Test Equipment Anonymous

Having spent too much time trying to decypher schematics of this quality....
... a unicorn appeared: a Solartron 7075 Technical Manual appeared on fleabay, and amazingly I was able to buy it.

Let's hope it has the schematics. Thereafter I'll have to decide whether to scan all/part of it and/or try to recoup (time) costs doing that.

That's great news. Make sure you send a copy to the relevant sites so other people don't have to suffer.

It does make me wonder who the hell scanned the original ones and if they actually gave a shit about what they were doing...

I'd be interested in a copy of the scans, if possible.

Thanks.

bd139 wrote: ↑Sat Apr 27, 2024 5:56 pm That's great news. Make sure you send a copy to the relevant sites so other people don't have to suffer.

It does make me wonder who the hell scanned the original ones and if they actually gave a shit about what they were doing...

Sadly, all too often that is the quality of schematics these days. I think that it might stem from that they take a little more care when it comes to the text pages as they can read those for QA, but schematics just leave them cold as they cannot figure out what they are or just how important they are.

bd139 wrote: ↑Sat Apr 27, 2024 5:56 pm That's great news. Make sure you send a copy to the relevant sites so other people don't have to suffer.

It does make me wonder who the hell scanned the original ones and if they actually gave a shit about what they were doing...

All sorts of things happen. Sometimes an original manual has degraded through age or was not great to begin with. It may be hard to scan because of the binding.

I uploaded a copy of a Marconi manual which had been unobtainable for years. People had appealed for it at the other place with no response. I bought the instrument with a photocopy of the manual which I scanned. It's not great, but it's usable and a lot better than nothing.

It also depends on what sort of scanner you have, and whether you know how to use it. Some people may not have a scanner and may use a camera. I've done that to email excerpts from books manuals to friends. Very often schematics are awkward because they are not A4. I've got a few which have had to be made up using A4 pages put together with Sellotape.

For bog-standard pages, I use a bog-standard scanner at 300dpi

For A3 I'll either use a camera, or - if possible - two A4 scans.

Stitching photos or scans together is via the hugin panorama converter.

I've a command line utility to reduce colour depth and store as TIFF file[1], insert that image into a PDF file, concatenate all the PDF files.

[1] it is the only way i have found of preventing compression of the image when inserting it into the PDF. Since the TIFF files are small enough, I haven't looked very hard

Good news bad news.

The bad news is that even with decent originals, some of the lines are very thin.

The bad news is that the hugin panorama stitcher is proving to be a pain with photos of schematics. Even with many manually added homologous points, the straight parallel lines confuse it.

The good news is that better schematics can become available, as and when time permits. The photo is 460kB, the thresholded TIFF variant is 14kB.
Those scans (original scan, and thresholded variant) are intriguing in one respect: some of the imperfections are the same as in the picture in the first post. Look at SK2 pin 14 and IC13 towards the bottom right. Either the scan was made from this original, or all manuals suffer from the poor reproduction. I'm inclined to believe the latter.

They look a lot better than the one you posted originally, which seemed a low resolution scan. I believe I could follow them with much less guesswork and writing things in. Occasionally you come across a manual where the diagrams weren't great in the first place. Connections go through legends, and the writing is very small and indistinct, so a pin number could be a five, and eight or a six.

I wouldn't bother with the panorama stitcher if it's proving troublesome. Just make the best scans you can and people can put them together themselves with Sellotape or Pritt.

Zenith wrote: ↑Thu May 02, 2024 4:57 pm I wouldn't bother with the panorama stitcher if it's proving troublesome. Just make the best scans you can and people can put them together themselves with Sellotape or Pritt.

I might end up doing that, but it doesn't feel right!

Standard British quality throughout then!

There is another option.

There's a local copyshop that I've used to print schematics on A3 before. They were very relaxed, amenable, and cheap. (Much more so than other places)

They have both A3 and A0 scanners, so I might just get them to copy the dozen pages that need doing.

One thing I try, if you have one, is to use whatever digital camera you have and take a photo, then edit the curves so everything is either black or white. If you get it front on, the distortion is barely discernible in anything produced in the last 10 years. 6MP should be enough to cover A3.

Attached photo was not shot front on hence the mess but it shows what you can get. This is just a phone photo cropped down from a double A4 page as well...

bd139 wrote: ↑Sat May 04, 2024 11:51 am One thing I try, if you have one, is to use whatever digital camera you have and take a photo, then edit the curves so everything is either black or white. If you get it front on, the distortion is barely discernible in anything produced in the last 10 years. 6MP should be enough to cover A3.

Attached photo was not shot front on hence the mess but it shows what you can get. This is just a phone photo cropped down from a double A4 page as well...

IMG_0121 (1).jpeg

Unfortunately....

Here's a 5k*3k camera image of A3 (i.e. not the largest), lightly cropped to remove the crap, then thresholded.[1] The second picture is a cropped version of the far right. Barely readable The third picture is scanned at 300dpi, then thresholded. Shows how much detail is missing from the camera version. Still perfectly readable, but wouldn't want to go below 300dpi. One or two places might need scanning at 600dpi.

[1]The TIFF variant is 43K, the PNG variant is 76K, a ratio that is constant. I put the TIFFs into the PDF files.

Well that explains it. The threshold, in the process of reducing image size, reduces enough information through quantisation that it's not possible to interpret it.

A suitable example

.... book page from camera ...
.... crop ....
... push contrast around ...
Disclaimer: requires competent image processing software.

The last is a JPEG for ref. It was 112k 1412x2000px but not sure what the forum does to it...

You're not going to get a good result in 1 bit per pixel for most things.

bd139 wrote: ↑Sat May 04, 2024 4:14 pm Well that explains it. The threshold, in the process of reducing image size, reduces enough information through quantisation that it's not possible to interpret it.

My thresholding keeps the same number of pixels.

OTOH the picture I uploaded was not 2000 pixels wide. Looks like something on this forum has reduced the width.

Disclaimer: requires competent image processing software.

The last is a JPEG for ref. It was 112k 1412x2000px but not sure what the forum does to it...

You're not going to get a good result in 1 bit per pixel for most things.

The problem is the number of pixels in the source image being too small for the very fine lines in the original. No, I'm not going to buy another camera!

It's not about the number of pixels but the information represented by them

Consider the two images. One represents a better "shape" than the other despite having the same number of pixels. One pixel definition has two states (1 bit). The other pixel has three states (around 1.58 bits...)

DSP theory and practice indicates the XY and Z resolutions can be traded off to a surprising degree.

Nonetheless, if the original camera image is like this, then I don't see any thresholding algorithm being effective.
H/M, R/P, 8/6/3/2, K/X - even an oracular human won't know with certainty.

Well that's sort of the point. But you can use some inference up top to read the semantics rather than the syntax. I mean there isn't going to be a pin 0 unless someone is sick bastard!

Ideally what would be nice is a nice vector PDF (says me hammering out LaTeX at the moment)

bd139 wrote: ↑Sat May 04, 2024 8:12 pm Well that's sort of the point. But you can use some inference up top to read the semantics rather than the syntax. I mean there isn't going to be a pin 0 unless someone is sick bastard!

That's the arguments the LLM proponents use to justify the "adequacy" security/networking code emitted/dropped by LLMs. "It is plausible and that's all you need; you can infer the rest".

There are 250 pages, 3-4MB/photo, so only the best part of 1GB - and then you have to make guesses on top of that. Do you work for MicroSoft?

tggzzz wrote: ↑Sat May 04, 2024 8:23 pm

bd139 wrote: ↑Sat May 04, 2024 8:12 pm Well that's sort of the point. But you can use some inference up top to read the semantics rather than the syntax. I mean there isn't going to be a pin 0 unless someone is sick bastard!

That's the arguments the LLM proponents use to justify the "adequacy" security/networking code emitted/dropped by LLMs. "It is plausible and that's all you need; you can infer the rest".

There are 250 pages, 3-4MB/photo, so only the best part of 1GB - and then you have to make guesses on top of that. Do you work for MicroSoft?

That lines up neatly with my afternoon imaging experiment with ML (Machine Lying) which had a somewhat hilarious conclusion.

There is a new feature in Photoshop called generative expand which takes an image and fills in an area outside of the image for you. I figured it had certain boundaries for the inference before it generated garbage and I was right.

So I carefully chose an image from my photos which I thought would go wrong...
This was "extrapolated" out. Not a terrible job.
Now this is where it gets fun....

.... zoom right out ....
Which is quite frankly hilarious

I've finally uploaded solartron_7075_DVM_dataSystems_userServiceCalibration_1980.pdf to ko4bb and BAMA. Don't know when they will appear for download, of course.

File is 15MB whereas the previous illegible versions were 13MB. It was going to be 65MB, until I noticed a stupid expansion of two scans!

Have fun

Don't download the "750_Technical_Manual..." from BAMA just yet, since it is 66% smaller than the one I sent, and there are quite a few pixels missing

I've informed the owner, waiting for response.

Ed has put the correct file up: https://bama.edebris.com/manuals/solartron/7075/ and then 750_Technical_Manual.pdf(14 MB)

Let me know if there are any "infelicities".

(At a quick look, the djvu file avoided the problem I noticed. I don't know how that manages to halve the filesize, and I'm not going to investigate!)