On 3/31/10, Tony Duell <ard at p850ug1.demon.co.uk> wrote: Indeed. Yep, and the Byte article design has the phototransistor for the sprocket track installed "backwards" with respect to the 8 data tracks such that the sprocket track fires later (due to mechanical asymmetry in how phototransistors are manufactured) which enables it to trivially work as a latch pulse (since the data channels were exposed to light earlier and have had a chance to settle/allow the holes to more completely match the sensitive spots of their phototransistors). Yep. I think that's a common approach for optical sensors. For my TIFF/GIF-based "reader", I did edge detection of a snapshot of the entire card to check for scale and orientation so that once the leading edge was identified in software, it shouldn't be difficult to scan along the long axis, looking for a dark spot or the absence of a dark spot every so many pixels (depending on the scan resolution). It doesn't matter if you scan at 75 dpi, 300 dpi, or more - it's all proportional to the detected height and width of the outer edges of the card image. You do have to keep the card mostly perpendicular to the edge of the scanner or there needs to be smarter code than I wrote to detect the silhouette of the card (it starts at each edge of the image and works its way to the middle, looking for active pixels in the high-frequency-filtered version of the card scan - it should still work if you are not 100% square to the scanner, but not if it's more than 1-2mm off-axis). That's an interesting approach. It might not be so expensive these days to make a PCB-based mechanical scanner, but I would personally rather work on an optical column-at-a-time scanner than a multiple-hundred-mechanical-connection electrical scanner. If I had a reader head to start with as you did, I might feel differently. It's an interesting approach, though. -ethan

I used to have a Wang card reader. It had 480 contacts, but they were every other column. It was for "Hanging Chad" port-a-punch type cards. It was a large clamshell that you manually opened, placed a card in it, and manually closed. Therefore, it was not for high-volume service-bureau card deck reading - slightly slower than keying in the data. I had previously used a Wang desktop programmable calculator that had one. If you wanted to read a normal 80 column card, you could read the card, turn it over, and read the other columns, since the margins for IBM cards were symmetrical. On the Wang calculator, I used an 026, and made "flippy" cards for some of the more commonly used macros. -- Grumpy Ol' Fred cisin at xenosoft.com

On Mar 31, 2010, at 4:08 PM, Tony Duell wrote: I wonder if this could be handled with a simple PLL-based clock-recovery circuit. I suppose the ones density isn't exactly guaranteed though. What if the card slips? -Dave -- Dave McGuire Port Charlotte, FL

On Apr 1, 2010, at 12:54 AM, Fred Cisin wrote: Ahhhhhh I see. Can they re-feed, or would that be the operator's responsibility? (I've never used a real IBM card reader) -Dave -- Dave McGuire Port Charlotte, FL

On 1 Apr 2010 at 0:17, Ian King wrote: Not necessarily. If you look at Hollerith's card reader, it read the entire card in parallel, using spring-loaded probes that would extend through the card holes and touch a pool of mercury. No need for clocking. Serial card reading came later when the card was not manually read. If you're going to automate card transport, then reading the card serially actually is easier. What was the proportion of readers that read colum-serial versus row- serial? I believe the CDC 415 punch had its read station as 80 brushes, reading the card just punched row-wise. --Chuck

On 1 Apr 2010 at 12:14, Fred Cisin wrote: It's the 250 CPM punch usually hooked to CDC 3000 and 6000-series gear. Prone to fail in big punch jobs because of overheating. Punches a row at a time. --Chuck

On 4 Apr 2010 at 1:29, CSquared wrote: One doesn't need a memory with the web. From the Columbia university web site http://www.columbia.edu/acis/history/sorter.html: Type 84 19?? 2000 cpm "For some reason, IBM didn't produce a card reader that operated at 2000 cpm. The 084 was unique in that it used a vacuum-assist card- feed mechanism that did not require the use of a card-weight, as the cards were held against the throat and picker knives by suction. The 084 also used an optical method of detecting card punches instead of the wire brush used by earlier models. " I never got to use an 084, just an 082--there must have been a jillion of those in service. --Chuck

On 5 Apr 2010 at 13:25, Fred Cisin wrote: I went to the bitsavers document on the 082, 083 and 084 sorters. The counter was an option for all three--it could count total read as well as the number in each pocket. There were some other fancy options, including alpha sort and a 10-column brush with a mini plugboard for the 082 and 083 to do all manner of fancy combination sorts. I never saw any of the esoteric options. The 084 had a 3600-card feed hopper. That means that the beast would empty it in less than 2 minutes! The operating notes state that you didn't need to "joggle" the cards (another IBM term) on the 084 as it was "self-joggling". Cheers, Chuck

Fred Cisin wrote: I was thinking 2000 per minute for the 084, but that seemed a bit outlandish and I was too lazy to search for it. I think I only had one in my territory, at an insurance company. They later got a 108 "Statistical Card Proving Machine" too which was a bit unusual in our area. I only recall seeing one other one. I'd forgotten the 082 ran at 1000 cards per minute, which is not too shabby either. Don't be sorry, I love it! I think the appreciation for things like this grows the older one gets.

On 31 Mar 2010 at 21:54, Fred Cisin wrote: The CDC 405 uses two rows of photodiodes and compares the read reslts as the card moves through. If the card slips, a read compare results. Two photodiodes are used at the front of the read station, positioned between rows. As the card edge moves through, the first diode, then the second diode is interrupted. If this doesn't happen, a fault is generated. Although the card feed is performed with a vacuum capstan, which always has vacuum applied (unlike the capstans in the 6xx tape drives) , pinch rollers move the card through the read station. Watching the thing read at 1200 cards per minute is pretty impressive. Another aspect that's interesting is that the top and bottom of the read station throat are wider than the middle section to accomodate dog-eared cards. --Chuck