Thank you all for all these suggestions. I have examined them and what I was doing.

.

The insert is the sort with three sizeable triangular tips, so when cutting a fairly small thread these penetrate the work only partially (just over 0.03″ for 3/8″ BSF). This type will therefore leave a sharp crest. So would a plain Vee-tool unless ground specifically for the thread.

I find the thread depth from the Zeus book, and make a small root-diameter extension as a guide as well as using the dial.

Usually I do cut the thread to near-depth than finish it with a die, but even with the die open in the holder, this time it still spoilt the thread.

.

My die-holder is the type with the through-hole for a handle. Normally I use a rod bearing on a safe area of the lathe.

Its tanged arbor does not hold very tightly in the Myford’s tailstock so the action of the die pulls it out. I don’t know if the tang stops it going back just that tiny bit more. Other tools usually hold properly. It grips on the Harrison lathe, yet neither lathe has a tang-slotted tailstock barrel.

.

The Harrison’s bed is worn at the head end, with a definite step onto the gap-filler. My first attempt gave much more accurate diameters, because to save material I used a fairly long piece supported in the fixed steady, putting the saddle a long way down the bed.

Perhaps I should not to try to turn the full length, but finish the thread section then the shank section separately, so controlling the depth over a much shorter length. It won’t remove the taper completely but will reduce its effect.

.

The saga continued, concluding in finding an enormous amount of play in the Myford lathe.

I took the embryo bolt still in the 4-jaw SC chuck to the milling-machine and screwed it onto the dividing head with the tip of the bolt supported by the matching tailstock, to mill the two pairs of flats. At the point needing 0.04″ total removing, 0.02 a side, I raised the knee 0.015″. The result measured not 0.03″ gone, but well over 0.03″ still to clear!

I was holding the cutter in an R8 collet. It seemed fully tight but I gave the drawbar an extra tweak.

The next cut worked, taking it down to 0.503″ wide. I wanted to finish at ~ 0.495″. Raised the knee 0.002″, skimmed the two flats, measured…. 0.474″ .  That’s rubbish!

Thoroughly dispirited: this was the second time that had happened. I cut the two steps (the stepped profile is not strictly necessary on a one-piece Tee-bolt, but never mind), took the chuck and work back to the lathe to part off the bolt.

I had not cut the preliminary groove deeply enough, so the two flats gave the 3/32″ wide, HSS parting-tool a sizeable interrupted-cut. Despite the small size and using very low speed, this lifted the saddle and the work visibly on each bump, until the parting-tool had gone below the flats. Then it started playing up again, and I had to saw the last bit.

The flats were too tight. I had to ease them with a file.

This Tee-bolt finished the correct length with a beautifully smooth shank within 0.001″ of size… but a very poor thread, rough, inaccurate and slightly unsymmetrical flats that produced ugly scars on the shank, and still tight in the Myford Tee-slots it is supposed to fit.

.

I had to stop there anyway as I was out for the afternoon, but I locked up the workshop knowing after over 50 years in model-engineering I cannot even make a simple Tee-bolt to Loan Section quality looks and function; let alone competition “Commended” level.

Hopper –

Nothing wrong with the die I used. It was not some ‘flea-Bray’ cheapie but a little-used one from a reputable supplier. It was my using it that failed, and I don’t know why.

I was also using very good, free-cutting mild-steel.

.

Jason –

That is what I was effectively doing: cutting the thread deeper than the nominal to compensate for the sharp tool point. I could see it grooving the witness stub on the end, at nominal root diameter.

The test nut would screw on for a couple of turns then jam. Yet this bolt’s brother, differing only in length, had worked, perhaps because I had not tried to screw-cut the thread.

Hopeless because I need also make two threads some 3″ long and this rate I can see the only way is to use studding and modify the rest of the design to suit!

I could measure the flats in progress only with the work in place, so carefully brushed all the swarf off and moved the table clear to allow full access for the micrometer. I also measured the thing several times after I’d finished it, but whatever was happening doesn’t make sense.

It’s possible the cutter was slipping in the collet, I suppose, but there was not enough room for the Autolock collet next to a dividing-head.

Or something was springing, maybe.

I just don’t know.

.

I’m afraid you mis-read me.

The nominal A/F size is 0.5″, to fit Tee-slots milled in the factory – when that was in Beeston, too.

So I wanted the head close to 0.495″ A/F – as on ones that worked.

Instead, despite all care to measure the head as I went, this specimen, and a previous one, came out at 0.475.

Yet I’d only raised the work by two thou to take that last stage down to that very gently. How does 0.002 suddenly become 0.01″ ?

””

I avoid videos because I can’t learn from them, even if they seem reliable. Most seem simply demonstrating skill, and in ideal conditions, not teaching it. That was even before Google etc. started ruining videos with constant advertising breaks.

While I am sorry but your last paragraph was rather hurtful. Of course I don’t rely on old text-books to tell me about things that did not exist in their day. I use them only where their information still covers the particular task, tools and materials – but a lot is still relevant.

””’

Actually it is becoming harder to find anything useful on the Internet. A company web-site name always works;  but otherwise unless you know exactly the right search phrases the system will use, it returns screenfuls of ads, usually not at all relevant (probably by weak word-matching).

I have just broken off to test this. Entering “indexable tool standard specifications” did soon reveal the goods. The mysterious ISO has its own web-site – but that only indexes information available only by sale only to they who move in the Right Circles.

Luckily I also found – and book-marked in my “Engineering” list – the Seco tool-makers’ site with comprehensive, clear tables of the ISO-standard designations etc. That’s mainly for the holders but does show the inserts’ codes too, at least partly. I’ve also a copy of a printed Sandvik catalogue with similar information.

Fine… doesn’t help though. I could have been using brand-new machines and fittings, and still used 18 inches of steel to fail to make a decent-quality Tee-bolt 2 inches long.

Sighs…

I have not changed anything.

Measured the blank, subtracted 0.500, halved the difference.

Took an initial cut, re-measured – it needed 0.040″ taking off.

0.02″ a side.

Set 0.015 cut, for safety…

That should have reduced the work by 0.030″. It had not!

So I examined everything to find why. No obvious reason but I did give the drawbar an extra tug with the spanner. I no longer believe that plain friction collets (R8, ER, etc.) are right for holding milling-cutters for accurate depths; but there was no room to use the Autolock collet.

Next cut worked, bringing the work to 0.503″, give or take a thou. (I remember because this was proving such grief to do!)

Table up by 0.002″ to leave the piece 0.499″ wide.

I was aiming for 0.495 as close as I could so expected one light skim more would do it.

So re-measured for that finishing skim: not 0.499 but that blasted 0.475″.

.

The micrometer is one with a mechanical analogue display, so difficult to mis-read, and is reliable. That error suggests mis-counting a conventional 25/1000 scale but by co-incidence because I did not use that type.

Somewhere, somehow I lost 0.012″ a side – and I do not know why.

.

Knowing now that these Tee-bolts do not need steps, I considered side milling the flats using either a Stevensons Block with an ER collet, or the jig-borer and a 3-jaw chuck; but the cutter would probably act like a gear and rotate the work-piece or chuck. Almost easier to file the blasted things! If I can’t make a simple Tee-bolt, I have little chance of making anything that really needs be accurate, like engine components.

.

I am afraid that if anything I build works, such as that long-suffering steam-wagon I should have completed years ago despite no drawings, it will only be by a miracle of work-rounds, inaccurate parts, re-works, holes filed oval, surplus holes from changes to overcome problems not foreseeable, tatty threads, loose fits, poor finishes and plain old-fashioned bodges… Despite every care to get every part right first time, at every stage, too.

……

Just having a post-lunch cuppa and trying to motivate myself to have yet another go with the last bit of that length of steel. It’s just long enough with great care. I’ll search my collection of chasers but don’t think I have any of the right size.

I am following a drawing that calls for 1/2″ – no decimals – so didn’t think to question it. The two T-bolts (different lengths) are to fit the ML7 saddle, and I found slight width variations on the upper part of the slots on mine. The bolt enters one but is tight on another.

The tightness on the bolt was on the steps I’d given it – since I am making two things in parallel, I looked at the salient dimensions for the essentially similar components but failed to spot the ones to fit the Myford are plain head type. That would not matter if the thing had actually fitted – it didn’t because I could not make it well enough!

I did not want the head too narrow because that reduces what little contact area they have anyway.

What was really upsetting was being unable to achieve the result I wanted, and trying to make just T-bolt had become a nightmare of my mistakes, worn machine-tools and things just going wrong for no clear reason.

……

I finally succeeded this evening.

I had only just enough steel of that bar left so had to fit the spindle back-stop to help hold the material in place, and support the end from the tailstock. Unable to rely on turning the full 2″ length of 3/8″ parallel I roughed it down to under half-inch then took the thread section alone down to 0.371″ diameter.

Using the parting-tool I cut a run-out groove of .032″ deep from that so in theory probably enough to accommodate the non-truncated screw-cutting tool. It proved about right.

This was where I solved –

 Problem 1 – the loose-fitting arbor on the tailstock-die holder. It had spun in the tailstock at some time, gouging it just enough to prevent full engagement. I cleaned it with a small file and and that was sorted. However I was not going to use it with such a delicate set-up…

and….

My mistake 1. It does help to use the holder with the Whitworth-form insert, not the 60º one as I found I’d done. No wonder the BSF die had munched through so much metal. It was trying to trim a UNF thread!

.

Then My mistake 2: I thought the developing thread looked a bit fine for 3/8″ X 20 BSF. I’d not set the gearbox back from the fine feed, and was now cutting a 28tpi thread. Fortunately the cut was still shallow and the error disappeared eventually.

Tickling the thread down and lightly running a smooth file over the crest eventually produced something for the nut to run on almost as it’s meant to! I have no idea how many passes and tests this process took but the tool seemed to find steel to remove every time.

Now turned the shank, which will eventually fit a reamed hole. We can but hope.

Experimenting while still well above finished size I found an insert tool that gave a slightly better surface than the HSS one doing most of the work, though itself giving a very good finish. Very gently, without letting the tool rub, I crept up on that 0.375″ figure, using the self-acting feed on its lowest rate.

It is too! Well, in the centre of the 1.25″ length it is.

The wretched thing is still tapered by 1 to 2 thou over just that length!

.

So Problem 2 – Why can’t I turn an accurate parallel?

.

I’ll just have to live with it. I turned the material round, reduced the head to thickness then cut the flats. I think I found the source of –

Problem 3 – weird widths. Cutting the wider flange for the stepped head seemed to have left narrow ridges, and I was probably measuring either on those or the lower surfaces without realising it.

This time I used a slightly different technique and anyway was milling a much narrower flange, and this time did indeed reach that nominal half-inch less five thou.

.

So after 4 attempts I now have a Tee-bolt that is to drawing and close enough to work as it’s intended – but would fail any proper QC inspection.

What a ruddy palaver, and convincing me I’ll never create any significant project to decent standard.