All posts by admin

So… are those Amazon sourced £20 drill batteries any good then?

In the case of the one I purchased: No.

Warning: this post contains a lot of technical details the average person isn’t going to care much for. Feel free to skip to the end.

I wouldn’t normally contemplate buying an “off brand” battery for anything, but this year genuine batteries for my 14.4V Bosch power tools (which I have a few of) were discontinued.

I could re-pack my batteries with quality cells, but for now, I’ve decided to give a common off-brand battery a go.

There are three main things we care about in a battery:

  • Safety
  • Longevity
  • Capacity

I’m not going to be doing any safety checks here, and as for longevity, come back here in 5 years, I might give an update. For now all we can look at is capacity.

My cheapo battery, purchased from http://www.drillbattery.co.uk/ (which is a front for a Chinese operation which buys crap off Amazon on your behalf and has it shipped directly to you) is advertised with a capacity of 3.0Ah (3000mAh) – the same as the highest capacity genuine battery, a battery which cost three times the money. I’m already suspicious.

While it is possible to get a licked-finger-in-the-air measure of the capacity of a battery with two multimeters, a resistor, stopwatch and an exceptional attention span; testing the true capacity of a battery is rather difficult. This type of test is typically performed with a DC Electronic Load which has the necessary circuitry and software to perform this type of measurement.

For my test I’ll be using a Keithley 2380 DC Electronic Load – one of the best available at the time of writing.

There are two parameters we need to enter into it to perform this test:

  • The discharge rate
  • Voltage we consider the battery to be “flat”.

As can be seen from the above graph the battery is completely discharged at 1.0 V per cell, and as we tend to use power tools until the battery is totally dead (and often beyond) we’ll use that number for this test.

To conduct this test I’ll be using my trusty false charger – which allows me to safely connect the tool batteries to other stuff.

Bosch AL1404 with its insides removed, and puny contacts replaced with the beefier variety of contact found in the tools, those wired to an AWG12 cable with a 50A Mini-Fit Sr connector on the end.

First I’ll run this test on a genuine battery. In this particular case a 3.0Ah battery part# 2 607 335 693. It is 4 years old, and I estimate it has been cycled about 250-300 times. It still performs well so am expecting it to be close to the advertised rating.

For the discharge rate I’ll go for 0.5C (discharge over 2 hours). Discharging too fast will give me a false low reading, and discharging too slow will take forever. I’m not a patient man.

The original battery tested at 2.12Ah. A little disappointing but given its age and regular use, about what we would expect.

Now for our crappy knock off:

From the feel and appearance of it I already have a bad feeling about this one.

And there we have it. The capacity of this battery from brand new, fully charged, measures at just 1.176Ah. Even at a puny discharge rate of 0.5C / 1.5A (a tiny fraction of what the drill would discharge it at) it only lasted 2823 seconds (47 minutes).

An astonishingly poor result and practically only a third of its advertised capacity of 3.0Ah. You get what you pay for. It’s a damn shame the option to pay more no longer exists.

Attaching a 5¼” floppy drive to a laptop, that wasn’t designed to have one attached to it

For years now I’ve had a honking big Tandon TM100 360K 5¼” floppy drive which has been screaming out to be used, and there is something I’ve wanted to use it for – on a “go between” PC so I can shunt files to and from my XT era PCs which I tinker around with.

The easiest way to do this is to have, say, a Pentium II era desktop PC on a network, etc – but that’d be a big lump of junk I don’t want taking up space in my cave.

I could also splash out on a Kyroflux, which would let me attach it by USB to a modern machine but Kyroflux being a specialty tool for data recovery, isn’t convenient in my use-case whatsoever because it only works on entire disk images, doesn’t do drag and drop – it isn’t really what I want. If only I could attach it to my Thinkpad 760 – which I can easily network, and has an external floppy connector.

I’ve been pondering this for a long time now, but knew it wasn’t going to be straight forward.

Difficulty #1 – What is the sodding pinout for that connector?

I had previously spent hours searching around for a schematic, spec, any kind of information on this connector but came up with bupkis. I could buy an external floppy for it, bust it open and have it in minutes but in the UK where I live, most 700 series Thinkpads were sold with internal floppy drives, so these are pretty much unobtainable. I’d have to ship one from the US at massive expense.

Given the ridiculous amount of spare time lockdown has created – I decide to do this the hard way.

Day 1: I completely dismantle the laptop. Inside I discover that it has a half a dozen PCBs connected together with ribbons and board-to-board connectors. It’s a nightmare. I found the floppy controller, but it’s not accessible when it’s all assembled, so I can’t just beep it out. I started to beep out the rear connector to some of the intermediate connectors but it’s a sucky, time consuming exercise.

I eventually discover that only two pins (RDATA, WDATA) are directly connected to the floppy controller, the rest are buffered separately from the internal drive, and the buffers are under the I/O PCB, between it and the CPU PCB which is really difficult to get at. I give up. While I’m at it, I can also get the +5V and GND pins easy enough. I re-assemble the laptop and ponder my next move.

Day 1 ends, I know more than I did at the outset, but unfortunately there’s still a lot of pins to work out.

Day 2: I discover that signals from the floppy controller to the internal floppy are observable on the rear connector, but signals going the other way, aren’t, because of the internal buffering. Doing some accesses to the internal floppy with a scope attached I can get a few more pins – STEP, WGATE, SIDE, DIR.

Day 2 ends, and there’s still more to go.

Day 3: I’ve now written a small DOS program which allows me to read and write the registers in the floppy controller. With this I am able to get TRACK0, WP, DSKCHG and DRVSEL/MOTEN.

A quickly lashed up tool to assist with the identification of pins on the floppy connector

Day 3 ends, I’ve got one signal left: INDEX. All of the inputs have 20K pull-ups on them, so I have to assume that the remaining unidentified pin with a 20K pull-up is INDEX.

Day 4: I’ve got all of the floppy signals now, I’ve got still got quite a few unknown and it bugs me that I wasn’t able to confirm the INDEX signal. It seems to work with it disconnected, so do I really have the INDEX signal? FFS. Pull it open again…

The laptop is dismantled for the second time – to beep out that blasted last INDEX signal. This time the I/O board is removed so I can get at those buffers on the underside

While I was at it, I got another two signals, DRATE0/DRATE1. Not sure what those are for. I haven’t used them.

The Thinkpad external floppy pinout

(On models which use a 26-pin mini-D ribbon connector) – I can proudly announce:

Note: Newer models (i.e. 770) use a slightly smaller D-type connector. This may not be the pinout for those.

Difficulty #2 – Cable

The pigtail coming from the DC shell carries the +5V line, which in future will be used to switch external drive power

There’s nothing special about the connector – it’s a Mini-D Ribbon connector (MDR). I found some cheap ones on eBay labelled “1Pcs SCSI 26 Pin MDR Male CN Solder Plug” which were exactly what I wanted – solder termination, also including a decent shell.

After much pondering as to what would go on the other end – I’ve settled on a standard DC-37 (yes, DC. Not DB) external floppy connector as used on IBM machines in the 1980s.

As one can imagine this cable is quite a bit of work to assemble.

Difficulty #3 – The BIOS doesn’t support 5¼” drives

This is hardly surprising as Laptop BIOSes typically don’t have features not corresponding to official accessories, and realistically, who on earth would have wanted to do that at the time.

I was fearful that there’s be a serial EEPROM in the floppy drive containing product identification data, that I’d be stuffed without the contents of, but that doesn’t turn out to be the case. I attached a generic 3½ floppy drive to my newly made cable – it works. An additional bonus, I can also boot from an external 3½ drive of my choice, should I ever feel the need to.

Back to the matter at hand. My Thinkpad 760 runs Windows NT 4.0, which has its own floppy disk driver, not using the BIOS at all, except, it does interrogate it to find out what kind of drive is attached, which is where we have a problem.

The type of drive is retrieved by NT very early on via an INT 13h call to the BIOS, that data gets stashed in a holding area, finally retrieved by floppy.sys a bit later on when it starts up.

Hacking the response to the INT 13h call would be pretty difficult, so instead I’ve gone down the path of hacking floppy.sys, whose source code is in the NT4 DDK.

After days of screwing around with my floppy NT driver hack not quite working I change approach – and go hack NTDETECT.COM – an obscure 16 bit executable which is used during the NT boot process to probe the BIOS for hardware information, including attached floppy drives. Using the NT leaked sources, this turned out to be absurdly easy, however re-compiling it most certainly was not.

Very little modification is required to force B: to 5.25 360K

Eager hackers have gone to the trouble of re-creating Microsoft’s build environment from the early 1990s. Even with all of the know-how in the public domain this is still an awful lot of work, so I put together a custom build environment for it instead. My new NTDETECT is quite a bit bigger (35K) than the original (26K) for some reason, but it works.

There is something rather satisfying about having a B: drive – especially when it’s a 5¼”

Read data troubles

I noticed fairly quickly that I was having quite a lot of read errors, so I got out the scope to have a look at RDATA

It seems that the 20K pull-ups inside the laptop aren’t sufficient for my cable rig. I had to add 1K pull-ups on the RDATA and INDEX signals to get things looking a bit cleaner.

Would you believe it…

The day after I complete this, after nearly two years of watching eBay – a first generation Thinkpad external floppy appears on eBay UK. I nab it for £6.50.

Let’s crack it open and see what’s inside.

It turns out the external floppy is just a caddy which can have the internal floppy drive inserted into it. Busting it open actually tells us nothing about the pinout of the rear connector, because we don’t know the pinout of that large 100 pin connector inside it.

To get the pinout I had to detach the Mylar FPC which runs between that connector and the physical drive, which has a 26 pin connector on it, which we do know the pinout for, because it’s a TEAC FD-05HF-8830 for which the datasheet is available. Finally, I can beep this thing out properly.

What did I learn from it?

The only new information I got is that the HDSEL signal is actually present on the external floppy interface. It’s a bit of an odd one. It’s an output from the drive which is LOW when a 720K disk is inserted, and HIGH when a 1.44MB disk is inserted.

What’s the point of it? Not sure. Generally floppy drivers do this detection in software, first trying the highest format for that type of drive, if that fails, step it down, try again, and again and so-on.

Because it’s not connected to the FDC (it’s connected to an undocumented IBM ASIC) nothing much would be able to make use of it anyway. It’s probably used by the BIOS to make format detection a bit quicker, but the BIOS is also able to do format detection in software, so, really not needed.

Other than that, a few mystery pins were found to go to NC pins on the drive. I think it’s for some kind of drive identification mechanism, which isn’t used on the drive I have.

A general note about using 5¼” drives on modern computers

You may be sitting here thinking: “I don’t have one of those ports on my laptop. Why couldn’t he have shown how to do this with an interface I have on my computer”. I hear you.

I would also dearly love to to just plug this drive into my Windows 10 laptop, have that 5¼” icon show up in My Computer and use the drive like a modern USB floppy, heck, even running one from a parallel port would be amazing.

Unfortunately, when the writers of the USB Floppy Specification decided only to include 720K and 1.4MB formats on 3½” drives in the standard, they made this extremely difficult. Because of this, a whole new software stack would have to be written, for which ever OS you want to use (it’d be easier on Linux – being open source you could modify what’s there). For Windows this would take even an experienced driver developer several months. Ditto for any other interface.

Maybe, just maybe, the source code for a newer Microsoft operating system (with the USB floppy driver) will one day find its way into the public domain. We can then modify it to support 5¼” drives – 90% easier than re-writing it from scratch. Until this day comes, hacks like the one this article describes are about as good as it’s going to get.

Failing that, there are other possible solutions. It is theoretically possible to attach a Super I/O (with floppy controller) to the LPC bus (assuming your mainboard has a header for it – many do). A bit of new hardware would have to built, as well as a bit of software hackery, but this would still be fairly easy compared to building a whole new soltion.

I have some EPROMs, from Russia

What is our fixation with chips produced by the Soviet union by the help of industrial espionage? Or is it just me that likes to collect these?

Today, I’ve received another consignment of them:

This is a set of 4 U552C’s – the Soviet version of the Intel 1702A which I recently built a programmer for.

Before I get started, there is a familiar problem which is that the pin spacing of these chips is a metric 2.5mm, not the usual imperial 0.1″. Engineers from the USSR apparently felt the need to correct the oddities of the imperial past. Sockets with this pin spacing are more difficult to come by than the chips that plug into them, but today I’ve got some, so I’ve built a little adapter:

Just for a laugh – I though I’d try program them using the x86 build of that programmer on another Soviet chip: the K1810VM86 – a clone of Intel’s 5MHz 8086 processor.

The test setup – an un-holy mix of components from either side of the iron curtain.

On the first run I nearly burnt out an EPROM because I’d forgotten that the x86 HvEprom build is hard coded for a 10MHz CPU, whereas the K1810VM86 only ran at 5MHz, so I had to go back and re-do all of the timings.

They all programmed and verified no problem.

The last test is to pop them in my 1702A reader, and we can see that the ASCII letter ‘K’ is in the first position as expected.

ROM Adapter for running ‘B’ firmware on a Keithley 2001 with a pre Rev-K digital PCB

If you have one of these, you will have discovered that the last revision of firmware than can run on these is A08, because the single PLCC-44 socket can only hold at most a 4 mbit EPROM.

Later PCBs have two EPROM sockets allowing them to run newer (larger) firmware images.

Strictly speaking however, there’s nothing stopping us from running ‘B’ firmware on these older units, if only there was some way we could get a little more ROM space.

The EPROM found on pre rev-K PCBs

There’s no PLCC-44 EPROM we can put in that socket which can hold more than 4 mbits, however…

My Keithley 2001 ROM adapter

I have built an adapter which makes this possible. It’s pretty simple. It’s got a PLCC-44 plug and a PDIP-42 socket to allow, for example, an M27C322 (32 mbit) to be used in place of the original ROM. I used a 42-pin EPROM because they’re common, dirt cheap and easy to program with cheap hardware.

It is secured in place using a hex spacer mounted on an existing screw hole on the digital PCB.

ROM adapter installed

After having built this I discovered that my choice of PLCC plug (Winslow W9303) is made (and only available) in the UK – so this project is probably a fat lot of good to anyone else, but the none-the-less the point is proven here, it is possible to get to the latest firmware on these oldest units.

One little wire mod

Sadly we do have to make a little change to the PCB to facilitate this new adapter:

An extra address line from the CPU has to be connected through to pin 1 (unused) of the EPROM socket. The other end of this wire is visible in the previous image (it is connected to the 5th pin from the top left corner of the CPU). Because this pin is NC on the stock EPROM it can be left in place if you wish to return to the original Axx firmware.

Calibration data troubles

After having booted my 2001 with (more or less) the latest firmware (B16) I discover that the calibration data hasn’t loaded, and on top of that it’s gone and wiped the calibration memory on boot-up. FFS.

xdevs.com went to the trouble of devising an ISP arrangement to deal with calibration EEPROM tasks. I moved mine to a socket.

It turns out that the format of the calibration memory is different between the original firmware (A08) and what I’ve upgraded it to (B16). Referring back to the great oracle of Keithley 2001 related information (xdevs.comone of the images he’s got up there (link) appears to be in the same format as what I’ve upgraded to, so I edited the binary to put my serial number into it, flashed that into mine, and I’m back in business, albeit with someone else’s calibration coefficients. Eh. I don’t care, my cals were a decade out of date (long overdue to be re-done) anyway so no big loss.

Once the calibration EEPROM is changed to ‘B’ format it’s no problem to change between all ‘B’ versions. I do not know how to convert them unfortunately, that would be a separate project.

It works

A very early Keithley 2001 booting the latest firmware

ROM

If building one exactly like mine – the 27C322 is large enough for 4 B-Revision ROMs. Jumper settings will determine which to boot.

BOM

The only tricky part to source is the PLCC-44 plug – which is a Winslow W9303. Unfortunately these are only available in the UK and there is no alternative. Sorry about that!

  • 1x Winslow W9303 PLCC-44 plug
  • 1x 0.1 uF ceramic capacitor
  • 1x M27C322, M27C160 or M27C800 EPROM
  • PDIP-42 socket
  • 2x right angle jumpers
  • M3 15mm hex spacer + screws + washers

If using an M27C160 it’s only possible to fit two firmware versions in, which would be toggled by the A19 jumper.  The A20 jumper should be set to ‘L’.

In the case of the M27C800 – only one firmware image will fit. The A19 jumper must be set to ‘L’, and the A20 jumper must be set to ‘H’.

Building one

Attaching LTO Tape drives via USB or Thunderbolt

TL;DR

If you don’t have a computer with PCIe expansion slots – there are no easy and inexpensive ways of attaching an LTO tape drive to it.

Introduction

For quite some time now I’ve been doing my backups (including all of the content on this website) by Tape. Why? Quite frankly I just like them. If like me you buy older generation drives second hand for personal backup, I find it actually works out cheaper than having say, two or three (or more) USB hard disks. The low cost of the media also allows me to have a history of my data (say, a copy from each year), because sometimes, things get lost or corrupted, I don’t realise it, then end up overwriting a good (backup) copy of data with bad (or no) data.

There are other significant benefits of tape – particularly in the robustness and simplicity of the media. Tape media is purely mechanical (aside from the RFID chip on the side) meaning that there’s no electronics which can be damaged. The physical spool of tape is also very robust, unlike the glass platters of a hard disk.

Dropped your tape and smashed it? Somehow managed to kill that RFID chip? No problem. Just buy another tape, undo the four Phillips screws on the bottom and transfer the tape spool to another shell, and you’re back in business. Good luck doing anything like that on a hard disk!

Happened to be subject to a massive electrical surge or lightning strike while your one and only copy of data on a tape was in the drive? Once again, no problem. Tapes have no electrical connection to the drive whatsoever, even during operation.

When talking about backup, IT administrators often use the term “air gap” meaning that data cannot be wiped out by a virus or other accidental software or power incidents. Tape, today, remains the undisputed the king of air gaps.

SFF-8088 SAS Connection on the rear of a tape drive

LTO Tape drives (these days) come with either Fiber Channel, or SAS interfaces – there are no other options. This makes them a little difficult to attached to (For example) a Mac, or any Laptop. Desktop PCs & Mac Pro’s are not an issue because you’ll likely have PCIe slots where a SAS or Fiber Channel host adapter can be installed.

I don’t actually have any PCs with PCIe slots in them anymore, and haven’t had for quite some time so have had to confront this issue myself.

Drive selection

Before we get into the topic of attaching drives, it may be worth considering the drives themselves – assuming you haven’t already purchased one.

Top: IBM LTO-7 Drive. Bottom HP LTO-6 Drive. Both branded Quantum

Despite a range of different brands, there are two manufacturers of LTO tape drives: HP & IBM. We can see this in the picture with our top drive having an IBM style blue button like what is found on their servers and desktop PCs, and the bottom drive featuring HP’s corporate font under the LEDs.

Dell models feature a slightly different chassis however they are still IBM drives.

Both are very well engineered as you can imagine for the large price paid for these units purchased new.

There are some things to consider (in the context of tabletop drives) when choosing one over the other:

IBM Tabletop Drive

Pros:

  • Slimmer more aesthetically pleasing design
  • Good full featured LTFS implementation for Windows
  • Easy single handed tape insertion – like a VCR
  • Slightly more pleasing operational noises
  • Very robust all metal chassis
  • Easy to dismantle (four screws on the underside, cover slides off)

Cons:

  • Extremely noisy high-RPM 40mm fan on rear, always runs and restricts the drive to server room use only
  • Terrifying high pitched sound when loading tape
  • Mine seems to jam during loading about 1 in 10 times – requiring a second attempt
  • Drive is very long (340mm), may not fit on some shelves
  • “Soft” power switch. Power supply is still on even when drive is off
  • An extra $1000 for an IBM branded one (grumble)

HP Tabletop Drive

Pros:

  • “Fat” design thermally superior to IBM’s. Uses larger, quieter fan
  • Fan is only on when tape is inserted, goes unto standby mode when empty (thanks to a fan output connector on the drive its self) – likely applies to LTO-6 drives and earlier only.
  • Shorter than IBM drive (300mm)
  • Full AC power switch on front (I think)

Cons:

  • Ugly
  • LTFS Implementation for Windows less featured than IBMs (LTO-6 and earlier). Depending on your use case you may also want to supplement it with this.
  • Operational noises a little more irritating than the IBM drive
  • LTO-6 and earlier drives have a “flap” which has to be lifted up to insert a tape making it a two hand job. If you use the drive a lot – this is going to piss you off.
  • Outer plastic chassis not as robust as IBM drive (there is also an inner metal chassis).
  • Complicated chassis design must be dismantled in a very specific way to avoid breaking internal plastic clips.

Side note: I have made several references to “LTO-6 and earlier” here. This is because the last generation of drive made by HP was LTO-6. From 7 onward, HP drives are re-branded IBM drives deployed in HP’s traditional black plastic chassis.

Option 1: USB

Some time ago a product existed to convert USB to SAS:

They originally sold for around US $250 but likely due to the decline of SAS usage in general, do not appear appear to be made anymore. They can still be found for sale, for typically very high ($500+) prices. If you can get your hands one for a good price, this may work out, but don’t  count on being able to get another.

4x Internal SAS to SFF-8088 cable

Assuming you can obtain one of these rather exotic items, you would then need an internal SAS to SFF-8088 cable, bearing in mind that SFF-8088 carries 4 SAS lanes, you’d just connect your USB to SAS adapter to port 1 – which is what the tape drive will be internally connected to. This would be a workable setup – but a bit ugly.

Due to the obscene cost and obscurity – I would not recommend going down this path.

Option 2: Thunderbolt (buy one pre-made)

If you’re not technical and/or not on a budget, there are a few ready-made Thunderbolt drives. These internally contain a PCIe to SAS host adapter as I demonstrate below. Expect to pay a $2000-3000 premium for this convenience. A product like this uses an ATTO or Highpoint SAS controller which is required for compatibility with macOS X – the primary target market for these products.

Option 3: Thunderbolt (pre-made  SAS adapter)

This will be a little cheaper than buying a pre-built drive but still a lot more expensive than the option below. You’ll have to source the appropriate SAS cable. More about that below.

Option 4: Thunderbolt (make your own)

This is a far more sensible (and cheaper) option. Because Thunderbolt carries PCIe we can use (For example) an eGPU enclosure to carry a PCIe SAS Host adapter.

LSI SAS9207-4i4e SAS Host Adapter

I personally use an LSI SAS9207-4i4e. The LSI SAS9207-8e (two external ports) would also be suitable, as would many others. I have chosen this because it is a fairly recent adapter, which also has the very same SFF-8088 connector found on the tape drive. I got this adapter off eBay for $30.

The full setup

To the left we have a PCIe Thunderbolt enclosure containing the Host Adapter. There are lots of Thunderbolt PCIe enclosures on the market, you can pretty much just pick the cheapest one as a SAS Host Adapter is not a very demanding card to install in one. Single slot enclosures seems to be the cheapest at around $200 at the time of writing.

To the right is the tape drive.

If we go down this path, in addition to only having spent $300 (excluding the cost of the drive), we also have the bonus of having a few new items that have other uses. For one the Thunderbolt enclosure can be used for other PCIe cards, also the SAS Host Adapter can be used as a very high performance connection for SATA hard disks too. There are many different cables and enclosures which can make use of this.

SAS cable selection

SFF-8088 cables come either x1 (one lane) or x4 (four lane) variants. A tape drive only has one lane so either an x1 or x4 cable will be OK. x1 cables are considerably thinner and lighter than x4 cables.

You can also buy fairly long SAS cables too. Bear in mind that tape drives are quite noisy, you may want to consider buying a longer cable (up to 10M / 33ft) so you can put the drive somewhere it’s not going to irritate you.

Left: SFF-8644 connector. Right: SFF-8088 connector

There are multiple type of external SAS connector presently in use and these days SFF-8644 is beginning to replace SFF-8088 despite it still being common on tape drives. It’s not a problem if you end up having to buy a SAS Host Adapter which has a SFF-8644 connector on it, you’ll just have to buy a cable which has the appropriate connectors on each end.

Operating system (Windows/Linux)

On Windows 10 I did not have to install any drivers for the Thunderbolt enclosure or SAS Adapter – it all just worked.

The only driver I did have to install was for the tape drive its self.

Linux is even easier with all of the necessary drivers built into the kernel.

Operating system (macOS X)

There will not be any driver issues with either the Thunderbolt enclosure or the tape drive – they will work out of the box.

The issue arrives with the SAS host adapter. Unfortunately native SAS support is quite poor in macOS with only a handful of ATTO and Highpoint chipsets supported. It is these chipsets which are found in expensive ready-to-go solutions I have previously mentioned.

ATTO ExpressSAS H644

If you are lucky you might get one on a PCIe card for a decent price. Pictured above is an ATTO ExpressSAS H644 which you conceivably may be able to pick up second hand for a less than bank-balance-busting price but I wouldn’t count on it. Honestly, if you’re an Apple person, it’s likely not worth the hassle for you. Magstor’s $5500 drive will work a treat.

As it happens I do own a Mac, and I’ve managed to pick one of these up for a very reasonable price second hand, so let’s try it out…

ATTO ExpressSAS H644 installed in my thunderbolt enclosure alongside my existing LSI adapter

LTO tape drive detected through ATTO SAS adapter on macOS

The ATTO worked out of the box. I didn’t have to install any additional driver packages. LTFS detected the drive and mounted it just fine. LTFS is not a very good “experience” on macOS – and since experience the reason you have a Mac, you’re probably going to want to look at some commercial backup software to run your tape drive, of which there are many choices.

As was to be expected, my LSI adapter wasn’t detected by macOS, nor are there any drivers available for download.

Hot plugging

I found that it all works like a charm. You can either hot plug the whole setup through the Thunderbolt cable, or you can just disconnect the SAS cable (or even power off the drive), as SAS is also hot-pluggable, if for example you use your SAS host adapter and PCIe enclosure for other things (as I do).

Fiber Channel instead of SAS

You could also substitute a PCIe SAS adapter for a Fiber Channel host adapter in your Thunderbolt enclosure if that’s the kind of drive you happen to have. You can do your own research on that. This is going to be a lot more complicated but the advantage of this option is that you could have your tape drive a very long way from your PC.

Are SATA adapters of any use?

For the most part, no. If you are starting with a SATA/eSATA controller, there is no way to adapt to SAS. You must start with a SAS controller. SAS controllers however, support either.

For the benefit of skim readers, and/or the impatient, here is the previous sentence spelled out unequivocally

There are some scenarios where SATA to SAS/SFF adapters are useful – for example:

Let’s say you have a SAS controller with a bunch of SATA HDDs attached via a SAS to SATA octopus cable, and you happened to have a spare port – it is possible to adapt that port back to SAS, to attach to a SAS tape drive.

Essentially – it doesn’t matter if intermediate the cabling or connectors are SATA, just so long as you have SAS hardware at either end. SAS and SATA cables & connectors are eletrically the same i.e. 2x 100Ω differential pairs each – differing only mechanically.

It also doesn’t matter that you are using a mix of devices on one host adapter – so long as you’re not trying to put those devices into a single RAID volume – SAS controllers don’t care.

Because the tape drive will only be using one port on the SFF-8088 connector, you can connect that one spare SAS port to port 1 on the adapter.

Older laptops with ExpressCard slots

There are some products which adapt ExpressCard to PCIe which would allow a SAS adapter to be attached.

There are some examples of SAS ExpressCard adapters:

Sonnet 4-Port Tempo SAS Pro ExpressCard/34

Very few true SAS ExpressCard adapters like the above were ever made. None are made anymore and anyone who has one may expect a high price for it.

Most products resembling the above are 4x SATA controllers with an SFF-8088 connectors, which is of no use for tape drives.

Keysight parts division: Clearly not profitable

Keysight 34401-88304 rear bezel in original packaging

Half width Agilent / Keysight instruments made up until the late 2000’s had a plastic back bezel (34401-88304) which the rear bumper attaches to. They are frequently broken in electronic labs because the rear bumper often slides off the end of the desk. Someone then grabs the instrument to pull it forward, not knowing it is caught on the back edge of the desk, destroying this part in the process. After the 100th time trying to glue mine back together – it was time to get a replacement.

They are frequently seen for sale on eBay for $100 or more, but the part’s true cost from Keysight is only around $12 – if you can be bothered with the hassle of ordering it directly from keysight – Which brings me to my point here.

Where U.S. Customers can order with ease directly from the website – getting this part in the UK involved 12 emails, 4 phone calls, to people in 3 different countries, and hours of both mine and Keysight’s time. In the end I was charged just £14 (including VAT and shipping). Ridiculous. Keysight likely spent £1000 of their own time getting me this part.

I am not faulting the service here – Given the rigmarole I am astonished at how little this cost! It even comes in a ludicrously oversized box:

I ended up having to collect in person from Mayfair post office –  who are never thrilled to have to hold large items – resulting in the spectacle of me carrying this large geeky box through the crowded streets of Central London (not that a single person would have recognised that logo).

Yeah… just the shipping alone would have cost £14.

In the end a rare example of dealing with a big brand like this, without punishing my credit card in the process. I also have some Keithley parts to order this month. That is not going to be quite so cheap.