get the stuff to work. We had some problems with the
stability, etc. but finally we won. Well, this is not the product that we need. The
settings of MPEG format is graciously simple, and maybe this is why the VBR of the created
stream is not too effective, let's say it looks like a CBR stream. As you may see, the
peak bitrate was set to 4000 to get the average of 5000, and finally it succeded at around
5500. So it is hard to make usable stream in this point of view. Another small problem the
created file has AVI header, so the user has to convert it with some small utility to
"avi-headerless" file.We don't know how this product works as NLE solution, for
MPEG creation we would not use it. Neither the quality of stream nor the reliablity
reaches that level what we could accept. |
| 16. |
MPEGator 2 |
RT |
|||||| |
6 |
mpegator2.mpv |
That
is a cheaper MPEG-2 realtime encoder. Uses a Visiontech chipset called MVCast which is an
older chipset, and maybe this is why it does not support the VBR. The picture is very
smoothed and the color of the mpeg is not the same as the original source, a litlle
greener. This board has YUV input too, and using that the picture would be significally
sharper, but it is impossibile to set the right color ballance, so unfortunatelly that
input is unusable. |
| 17. |
Vitec MPEG2 tools |
03:52 |
|||||| |
6 |
vitec.mpv |
Is not
recommended for DVD, although there is a switch on the settings window referring to DVD,
but choosing that the software switches to CBR mode. We know well why does it. The reason
is the VBR method of this encoder. Completelly the same way as the BBMpeg's. See below. |
| 18 |
DV 500 |
00:48 |
|||||| |
6 |
dv500.mpv |
C-CUBE
DVXpress chipset, just like on the Apollo. Hmmm, so we really don't know how share our
opinion without offending Pinnacle. When you watch this stream on a TV do not think your
TV is damaged. We were not able to create better q. stream. The most important problem is
the flickering. There is some problem with the field order what we couldn't solve even we
tried everything. Not to mention these guys don't know that the VBR needs two parameters
(at least!!!), the average bitrate and the peak.They are asking for just one and the
result will be 5000 as average and peak. We wouldn't buy this board just because of the
MPEG encoding capabilities. The used chipset itself would be able to create better
quality, by the way. |
| 19. |
BBMpeg 1.2 |
06:25 |
||||| |
5 |
bbmpeg.mpv |
Smart
& Free! Bravo! Can't create good VBR, btw. This software creates Constant Quality
stream in the VBR mode and this results huge file. Is not worth to use this setting.
Warning! CQ is not equal with VBR! Anyway, we suppose, the lot of settings at the advanced
window could help the better quality creation. We haven't had time to try those settings
because ths encoder is too slow. |
| 20. |
MEGAPeg |
21:40 |
||| |
3 |
digigami.mpv |
It is
very slow and creates terribily bad quality. VBR not implemented. This is a
joke. |
| 21. |
MediaCleaner Pro 5.02 |
11:13 |
||| |
3 |
mclean502.mpv |
It is
also very slow. VBR not implemented in this version. Is not worth wasting space on
creates on it - so bad quality stream. Even if the bitrate is the highest, close to 6
Mbit. They have big name for no reason - regarding to MPEG encoding. |
| 22. |
DVMpeg 5.0 |
03:14 |
|| |
2 |
dvmpeg.mpv |
Strange,
Darim is expert in MPEG technology. It seems this is not their favorite product.. VBR is
not implemented. We suppose they will improve the abilities of this encoder. It would be
worth... |
|
The opinions of visitors
We receive a lot of emails day by day, so we opened an "opinions page".
If you have any comment, experience or suggestion about the tested encoder, even if you
are the manufacturer or just a user you should send your thoughts to opinion@tecoltd.com and we will post them on this
page:
|
OPINIONS
To publish your comments
there are some rules:
If you are a user you have to send us your real,
working email address because we will publish it under your thoughts.
If you are the developer or
the manufacturer of the tested encoder you have to include in your email your full name
and your agreement of the publishing of your name and opinion on this page. We will not
publish your email address if you ask it.
It is possibile that we will
complete the sent opinion with our own, in this case we will use bold blue fonts. You have
to accept that we need to insert our standpoint.
If the sent thoughts are
based on missunderstandings or on the lack of right information on the how to create MPEG
we will not publish your comments, unless if you insist on the publication after some
correspondence. In this case we will sign our disagreement under your comments.
If you have bad english, we
will try to correct your grammar to make it understandable :-), we are not too successful
at this either, but we can agree that this is not so important. German, French, Italian,
Spanish and Hungarian languages are accepted besides the English. We will try to
translate.
We will publish your comments
even if you have a very bad opinion about our tests or you protest against them as
manufacturer or developer, so do not worry about this.
If you send any email to the
mailbox mentioned above, this will mean you accepted the rules and you will not start any
protest against the publication.
If you do not want to see your opinion on the web, but
you would like to inform us about it, you can send an email to the
enctest@tecoltd.com
mailbox. We promise your thoughts will not be published
on the web.
|
| Last updated 02.06.2002 |
Special
thanks to Jonathan Stone |
|
|
