Hxc Rev3: track change too slow?
Re: Hxc Rev3: track change too slow?
Thanks for the firmware. However, using it means that the users will also have to use it, so it is not really doable...
But I don't understand why you talk about "head teleportation". I don't know much about real hardware timings, but how long does it take to read one size-2 sector? This is the time I use to move the head. Plus the GAP the beginning of the track, plus the gaps at the end of the previous one, shouldn't it be enough? In "my" format, the first sector is not read when the head comes to the track, so it should have enough time to be ready for the second sector! Anyway, once again, what I am doing is really reliable on a CPC.
But I don't understand why you talk about "head teleportation". I don't know much about real hardware timings, but how long does it take to read one size-2 sector? This is the time I use to move the head. Plus the GAP the beginning of the track, plus the gaps at the end of the previous one, shouldn't it be enough? In "my" format, the first sector is not read when the head comes to the track, so it should have enough time to be ready for the second sector! Anyway, once again, what I am doing is really reliable on a CPC.
Re: Hxc Rev3: track change too slow?
About the time to read a sector : the track is shifted by one sector, so when you read the last sector you make a step, then you got directly the fisrt sector to read ! the only space between the last and the first sector is the sector gap !Targhan wrote:...but how long does it take to read one size-2 sector? This is the time I use to move the head. Plus the GAP the beginning of the track, plus the gaps at the end of the previous one...
About the start & end track gap, you are reading from sector 177 up to 186 then jumping to the next track. The sector 186 position is different on each track so you can't take the start and end track gaps.
am i missing something ?
Re: Hxc Rev3: track change too slow?
Hi guys,
What is a 15 ms settling time?
Even with the slow 3" drives of the CPC the step rate is with 10 ms on the safe side. 3.5" and 5.25" usually work with 4 ms.
When reading the last sector a track (that's the last sector I do read before I move the head to another track) without an error I tell the FDC to step to the next track, there I do read a sector ID, and then the next sector which is to be found in physical order.
What ever a settling time is has probably no impact, because after moving the track the first thing I do is to read a sector ID.
This works very quick on CPC (DS, 80 Trk, Vortex format), but is an order of magnitude slower on HxC.
IMHO some time constant in the HxC firmware is too big.
EDIT: What matters is that the HxC simulates reality, not floppy-disc-producer-paperwork, no?
What is a 15 ms settling time?
Even with the slow 3" drives of the CPC the step rate is with 10 ms on the safe side. 3.5" and 5.25" usually work with 4 ms.
When reading the last sector a track (that's the last sector I do read before I move the head to another track) without an error I tell the FDC to step to the next track, there I do read a sector ID, and then the next sector which is to be found in physical order.
What ever a settling time is has probably no impact, because after moving the track the first thing I do is to read a sector ID.
This works very quick on CPC (DS, 80 Trk, Vortex format), but is an order of magnitude slower on HxC.
IMHO some time constant in the HxC firmware is too big.
EDIT: What matters is that the HxC simulates reality, not floppy-disc-producer-paperwork, no?
Re: Hxc Rev3: track change too slow?
Mostly mechanical head stabilization + AGC (automatic gain control) circuit stabilization.TFM wrote: What is a 15 ms settling time?
BTW lots of drives "gate" the data line during the settle time and don't output anything.
http://eab.abime.net/showpost.php?s=480 ... ostcount=9
http://eab.abime.net/showpost.php?s=480 ... stcount=14
It does, and if you don't emulate correctly the settle time you make lots of amiga trackloader going crazy .TFM wrote: EDIT: What matters is that the HxC simulates reality, not floppy-disc-producer-paperwork, no?
Anyway if you want you can take the cheater firmware there :
http://torlus.com/floppy/forum/viewtopi ... 592#p11592
Re: Hxc Rev3: track change too slow?
Step and Settling time increased to 16ms for all machines (and not only Amiga) :
https://hxc2001.com/download/floppy_driv ... rmware.zip
Seems to load faster
So the HxC wasn't too slow, but too fast for the CPC routines.
Waiting for feedbacks.
https://hxc2001.com/download/floppy_driv ... rmware.zip
Seems to load faster
So the HxC wasn't too slow, but too fast for the CPC routines.
Waiting for feedbacks.
Re: Hxc Rev3: track change too slow?
Yes, you missed something indeed! I'm glad we could point that out . Let me use a simplified schematic.About the time to read a sector : the track is shifted by one sector, so when you read the last sector you make a step, then you got directly the fisrt sector to read ! the only space between the last and the first sector is the sector gap !
Track 0: #a1 #a2 #a3 #a4 #a5 #a6 #a7 #a8 #a9 #aa
Track 1: #aa #a1 #a2 #a3 #a4 #a5 #a6 #a7 #a8 #a9
...
- On track 0, I read #a1 to #aa.
- Goto Track 1.
- I give the order to read #a1. HOWEVER, the track is still moving/is not ready. *If* I wanted to read #aa on Track 1, *I would miss it* because the head is NOT ready yet, so it would have to wait a full turn. HOWEVER, I want to read #a1. By the time #a1 comes under the head, the head has finished its motion and is ready! So you see, there is no "head teleportation". I just shift the sectors so that the head has enough time to move and settle on the new track.
As a test, I created a format that shift the sectors of 2 slots, not 1. Like this:
Track 0: #a1 #a2 #a3 #a4 #a5 #a6 #a7 #a8 #a9 #aa
Track 1: #a9 #aa #a1 #a2 #a3 #a4 #a5 #a6 #a7 #a8
It IS working on the HxC. But it's 1/10 slower than the speed I expect with the first format shown here.
Is it clearer now?
Thanks a lot for your time.
Re: Hxc Rev3: track change too slow?
Have read my last post ?
Have you tried this firmware ?
Have you tried this firmware ?
Re: Hxc Rev3: track change too slow?
I over-read it, but is it an official release (I mean, will ALL the users use this specific firmware?)?.
Re: Hxc Rev3: track change too slow?
Ok, I used the last firmware and... it works! That's great.
Thanks a lot for your great work. When do you think the beta will be released?
Thanks a lot for your great work. When do you think the beta will be released?
Re: Hxc Rev3: track change too slow?
This beta is already available at the download pages.Targhan wrote:Ok, I used the last firmware and... it works! That's great.
Thanks a lot for your great work. When do you think the beta will be released?
Re: Hxc Rev3: track change too slow?
Great. Could you explain to me why exactly you said the "HxC was too fast", and how the new value made our loading track actually faster? Thanks.
Re: Hxc Rev3: track change too slow?
Thank you very much for the update, will test tonite
Re: Hxc Rev3: track change too slow?
Hi!
Tested the new firmware, but sorry, no change. Loading is still terribly slow.
Tested the new firmware, but sorry, no change. Loading is still terribly slow.
Re: Hxc Rev3: track change too slow?
Seriously, are you kidding to me ? Tested here, all is fine. And Targhan have confirmed that the speed was fine some weeks ago.TFM wrote:Hi!
Tested the new firmware, but sorry, no change. Loading is still terribly slow.
Some questions to have some valuable information about your test case :
Which HxC, Which CPC, Which image / software ? (Share the DSK and the HFE files !), At which moment into the software ?
Which firmware have you used exactly ?
You must use this one :
https://hxc2001.com/download/floppy_driv ... rmware.zip
V1.8.6.15a
Re: Hxc Rev3: track change too slow?
Of course I used this software. My HxC is Rev C from Lotharek.
Targhan uses custorm format. I use standard Vortex format (80 Trk, DS, 704 KB, 9 sectors a 512 KB).
Maybe you need to adjust a little more.
You can use this DSK image for testing:
http://futureos.cpc-live.com/files/DarthVader.zip
Please compare loading speed from real 3.5" disc on real CPC6128 with HxC. Needs 512 KB RAM.
Targhan uses custorm format. I use standard Vortex format (80 Trk, DS, 704 KB, 9 sectors a 512 KB).
Maybe you need to adjust a little more.
You can use this DSK image for testing:
http://futureos.cpc-live.com/files/DarthVader.zip
Please compare loading speed from real 3.5" disc on real CPC6128 with HxC. Needs 512 KB RAM.