convert cps to kbps

[lost_loven]

Like the title says.. Is there a formula for it? Just wondering because i randomly tried a 3meg file off aminet and it was bouncing from 32000 cps to about 46000 cps in ibrowse.

lost

[hooverphonique]

presuming cps is characters per second and kbps is kilobits per second, that question is impossible to answer without knowing what protocol you're talking about.

[Mozzerfan]

Originally Posted by lost_loven

Like the title says.. Is there a formula for it? Just wondering because i randomly tried a 3meg file off aminet and it was bouncing from 32000 cps to about 46000 cps in ibrowse.

lost

cps stands for character per second. A character is 1 byte. 1 byte =8 bits.
with transferspeeds 1 kilobit=1000 bits (with datastorage 1 kilobit=1024 bits).

edit: so roughly you can say: (cps x 8) / 1000 = kbps

[thomas]

Like hooverphonique wrote, you have to take protocol overhead into account. For http over TCP/IP through ADSL you may calculate ca. 10 bits per character. Therefore the right formula would be cps * 10 / 1000 = kbps. Or the other way round bps / 10 = cps.

[Mozzerfan]

Originally Posted by thomas

Like hooverphonique wrote, you have to take protocol overhead into account. For http over TCP/IP through ADSL you may calculate ca. 10 bits per character. Therefore the right formula would be cps * 10 / 1000 = kbps. Or the other way round bps / 10 = cps.

I think 10 bits per character is a little bit too much

I just tested this on my a4000/060 with x-surf2 and adsl with 3 different downloads (I timed them with a stopwatch) , and my calculation was pretty close.

edit: First I converted the cps average Ibrowse gave at the end of the download to kbps (by your calculation and my calculation). Then I calculated the filesize in kilobits and divided that by the calculated kbps. Then you have the number of seconds it takes to download the file. My calculation wasn't too far off. With your calculation, all 3 downloads were a few seconds off.IMHO tcp overhead in this case doesn't matter. The downloadspeed is already calculated by Ibrowse. It's only in cps. Lost_loven just wants to know what that is in kbps. With my calculation you get the downloadspeed Ibrowse calculated in kbps, without having to take tcp overhead into account.

[thomas]

Ok, we are talking about apples and oranges.

Of course if you transfer a file, then the file will arrive at the same size. A file of 1024 bytes consists of 8192 bits and thus you transfer 8192 bits of user data.

But the line needs some more capacity because of protocol overhead. If your ISP says your internet line is 16 MBit per second, then you get only ca. 1.6 MByte per second file download speed and not 2 MByte as your formula expects.

In the same manner if you see a file transferred with 32 KByte per second, then you can expect your line to have 320 KBit per second and not 256 KBit per second as with your formula.

The transmission size increases on the lower protocol levels. At the highest level (where you looked at) the file size certainly remains the same. But to transfer one byte user data the line has to carry ca. 10 bits. The additional 2 bits are filtered out by the protocol stack at the receiving side. They are used for error recognition, packet ordering and so on.

[Mozzerfan]

@Thomas

I understand, so when Ibrowse calculates the download speed, the tcp overhead is already included. It's just that the downloadspeed is expressed in cps. Or am I seeing it wrong?

BTW:When using your formula, when converting the filesize to kilobits i have to take the extra 2 bits in account. That's what I didn't do the first time.

[thomas]

Yes, but overhead is excluded. IBrowse only counts user data, it does not see protocol data. It only calls send and recv functions of bsdsocket.library. The library manages the protocol. There might be even more to add on the hardware side. For example a serial connection adds parity and stop bits. I am sure ethernet does something similar.

[diablothe2nd]

Originally Posted by thomas

cps * 10 / 1000 = kbps.

wow that's a bit complicated.

cps /10 would suffice

edit or cps * 0.008 = kps for the more accurate answer

[thomas]

cps / 100 would suffice. cps / 10 is wrong.

cps * 10 / 1000 explains what is done:

cps * 10 = bps
bps / 1000 = kbps

cps = characters (bytes) per second user data
bps = bits per second raw transfer data (= user data + protocol overhead)
kbps = kilo bits per second (1000 bits per second)

[diablothe2nd]

d'oh missed a zero!

[lost_loven]

So then say in Ibrowse i download something in at in around 38000cps.
Then it would look like this.
38000 x10/1000=380.
answer being 380kbs then

lost

[Mozzerfan]

Originally Posted by thomas

Yes, but overhead is excluded. IBrowse only counts user data, it does not see protocol data. It only calls send and recv functions of bsdsocket.library. The library manages the protocol. There might be even more to add on the hardware side. For example a serial connection adds parity and stop bits. I am sure ethernet does something similar.

Am I correct if I say that the downloadspeed in ibrowse (in cps) doesn't include 25% overhead?

If so, is it then correct to say that (cps ibrowse) + 25% overhead=(cps ibrowse) x 10 /1000 ?

Edit: Because if that formula is correct, my formula (cps ibrowse)=(cps ibrowse) x 8 /1000 is correct. All one has to know is if speed from ibrowse is with or without overhead.
So my formula always goes. Even when ibrowse would include the 25% overhead the formula goes, because the number of cps would be higher. That what I was trying to tell you in my previous post. So I guess I'm not the one talking apples and oranges, I'm just trying to do simple math.

Originally Posted by Mozzerfan

I think 10 bits per character is a little bit too much

I just tested this on my a4000/060 with x-surf2 and adsl with 3 different downloads (I timed them with a stopwatch) , and my calculation was pretty close.

edit: First I converted the cps average Ibrowse gave at the end of the download to kbps (by your calculation and my calculation). Then I calculated the filesize in kilobits and divided that by the calculated kbps. Then you have the number of seconds it takes to download the file. My calculation wasn't too far off. With your calculation, all 3 downloads were a few seconds off.IMHO tcp overhead in this case doesn't matter. The downloadspeed is already calculated by Ibrowse. It's only in cps. Lost_loven just wants to know what that is in kbps. With my calculation you get the downloadspeed Ibrowse calculated in kbps, without having to take tcp overhead into account.

Edit2:

Originally Posted by lost_loven

So then say in Ibrowse i download something in at in around 38000cps.
Then it would look like this.
38000 x10/1000=380.
answer being 380kbs then
lost

No, it should be 38000x8/1000=304kbps. But the speed ibrowse calculates is without tcp overhead. So the REAL downloadspeed could be higher. Thomas calculation is based on the assumption that tcp overhead is 25%. But that's a guess. It could aslo be 12,5%, which means the formula changes, because then the formula would be:
cps+ 12,5% overhead = cpsx9/1000.
Tcp overhead percentage depends on all kinds of factors and could be different (like hooverphonique said). But let's say for your system Thomas' assumption was right and tcp overhead is 25% , then the REAL downloadspeed would be 380 kbps. But then the speed ibrowse reports should also be recalculated, because the REAL downloadspeed ibrowse should report would be:
38000 + (38000/4)=47500 cps.

[hiho]

Slightly off topic, but I downloaded a 2.3MB file from Aminet using may A4000T and my new X-Surf 100 card using A-WEB II. Once the transfer got going, A-WEB II reported speeds between 72,000 CPS and 79,000 CPS. I never got speeds that high with the original X-Surf card.

[Mozzerfan]

Originally Posted by hiho

Slightly off topic, but I downloaded a 2.3MB file from Aminet using may A4000T and my new X-Surf 100 card using A-WEB II. Once the transfer got going, A-WEB II reported speeds between 72,000 CPS and 79,000 CPS. I never got speeds that high with the original X-Surf card.

That's really strange. With my a4000/060 + x-surf2 (edit: and a wifi bridge) I get speeds around 555000 cps with Ibrowse.

[lost_loven]

Originally Posted by Mozzerfan

That's really strange. With my a4000/060 + x-surf2 I get 555000 cps with Ibrowse.

what the heck??? Am i bottle necking the connection some how??? What file can we use to compare transfer rates with?


edit: what driver version you using? 1.11 ?

lost

[hiho]

Originally Posted by Mozzerfan

That's really strange. With my a4000/060 + x-surf2 I get speeds around 555000 cps with Ibrowse.

I have the stock 040. I'm sure that is part of the difference.

I'd have to download the same file in both Ibrowse and A-Web II to be sure they are reporting the same statistic.

[lost_loven]

nevermind.. didn't sink in xsurf 2 ... but still 555000 wow.. i better take a look at my set up again and see if something can be tweaked.

lost

[Mozzerfan]

@lost_loven

I have an 8 mbit adsl-connection. Try downloading something from os4depot.net.
The speed is usually very good there. On aminet I'm not getting this speed with every download.

With the xsurf2 I'm using the latest driver from Individual computers.

edit: Forgot to mention I'm using Roadshow.

[hiho]

I'm using MiamiDX.