USB twice as fast as parallel?

[time]

I've been playing with an Epson C61 (C62 in the US and sells for about US$80 after rebate). It's way, way better than I expected. But what I want to talk about today is speed.

1. A4 photo (10.67x8" plus margin for all you metric-phobes) with maximum options: just under 16 minutes. This excludes the processing time of about 45 seconds for the Nikon software I was using. Maximum options means RPM (Resolution Performance Management, i.e. 5760 dpi) was enabled.

2. A4 photo with no options: under 8 minutes. RPM was disabled, i.e. 2880dpi. At first I thought there was a visible difference, and maybe there was, but when I looked at them a couple of hours later I couldn't tell which was which, magnifier or no. Maybe when we get a clear day ...

3. Swapped parallel cable for USB: 3.5 minutes. Say what?!

4. A4 photo with maximum options: about 7 minutes. (Sorry if this is a bit vague; I didn't set out to benchmark the thing precisely).

Conclusion 1: On the C61, RPM is mostly a waste of time (literally).

Conclusion 2: On the C61, USB is at least twice as fast as parallel.

Conclusion 3: This US$80 printer can produce quite reasonable photos (on Epson photo papers) in 3.5 minutes, very nearly what the manufacturer claims. <faints>

I compared a print with the result from a three-year old Epson A3 photo printer (can't remember the model). No contest, the C61 killed it (although maybe that means the older printer is sick - who knows?).

I also tried printing with the "plain paper" setting on 80gsm copy paper, the nemesis of ink jets everywhere (21lb bond for those of you who still believe in that ridiculous and inconsistent method of rating paper ). The photo oversaturated the paper but was pretty darn good, considering the media. Way, way better than its predecessor, the C60.

I also printed on 280gsm card (175lb "tag" stock, not inkjet compatible, and typical of higher quality business cards), and achieved equivalent results to the cruddy copy paper, albeit without the oversaturation - in fact I printed both sides. If you look, there is a smattering of white specs where the ink hasn't taken, and the picture lacks the depth and clarity of photo paper. But there are inkjets out there rated to only 100lb tag stock (Lexmark Z55 anyone?), nevermind the 440gsm (275lb) the C61 is rated to (in envelopes).

A 'normal' print in "text and graphics" mode took 40 seconds, and looked more like I expected. I haven't tried "draft" yet ...

Anyway, this impromptu mini-review raises the question: how many other printers need USB to reach maximum speed?

 

[LiamC]

Time, what LPT settings do you have? It makes a big difference.

Example 100MB LPT Zip drive

LPT - normal/SPT/uni-directional setting. Copy 90MB took nearly 5 hours
This is the sloest setting. Comms is one-way and usually in nibbles (1/2 bytes)

Bi-drectional - not much better but at least comms is two way

ECP - good for printers (and recommended setting). Buffered two way comms up to around 8MB/minute. As above took ~40 minutes

EPP - burst mode up to 22MB /minute, but no buffering. Good for external drives. 90Mb copy took less than 15 minutes. Version 1.9 differs from 1.7 in the cable required, and almost all cables these days are capable of 1.9.

Whilst a lot of printer manufacturers recommend ECP for printers and scanners, I've sometimes had better results with EPP if the files are large (photo's).

 

[Tannin]

Thanks, Time, that's a very interesting set of numbers and mini-review. It does nothing but confirm my desire to keep on selling Epson inkjets and not bother selling any other brand. But in truth, I would need three bishops swearing on a stack of glossy six by fours that, yes, the Epsons really do cause birth defects in mice before I'd consider selling anything else - because I have my own little set of numbers: $115ex, 30-odd units in service, zero returns, zero support calls. And, as you know, 115 + 30 + 0 + 0 = 0. Zero chance that I'd bother selling anything else, that is.

PS: I'd be very interested to see what happens if you can find time to toy with Bill's suggestion, by the way.

PPS: I just realised you were talking the C61. Owing to a lamentable slip of the brain, I was thinking of the C41. It was only when it occurred to me that you must have two of them, because the C41 only comes in parallel or USB (for some absolutely stupid and completely incomprehensible reason), that I twigged. Oh well. Here's my set of C61 numbers (sans correct price, because I forget the exact price): $150 + 20 + 0 + 0. Oddly enough, through some trick of mathematics, those numbers add up to the exact same answer: zero.

 

[time]

Time, what LPT settings do you have? It makes a big difference
...
EPP - burst mode up to 22MB /minute, but no buffering. Good for external drives. 90Mb copy took less than 15 minutes. Version 1.9 differs from 1.7 in the cable required, and almost all cables these days are capable of 1.9.

Whilst a lot of printer manufacturers recommend ECP for printers and scanners, I've sometimes had better results with EPP if the files are large (photo's).

Currently EPP/ECP. When I checked with Epson tech support, the guy said ECP and muttered about printer cables. Win2k is recognizing the port as ECP and I tried two different cables, at which point I lost interest because if I don't have a suitable cable then I can hardly expect users to.

But then the Epson guy also claimed there should be no difference between parallel and USB (we've asked for a call back from an Epson guy who does have a clue), so I'm keen to try what you suggest.

 

[Pradeep]

Wow, that certainly is fast. Much faster than my C80, which takes I would estimate 10 mins for a 1440dpi 8*10 print on Epson Glossy Photo. I'm close to ordering a Canon S9000 because of the speed, or lack of. Apparently the new Epsons do suffer from a "bronzing effect" on glossy papers, I've got some heavyweight matte to see how that goes.

 

[LiamC]

Time,

the cable spec is IEEE 1284a I think (can't remember off-hand). Another thing is, I've never had much luck with combined mode (ECP/EPP). You may want to try just ECP or just EPP.

0.05

 

[P5-133XL]

I'm sorry but there is something wrong here. There is no way that a parallel port could cause such a slowdown. A printer is mechanical and and both USB and parallel cables are electronic. There is no way that even a slow parallel cable will cause a large scale slowdown of your printer unless there is very serious HW/driver problems. The transfer of data should occur on the order of seconds not minutes.

A normal modern parallel port (non ECP/EPP) should be able to transfer data at aprox 100KBps and an EPP/ECP port should be able to transfer at aprox 500+KBps while a USB 1.0 should be able to transfer at aprox 1.0+MBps. With this in mind, the transfer of 5 MB of data should take only 50 seconds on the slowest interface. Regardless of the transfer method, the mechanical output should be taking the vast majority of time rather than the interface.

 

[blakerwry]

im not sure if you realize the amount of data that is being sent to the printer at 2880dpi

10" * 2880 * 8" * 2880 = 663552000 dots per page

The printer has to have color information for each one of these dots... if it were as simple as either on or off for each color then there would be 1 bit per each color...

In a 3 color printer this would be: 1990656000 bits... / 8 = 248,832,000 bytes


I don't know exactly how inkjets send picture information to the printer, but I assume RAW mode would be something like this, while more advanced modes would use compression or another technique to reduce the mess.

 

[jtr1962]

Actually, it would be 3 bytes per dot, not bits, and for an 8" x 10" photo at 2880 dpi that would be 1,990,656,000 bytes that need to be sent to the printer in raw mode. Since USB maxes out at around 1 Mbyte/sec, in 3.5 minutes you can only send about 210 MBytes to the printer, so I would imagine that the raw data is converted into more efficient printer instructions in order to cut down on the printing time. In any case, I have no problem believing that the data transmission rate is the limiting factor here, at least for photo quality. Text and graphics, on the other hand, are likely limited by the mechanics of the printer. The instructions for my HP940C mention that when printing large photos using photo quality you will temporarily need a few hundred MB of hard disk space. I would imagine this is indicative of the amount of data that must go to the printer. Also, in photo mode the printer prints a bit at a time and stops, suggesting that it must wait to receive new data. I may get a USB cable and see if there's any noticable difference using USB instead of ECP. Theoretically, ECP is supposed to be capable of 2Mbytes/sec, but in practice it may not even come close.

 

[LiamC]

Digging around in the Linux drivers for a Lexmark 7000 ColorJet, there is mention of the need for up to 700~770MB of temporary swap space for very large photo files when printing in 1200x1200 dpi mode.

Most cheap inkjets get the host CPU to do all the rasterisation before passing the data to the inkjets "print engine" (I use the term rather loosely).

Also, the IEEE 1284 spec seems to be rather loosely written so to achieve good speed out of the LPT port you need good hardware and good drivers (which is why combined modes might not be such a good thing). USB isn't in this boat because the spec has now been tied down a lot., as is the case with PnP, PCI, HyperTransport etc. Try different modes.

more info

http://www.bb-elec.com/tech_articles/parallel_port_explained.asp

http://www.fapo.com/ecpmode.htm

http://zone.ni.com/devzone/conceptd.nsf/webmain/0989d3d9dafae64e8625680400679736?OpenDocument

 

[blakerwry]

Actually, it would be 3 bytes per dot, not bits

why?


Does each color have 24bit (~16.8 million possibilities)?

Or were you saying that the total of all three colors was 24 bits(256 possibilities per color)?

 

[jtr1962]

Actually, it would be 3 bytes per dot, not bits

why?


Does each color have 24bit (~16.8 million possibilities)?

Or were you saying that the total of all three colors was 24 bits(256 possibilities per color)?

I was thinking of how colors are displaying on the monitor, so you might be right. They probably get millions of colors by dithering the three basic colors and black at 2880 dpi so that the eye doesn't notice(plus the colors will always blend slightly before they dry, further hiding the "dot" effect). I have also heard of some inkjets with variable dot size, although I don't think that 256 dot sizes per color is possible. Even with 1 bit per dot, you're still talking about nearly 250 MB of data, which incidentally USB can send in 3.5 minutes, give or take. Obviously the parallel port limits the printing speed for photos in this case.