Review: SanDisk Ultra Dual Drive m3.0 128Gb USB Flash Drive

It doesn’t happen very often, but once in a while, I need a new USB flash drive. This time around, I was looking for something cheap, of reputable make and 128Gb capacity. Interestingly, looking around my local computer shop, the cheapest option was a SanDisk Ultra Dual Drive m3.0 for AU$42. This was a surprise since this is a special USB drive with a regular USB Type A and a micro B connector for on-the-go (OTG) use and it was a few dollars cheaper than a regular drive with just the normal Type A connector. As a result, I just had to get one!


As with most other SanDisk products, it’s packed inside a plastic bubble encapsulated by glossy coloured cardboard with a hanging ear. The unit is called a SanDisk Ultra Dual Drive m3.0 and seems to be a new revision of their previous Dual Drives.  The unit advertises itself as a flash drive for “Android Smartphones” with up to 150MB/s read speed over USB 3.0. Otherwise, it makes no speed claims. It comes with a 5-year warranty, and is supported by their own free “SanDisk Memory Zone” app available on Google Play.

Splitting the card apart reveals a folded warranty guide, an electronic anti-steal (EAS) sticker and the sealed bubble the drive comes in.

The drive is a relatively small unit, only the size of the two connectors back-to-back, surrounded by a smoked grey plastic frame. The grey central portion is attached to the body of the drive, while the frame slides down to reveal each port (to prevent simultaneous dual-connection which may cause device damage). The capacity and serial number are etched on the top of the Type A connector, although the connector is a modified type with no holes on the top.

The grey plastic extends to one side to help slide the drive into the desired position. This is held by small detents in the outer ring portion, seen at the top of the image.

The opposite side provides a loop for a lanyard, not supplied.

A look from the underside shows that two dimples are provided on the Type A connector to improve retention in ports, however, they are not completely opened. This is because the integrated “chip” probably resides in the lower-half of the Type A connector shell.

We get a hint of this when we look at the blue USB 3.0 connector insert in the Type A plug. The plug has raised lips to retain the storage chip inside the drive and the connector insert appears to have been soldered to pads on the chip itself. This is necessary, as unlike USB 2.0 connectors, a functional USB 3.0 connector cannot be made solely from pads on a PCB due to the springy connectors for the USB 3.0 differential pairs.

The other side has the micro-B connector, along with cut-outs for the retention tangs. It’s quite obvious, but worth mentioning, that this connector will only provide USB 2.0 speeds in OTG operation.

This is what it looks like, extended for use with a computer.

This is what it looks like when extended for OTG operation with a phone or tablet device. Note the design includes a lengthened micro-B plug to ensure that it can plug into recessed ports and ensure that it can be used even if a phone case is in use.


As usual, the drive is pretty much plug-and-play when it comes to desktop and laptop computers, being just a regular USB Mass Storage Device.

It seems to have a VID of 0781 and a PID of 558A.

In its default format, it is supplied in FAT32 for broadest compatibility, but will not support single files larger than 4Gb. The default format provides a total capacity of 114GiB (122,980,499,456 bytes), with some already used.

This is due to the inclusion of a copy of the Sandisk Memory Zone app as an .apk file and a Quick Start Guide PDF on the drive. This could be useful if you intend to use the app on an Android device, although the latest version can always be downloaded from Google Play and other file managers should be able to use it provided your device supports USB OTG storage.

Performance Testing

Prior to commissioning any drives for data storage, I like to test them to determine their performance and to ensure they are performing correctly. Testing was done on my new Ryzen-based workstation using the on-board chipset hosted ports under Windows 10.

HDTune Pro

HDTune showed an average read speed of 171.4MB/s, well in excess of the 150MB/s claimed on the package, making this drive a fast reader especially considering its price.

Testing the write speed with HDTune shows an average write speed of 13.4MB/s, which is slightly disappointing. This only makes it about as fast as a low-end Class 10 microSD card.

Filling the whole drive with random data and verifying it took a very long time, even longer than expected than the write speeds benchmarked. Maybe the drive is sensitive to the data written, the block size or alignment. At least no data appeared to have been corrupted.


Because I made a small error in my test protocol, I did the HDTune tests above first, which wiped out the formatting on the drive. As a result, I discovered that the drive’s performance seems to heavily depend on the formatting of the drive.

FAT32 USB 3.0

From left to right, the ATTO results of the drive formatted as FAT32 in Windows 10, as FAT32 VFAT using mkdosfs v4, and as FAT32 VFAT unaligned with minimum reserved sectors using mkdosfs v4.

The Windows 10 FAT32 alignment is at 2048 sectors and it seems to offer peak speeds of 184.5MB/s read and 16.4MB/s write. Going with VFAT pushes this to 184.7MB/s read and 21.4MB/s write. The most space-greedy VFAT format results in 184.7MB/s read and 16.4MB/s write. As a result, it seems mkdosfs v4 gives the best speed in VFAT mode with default settings.

FAT32 USB 2.0

Repeating the same with the drive connected as USB 2.0 resulted in the Windows 10 FAT32 formatted drive reporting 39.9MB/s read and 6.3MB/s write, the FAT32 VFAT drive reporting 39.6MB/s read and 24.1MB/s write and the maximum FAT32 VFAT unaligned drive reporting 39.7MB/s read and 6.2MB/s write. As a result, it seems the drive’s performance suffers in USB 2.0 mode especially when it is operated outside of its preferred alignment.

NTFS USB 2.0 and 3.0

With Windows 10 formatting the drive as NTFS, it doesn’t seem to exhibit any negative effects despite the partition starting at sector 2048. In USB 3.0, it offers a peak read of 180.9MB/s and a peak write of 21.5MB/s. In USB 2.0, this reduces to 39.9MB/s read and 24.0MB/s write.

exFAT USB 2.0 and 3.0

Using Windows 10 to format as exFAT doesn’t seem to cause any alignment issues despite having a partition start at sector 2048. In USB 3.0, it offers a peak speed of 178.9MB/s read and 21.5MB/s write. In USB 2.0, this reduces to 39.9MB/s read and 24.2MB/s write.

Factory Format

Luckily, before publishing, I purchased another fresh sample of the drive. I determined that the drive is partitioned as MBR FAT32 with an alignment of 32 sectors.

The performance from ATTO over USB 3.0 shows that the second untested/unworn drive offers only marginally higher performance than the FAT32 VFAT example above, but that’s not unexpected given the fresh state of that drive.

The drive is capable of peak speeds of about 180MB/s read and 22.5MB/s write with full performance at around 256KB accesses. As a result, it seems that the mkdosfs v4 results and the Windows 10 format to NTFS and exFAT are all fairly close to factory for performance, whereas the Windows 10 format to FAT32 reduces performance significantly and could reduce the lifetime of the drive through unaligned writes.


CrystalDiskMark shows the performance of the drive as FAT32 (formatted by Windows) and FAT32 VFAT (formatted by mkdosfs v4) on the right. It seems to suggest the sequential performance can reach up to 182.8MB/s read and 21.39MB/s write, but suffers heavily under small block accesses. It also seems to suffer under multiple queue access as well.


I tested the drive in FAT32 (formatted by Windows) and FAT32 VFAT (formatted by mkdosfs v4) using H2testW and it passed both tests. Read speeds did not change between either run, registering around 155MB/s (more than claimed on the package), although write speeds increased for the VFAT drive to 14.7MB/s.


I tested it with my normal daily-use phone, a Xiaomi Mi Max running ROM.

When plugged in, the drive is immediately detected and ready for use.

It turns up in the storage section under settings and can be accessed through the file manager. As it turns out, it works if formatted in FAT32 or exFAT, but does not work if formatted in NTFS.

In the case of NTFS format, the following prompt appears when plugged in.

I tested the speed using A1 SD Bench, although I don’t find the figures highly reliable.

FAT32 in VFAT format

exFAT format

I also tested it in a Windows 10-based Lenovo Miix2 tablet using ATTO, which worked just fine delivering USB 2.0 rates as expected. The drive was formatted in FAT32 VFAT format.


I was looking for an inexpensive 128Gb USB 3.0 drive and the Sandisk Dual Drive m3.0 was the answer. The drive offers very fast read performance especially over USB 3.0 for a drive of this price range, but the write speed was very average (up to 22.5MB/s peak, but around 15-18MB/s sustained) and of the sort you might expect from a low-end Class 10 SD card. Even a Samsung EVO+ microSD card would run rings around it in this regard. This is not unexpected considering that the drive is cheaper than most regular plain USB 3.0 drives.

The differentiator, however, is the presence of the USB microB port for USB OTG usage. This allows you to use the drive with supported devices to copy files to and from the device for backup or expanded storage. While I won’t recommend “hanging” the drive off the port for extended periods as an accidental knock might break something, it has come in handy for a quick file transfer (e.g. instead of uploading to the cloud or getting a USB cable to plug the phone into the computer directly). However, this all depends on your device and whether it supports OTG storage devices. Filesystem support also varies from device to device. You can test it out with any ordinary USB key and a USB OTG adapter.

As a result, it seems a decent buy even if you’re just interested in using it as a regular USB key instead of a dual drive. But whatever you do, avoid reformatting the drive if you wish to retain its full performance. If you must, it seems Windows 10’s default 1MB/2048 sector alignment works just fine for NTFS and exFAT, but not so well for FAT32. mkdosfs v4 in Linux seems to do a better job for FAT32.

Posted in Computing, Tablet, Telecommunications | Tagged , , , , , | 1 Comment

Event: CeBIT Australia 2018 Exhibition (15-17 May)

It’s become a ritual – every year, around this time, CeBIT comes to Sydney and I pay them a visit. It’s always a privilege to be able to walk around the expo to see who’s around, what’s new and catch-up with friends. As usual, I’m writing a post about it and the same disclaimer applies – namely, I’m only going to be mentioning the most interesting exhibitors in my opinion based upon my interests but that doesn’t necessarily constitute endorsement of their products.

Before the Show

The show was held at the ICC just like last year, but there was a difference already noted before the show. This year, they used a new system called CeBIT Match, which is an online web application accessed via an e-mail link which provided access to ticketing and information about the show. No need to install any one-time-use apps or print out paper tickets, which is great news.

The entry ticket itself was merely a low-density QR code containing the ID number encoded as text data. This makes it quite easy to scan, but I wonder whether this is in any way secure – those ID numbers could be guessed or follow particularly simple rules?

Another difference is in the entrance badges. Gone are the old 2D barcodes which required specialised scanning equipment and were difficult to work with and gone are the 1D barcodes that replaced them. This year was the year for RFID through the use of an NFC sticker.

Some people might have some privacy concerns, as NFC can be read at a distance with the right equipment, but surprisingly it seems that this implementation is not an issue. An analysis of the tag shows that it is a clone MiFare Classic 1K tag (i.e. not from Philips/NXP) but is basically unprogrammed. The only way they know who is being scanned is by use of the UID bytes (unique identification) which are programmed at manufacture and not otherwise changeable (unless you have a special backdoored chip). Since the UID bytes don’t correspond to the ID number, UIDs are not too easy to spoof and the tag isn’t programmed with your details, it’s quite safe. The contents of my chip with the two LSBs of the UID are shown at the end of the article.

As a result, you should probably keep your tag since basically you’ve got a free MiFare Classic 1K tag that you can reprogram for any use you desire – if I had known, I might have chosen to program an NDEF VCARD onto it to give away my details directly.

I visited the show on the first day of opening – Tuesday 15th May, which proved to be quite a nice and sunny day.

Getting the badge proved to be a relatively smooth process although there was a short wait. After passing a telepresence assistant robot, getting the badge scanned, I was in.

The Show

Right after entering the show, my own instinct is to start on the right side. But first, you should never pass up the chance to grab a printed show guide – it’s been very useful in the case of “I remember seeing a stand here, what was the company’s name again?” I’m glad they still print these paper guides.

The first thing that hits you is the massive NBN stand, where representatives from the NBN were showing off some of the technological possibilities that the NBN could empower. They also had a few tablets allowing you to check when the NBN will arrive at your place.

For one, the most eye-catching part of the display was the ute with a self-aiming satellite dish and Cisco VoIP phones inside. This is apparently an emergency voice communications unit, which makes sense, since the Skymuster satellites cover practically all of Australia, however the quality of service is probably not going to be great owing to the great round-trip delay to geostationary orbit plus possible access contention in the case of a real emergency.

In fact, to be honest, while NBN bills itself as “Australia’s broadband network”, its management and deployment has left a lot to be desired. With the change to a mixed technology mode, delays in HFC rollout and introduction of a new FTTdp tier, it’s become both somewhat confusing and unequal in what is being provided. I, like many others, would have preferred FTTH if done properly. Now with a mixture of technologies, emergencies could be an issue especially in FTTN and HFC deployments where the network itself requires power to operate, so even if you have a local UPS, this may not be enough. The loss of voice ports and the problems with voice-band modems over VoIP (e.g. SIP ATAs) also deserve mention, as many customers are being forced to find alternatives to maintain their desired services (e.g. fax machines, remote alarm monitoring/telemetry).

I did take some time to ask the representative about these issues, but it seems broadly that these issues are acknowledged but the risks must be handled by the customer in conjunction with their service provider – and what is offered will vary from service provider to service provider. Fair enough, although as they are basically switching off the copper, I think it would be nice if they did have a little more to say about the issue.

However, what was a surprise was the mention that it would be possible to pay a cost-differential to build out a given location for fibre (depending on your present access technology) or any other access technology. This is, however, again dependent on whether your service provider is bothered to handle such a case … which I suspect no “consumer-grade” plans will bother with.

But my biggest disappointment is just how long it’s taken. I’m still without NBN … until early-to-mid next year when HFC comes around. In the interim, I live off LTE which has both limited quota and speeds that go from 120kbit/s during congested periods up to 30Mbit/s at mid-night over USB 2.0 tethering (so it’s not a Wi-Fi restriction). If not, I would be suffering ~3Mbit/s ADSL2+ in this location.

Regardless of the issues, I’m still waiting for it to get here, no matter what the access technology is, because it will mean affordable high-speed access at long last.

Walking along, I passed PCCW Global which was showing off their Now TV platform which has been around in Hong Kong for a while. MAXO Telecommunications are back again this year, along with their competition MyNetFone. I didn’t see anyone that I knew from the past at the latter, which was a little disappointing, but at least the nice LED signboard and mini-golf are here again. Ninebot were also on show, with a number of their self-balancing scooters shown in a static display. Cool, but it would have been cooler to see them in action. It was also nice to see RFI back again this year, as they’ve been around CeBIT for almost every year I’ve been.

The next big attraction was Horizon Wireless, exclusive distributors for Tarana Wireless. They demonstrated a link over the 5Ghz unlicensed band using their own proprietary self-aligning radios, delivering around 400Mbit/s per link with interference mitigation and non-line-of-sight technology. These types of point-to-point links aim at displacing the need to install fibre for backhaul purposes. Apparently, there’s a lot of magic in the algorithms and the claims made on the site really are quite amazing – something in the order of 20dB noise cancellation and a 200Hz self-alignment refinement. I suppose that’s the magic of having a whopping 16 antennas at each end, but there’s bound to be a lot of DSP to make it happen in real-time. Apparently, they are trialling the new AbsoluteAir 3 which pushes the throughput even higher to 1.6Gbit/s peak. All very exciting and shows just what’s possible … although I suspect this may also interfere with regular wireless LAN users as many “proprietary” 5Ghz TDMA-style radios do, as they often will not listen before transmitting.

Like previous years, it was nice to see Icom around this year as well. It’s pretty tough being in the radio market, but it seems that Icom aren’t standing by idly. In the previous years, they demonstrated their Wi-Fi handheld transceivers, this year, they have their new IP501H LTE Transceiver. While the LTE and IDAS systems don’t talk to each other yet, the gateway is in the roadmap. This particular unit is pretty compact and robust feeling and it blurs the line between a mobile-phone and a handheld transceiver. It uses their own SIM card with their own plan, offering full duplex communications just like a phone with priority interruption ability. Voice encoding is G.726 at 32kbit/s (as with some SIP services) thus it seems to be a packet-data delivered voice.

Western Digital and Sandisk shared a stand this year. I don’t remember them being here in the past, so that’s new. At the stand, there were some G-technology external solutions and HGST product catalogues on show. Their stand was quite busy with people waiting to spin the roulette for a chance to win a prize or increase their entries into a draw to win something. Unfortunately, not being swayed by such gimmicks, I moved on …

It seems Sydney area universities are pretty well represented this time around. UTS had a stand with a lot of interesting things going on. As an engineer, I approve of technical displays. One was especially interesting, using sensors and computer vision to detect train carriage occupancy based on how many boarding/disembarking for dwell time optimization and to direct passengers to less crowded carriages. Good in theory, but could be quite difficult to reliably implement and could raise privacy concerns.

Western Sydney University also had a stand with a VR experience of the Penrith Observatory, the Rhythmotron amongst other things.

UNSW haven’t really changed anything much, bringing out the Sunswift solar car once again, although in much more eye-catching colours this year.

Speaking of cars, there was a demonstration from EasyMile of a driverless transportation vehicle which reminds me of some of the people-mover pods that are used to transport people between certain airport terminals. It’s pretty cool to see it in action, even if it is a little slow (maybe for safety) but I didn’t have a chance to ride in it though.

With such an eye-catching display, there’s no way I could have missed Vodafone this year. It’s been challenging times for Vodafone over the past few years with service quality issues, but through a vast network remediation plan and now, with a lot of introductory offers, they’re trying their best to keep themselves going. Their display focused a lot on different IoT network possibilities, demonstrating various products which even included a home automation and security gateway and tiny tracker products which use a solder-able SIM. In fact, it seems that their IoT SIMs and platform offer extra features including global roaming abilities which is pretty cool. I was (unexpectedly) interrupted by a former schoolmate which I hadn’t seen in over a decade who works for Vodafone and was in charge of the stand. It was great to catch up.

Aside from them, Norton had a large showfloor as well, with spinning holographic displays which seems to catch a lot of attention. However, I have to say that the Norton name has lost a lot of shine in recent years … so I just walked on by. I also got to see some of the new Panasonic Toughbook laptops and tablets – good to see they’re still around, with the latest units running Windows 10 as well.

An unexpected exhibitor was Fing, the guys behind the best network scanning app around. If you’re the kind of person that does “tech support”, this is an app you literally must have. I’ve been using the app myself for over three years, and it’s free. I’ve watched it grow from a very capable ARP scanner to gain a lot of features such as NBNS/Bonjour name resolution, service scanning, traceroute/ping, wake-on-LAN, etc. It’s the one-stop-shop to find out whether a given device is on the network and what address it has – especially useful for networks where you don’t have access to the DHCP lease list or the IPs are statically configured (as some of my networks are).

They were on-side promoting their Fingbox, which is a hardware device attached to your network which can be used as a cloud-connected node to scan your network remotely, use as an IDS, bandwidth monitor/Wi-Fi speed tester, block devices from the network (through ARP spoofing and continuous gratuitous broadcasting) amongst other features. Definitely an interesting device … and maybe one I can look into in more detail in the future.


Just like previous years, there was a lot of emphasis on software platforms and services, which weren’t really my main interest. Despite this, there was still a good number of exhibitors that I expected to see and did visit, so it was a good show. It was a chance to have a chat and get up to date with what’s new – thanks everyone for taking the time to talk. The new CeBIT registration process was quite smooth and the use of NFC tags was nice to see.

Hope to see you all again next year! You can sign up for next year already!

Continue reading

Posted in Event, Telecommunications | Tagged , , , , , , | Leave a comment

Project: Generic Electronic 57 LED Hourglass Kit (TJ-56-205)

I couldn’t help myself, so once I finish one kit, it’s often straight onto another. This kit was sold as an electronic LED hourglass and was a little more expensive at AU$3.07 including postage. Compared to local suppliers, this is a bargain – one major supplier in Australia wants AU$8.95 for a two transistor flasher kit, so I thought I’d build one of these instead and try it out.

The Kit

Another clear plastic zip-lock bag. Maybe I should call these “fun bags” from now on … or maybe not. *looks awkwardly around in case anyone heard me say that*

Inside, there is a double-sided quarter page leaflet in Chinese with (presumably) instructions, PCB layout, bill of materials (BoM) and schematic. Already, we can see that the circuit uses a microcontroller, an obscure STC 15W201S, with the ability to be reprogrammed through a serial ISP header. This unfortunately means that the kit is probably not as educational as something with more traditional chips – the source code for the current program isn’t provided either. The circuit runs LEDs using a complimentary visual multiplexing matrix drive, so while not as complex as Charlieplexing, does at least allow for twice as many LEDs to be controlled as compared to regular visual multiplexing and makes debugging somewhat easier. Curiously, the BoM lists a capacitor that has no position and omits a USB connector.

The supplied components are, as listed in the BoM, so I suppose that’s what they intended to supply. It is a good idea, wherever you use microcontrollers, to have a bypassing capacitor installed near the chip itself but the leaflet makes no mention of a provision for it. The inclusion of an IC socket is a good feature, although I’d have to criticize them for including such a rusted, crusty push-button switch. There is also a quality control slip as well.

The PCB is a double-sided fibreglass type PCB with green soldermask on both sides and white silkscreen on the top. The pads are tin plated-through-hole, which makes this a pretty decent quality board for a kit. The board has the identifier TJ-56-205 on it, which suggests it is related to the heart flasher kit built in the last post.

Rather sadly, in the packing process, it seems that the board did suffer some abrasion but was not functionally damaged.

Construction and Testing

This kit is one that seems to require a substantial amount of work to complete –

57 LEDs x 2 connections
1 IC x 16 connections
1 switch x 5 connections
1 power jack x 3 connections
1 ISP header x 4 connections
1 switch x 4 connections
TOTAL - 146 joints

Construction was, however, very straightforward because the orientation of LEDs was consistent, the board was well labelled and the pads (while small) were tinned. The only thing it took was time – about 55 minutes for me to complete including distractions and photographing.

In the end, three spare LEDs were included, which is a nice touch. However, the negatives were the rusty push-button switch which was found to be somewhat intermittent in functioning, the lack of USB connector (for a more popular way to power the circuit), the lack of a capacitor position on the board, the “blind” use of a rather obscure microcontroller to perform all the functions and the almost unnecessarily thin traces (which might make damage more likely).

The unit, when operating, runs like this, displaying a rather long animation that simulates grains of sand falling from the upper to the lower portion. This “abruptly” resets to a full top-half once the bottom-half is full. The push button allows you to toggle between three speeds.It’s not particularly inspiring, but the LEDs are quite bright and the visual effect is somewhat eye-catching. It’s a shame that the microcontroller is a little obscure, otherwise the LEDs could be used to display something else entirely (or maybe a more flexible hourglass that actually functions as a timer).


While this kit was slightly more expensive than some others, it’s still very cheap by local standards. The cost is understandable, especially when you consider how many high-intensity LEDs were included and the quality of the double-sided fibreglass PCB. Construction was very straightforward, with only minor negatives in the fact that everything is hidden inside an obscure microcontroller with no source code, that the traces seem unnecessarily thin on the board, that a USB connector for easier powering was not included, that the push-button switch included was rusty and intermittent and that no position is provided on the board for the capacitor which I would deem good practice.

That being said, the tin plated board was a joy to solder to and the consistent LED direction made things a lot quicker than it would have otherwise been. The effect is also quite bright and visually distinctive. Having three spare LEDs and an IC socket makes this kit quite recommendable for someone who wants a decent amount of through-hole soldering practice, with a good chance of success.

Posted in Electronics | Tagged , , , , , | Leave a comment