Now that most of the hard (review) work is done, I can do what I’ve wanted to do for a while which is to take a peek under the covers (perhaps venturing into full teardown) to see what makes the instruments “tick”. How did they manage to get the performance that they did into such a box and at such a price? Perhaps we can learn something by opening the instruments up.
Warning: Despite the lack of warning labels and service guides with instructions for disassembly for replacement of the internal CR2032 coin-cell battery for the RTC being available, I would strongly advise that readers do not disassemble their instruments unless absolutely necessary. Doing so may cause damage (especially if you tear a flexible-flat cable) and affect the instrument’s accuracy specifications. This was something I was willing to risk to get a look inside my instruments, so you (hopefully) won’t have to!
EDU34450A Digital Multimeter
My first “victim” will be the EDU34450A DMM. Removing the back plastic cover from these oscilloscope-like units only takes the removal of four Torx screws.
This reveals the internal metal chassis which serves as a shield cover.
The hardware power switch is seen mounted to the side with an opening in the cover for the wires to go through.
The shield is not entirely closed with some cut-outs to allow the handle to fit. To remove the cover requires removing numerous Torx screws along the perimeter, then the rear cover can be lifted away, taking care not to pull on any wires.
The need for a rear voltage selector becomes apparent, because the power supply to the DMM is a transformer-based supply consisting of many primary windings, a screened secondary, a centre-tapped secondary and another separate secondary winding. The screening likely allows for the “digital” noise to be attenuated so it doesn’t affect the power that runs the meter front-end. A suppression toroid can be seen for the meter front-end power cables.
The DMM is broken up into two boards – the left is the digital “brains” that seem to be in charge of the communication, LCD and front panel. It drives the DMM front-end board on the right (that takes the majority of the width) via a ribbon cable. This board has a number of components visible with the more sensitive components located behind a cover, likely to reduce noise and improve temperature stability.
The DMM board has the input sockets soldered directly to the board, so every insertion or removal of the connections puts a tiny bit of stress on the board. This is probably a design decision that is not ideal, but reduces cost. The series fusing for the current measurement is connected via spade connectors. Large wire-wound resistors appear to be used as current shunts, with Chinese HongFa relays switching signals around (the latter, possibly a cost-motivated choice).
In-spite of this, the input terminals still seem to be well protected with gas discharge tubes and MOVs to suppress transient voltages.
The DMM front-end board was extracted from the chassis and an interdigitated finger pattern can be seen in the corner of the PCB. I wonder if this is a functional element (e.g. a capacitor) or whether it is used for testing the characteristics of the PCB.
Regardless, the other side of the protective can is seen with the single screw that “clamps” both halves together. Removing this screw exposes the important precision parts that ensure the meter’s accuracy.
On this side, it seems the major component is a Keysight branded chip – perhaps their own ASIC, marked 1NJ5-0001. Even around this chip, a number of guard-like rings can be seen across some important inputs to the chip.
Unfortunately, the manufacturing quality does seem to leave something to be desired, as flux (or flux cleaner) residue can be seen on the board and seems to have partly discoloured the copper. Perhaps this is another sacrifice in the name of price.
The other side reveals that one of the guard rings goes to a pair of glass diodes (perhaps special ones). Nevertheless, there are a few resistor “ladders” visible, made of discrete SMD resistors, possibly to handle high voltages. Above that is RN401, a Caddock 1776-C48 cost-reduced decade resistor network which serves as a reference. Also within the confines is an EEPROM and STM32F401 which seems to be driving the show via an isolated interface to the ribbon cable.
Further disassembly required removing screws to allow the front plastic fascia to be separated from the metal frame.
The front panel keypad seems to also be operated by an STM32. This board has two areas with patterned stripes – perhaps this is intended for use with grounding foam for shielding, but doesn’t seem to be used.
The front keypad is made of the standard rubber membrane that is similar to that inside TV remote controls, with rear carbon pads touching carbonised traces on the back of the PCB to form buttons.
The other key element you will find on the front metal chassis is the LCD screen. Visible around the screen is a thin bead of sealant.
This is used to hold a fairly substantial piece of cover-glass over the LCD to improve its strength and robustness. This way, a stray cable connector whipping about wouldn’t scratch or crack the screen so easily.
Unfortunately, the LCD itself is foam-taped into the frame assembly, so we don’t get to find out who makes the display itself. I suppose the whole module is designed to be replaceable, but whether it is worth the technician’s time is another question given the amount of disassembly required to reach it.
Finally, we reach the back of the “brains” PCB – this one also exhibits some flux cleaning residue which is not ideal. I suppose this counts as a full teardown of the DMM!
EDU33212A Arbitrary Waveform Generator
Having fully torn down the DMM, I suppose there is no need to go as far for the other “oscilloscope-like” instruments, assuming they share similarities internally.
The AWG already shows its difference up-front with a HengLiXin HD6020B12M fan facing outwards.
Inside, the architecture is noticeably different as this unit uses a switching power supply behind the metal shield to power the unit. This is the reason there is no voltage selector on the rear. Interestingly, the power cables from the power supply to the board and out to the fan are of the same type of connector which may make them vulnerable to being transposed during assembly, so take a note of that! But that does suggest to me that perhaps this is an “added feature” to regulate the fan speed based on temperature (at a guess).
The main board for this unit spans across the whole device. Another STM32 can be seen running the show, with some RAM and an Ethernet PHY/MAC. Under the heatsink must be a hefty chip, which drives the ADC. To the left-left is power-related circuitry, underneath is the components that “massage” the signal to the right levels. Near the bottom centre appears to be the clock reference circuitry. Of interest is the use of Nexem (formerly NEC-Tokin) Japan-made relays – these are quality signal relays which I would have expected to see in the DMM instead of those HongFa units.
Now, can you guess what’s underneath the heatsink? Drumroll please …
… for the Intel (formerly Altera) Cyclone 10! No, I wouldn’t have guessed this either! There’s RAM on both sides of the chip, flash to the left and an Analog Devices TxDAC similar to that in the Digilent Analog Discovery Pro 3450 (that I previously reviewed) on the right. The PCB has a lot of complicated trace “squiggling” to match trace lengths (presumably) that it’s sort of artistic as it is!
What’s under the power supply shield? Glad you asked. It’s an FJ-SW916-A7 which is covered in copious amounts of aluminium electrolytic capacitors and white silicone goop to keep everything in place.
Removing this for a closer look from the underside, it seems it is covered with a clear conformal coating, except for the screw holes with pads. Even the “spark gap” areas near the optoisolators (there seem to be three) have been covered, but given the width of the gap, perhaps they were never meant to function as spark gaps anyway.
The construction quality seems very slap-dash on this board. As it is a simple board, the diodes seem to be all over the place – mismatched footprints, packages at a skew … this is not an example of high-quality construction.
Examining the power supply from the top shows that there is a T3.15A fuse on the board itself. Aside from this, there seems to be some ChengX capacitors on the primary side – these are not a particularly reputable brand, generally speaking.
Other than this, the board appears to be littered with many OKCAP capacitors. I can only imagine that an engineer somewhere asked whether the capacitors were OK … jokes aside, this is not a highly reputable brand either. Whether this proves to be an Achilles Heel would be difficult to know, but given that Keysight has been building instruments for a while, I would suggest that these may be selected based on their reliability experiences.
EDUX1052G Digital Storage Oscilloscope
Now, it’s time for the DSO to have its covers removed.
In contrast to the AWG, the DSO has the fan blowing inward, but is otherwise superficially the same.
Once the cover was lifted, the leads were found to be so short that I had to disconnect them to continue. As a result, we are now staring at the main-board which spans the whole width again. The two channels can be seen on the left with their shielding cans – I didn’t dare touch this in case it affected the calibration. Above that is copious power supply circuitry. In the middle, towards the lower part of the board, is likely to be responsible for the wavegen functionality including more Nexem relays.
The elephant in the room? The Keysight BLT (sandwich) which seems to be a system module that sits on a mezzanine! This module has a hefty ST SPEAr600 dual-core communications processor, a Xilinx Spartan 3 FPGA flanked by Winbond Flash memory chips. Two heatsinks seem to be concealing their own ASIC and perhaps a higher-performance FPGA but I can’t be sure as I wasn’t willing to try breaking the thermal adhesive.
The module can be separated with care, with the via patterns and board-to-board connectors showing just how many pins are involved. Impressive. There also seems to be some memory on the back in U701 – perhaps this is relevant to the configuration of the device?
Having removed this mezzanine board, the mainboard itself looks pretty good. The only thing I noticed was X3, the RTC crystal, being slight askew, but otherwise the manufacturing quality seems fine. All boards so far seem to use the same CR2032 cell from Chao Chuang – perhaps this is their preferred supplier.
In case readers are interested, here are some more close-ups of the main board.
The rear shell seems to house a familiar power supply and another HengLiXin HD6020B12M fan, same as in the AWG. Perhaps this particular fan supplier is one approved by Keysight.
To be sure, I removed the cage around the power supply and found a perfectly identical FJ-SW916-A7 power supply. Therefore, it can be concluded that the AWG and DSO actually share the same power supply unit and fan (also the direction is reversed).
EDU36311A Triple-Output Power Supply
So far so good, but now it’s time for the hefty beast of a power supply. This required removing the four feet underneath, three screws on the rear lip, two screws near the front plastic feet and two screws securing the carry handle. Only then can the gunmetal grey shell be slid off the chassis of the unit.
The unit is heavy and chunky, so I decided I won’t tear this one apart too far because it doesn’t look particularly friendly to disassemble … and I only just figured out what a decent unit it is!
Securing the toroidal transformer sideways is a bit of an interesting design choice. I’ve never seen it before, but it looks quite neat!
The underside is mainly a steel plate with a few screw-heads and cable ties poking through.
Removing the cross-brace reveals the inside of the unit more clearly. There is a rear board responsible for bringing the USB and Ethernet connections to the back, a front board which has all the smarts, a keypad board and LCD in the front section, and a main board underneath. The heatsink board above appears to be a cleverly designed spacer to ensure the heatsinks remain well spaced even during vibration in transit. From this cursory glance, it seems that this design is entirely linear! Look at those hefty capacitors!
The main toroidal transformer comes from Eaglerise Electric & Electronic and is a complicated transformer. Marked as model HP248-01, it has a 248VA rating - for a 90W PSU it seems a bit of an overkill but should serve to help it run cooler and have better on-load regulation. The primary side has a screen connection, while there are seven secondaries! In a word, wow!
The font main-board is version 1.81 with separate areas for each channel. The sections are well marked by silkscreen and the various capacitors are all glued to prevent vibration-induced damage. Each channel seems to be independently controlled by a separate STM32 with an isolated data bus.
The whole show is then orchestrated by yet another STM32 – it seems this is a favourite for Keysight. Notice the positioning of the CR2032 clock battery, however – this location is just about the worst that can be chosen for serviceability.
Each channel has a reverse-parallel diode to protect against reverse polarity and a capacitor. All electrolytic capacitors in this unit seem to come from JiangHai which is not a common brand either. I suppose this is how the unit can be built down to meet a price point.
Not being able to get closer without potentially breaking something, I settled for this shot of the channel output heatsink which seems to show a bridge rectifier, dual-diode package, thermistor and transistor/MOSFET of some sort.
I decided to grab a closer look at the rear PCB as it seemed to be a bit wider than absolutely necessary. It seems there were positions for further components – perhaps this was meant to provide a USB-host port on the rear rather than the front?
The voltage selector board can perhaps get a little love as well. I like the fact that the wires are neatly crimped into ferrules and then soldered down. Construction quality of this unit appears rather good.
Finally, the rear fan is a HengLiXin HD1225B12H rated at 6W. No wonder it’s a bit noisy – this is a pretty powerful fan, but I suppose it’d need to be if this is a fully linear power supply!
Conclusion
Taking a peek under the covers was quite interesting to see. The DMM appears to have a transformer-based power input with screening. There is a separate PCB for the brains based around an STM32 and the DMM front-end, as well as another for the front-panel keypad. The main banana jack connectors are directly connected to the front-end PCB which is perhaps a cost-saving design, along with the use of Chinese HongFa relays. The inputs appear to be protected by MOVs and gas discharge tubes, so safety seems well taken care of. The current shunts appear to be wire-wound resistors. Looking behind the covers, there is a Keysight branded 1NJ5-0001 chip and a Caddock 1776-C48 cost-reduced decade resistor network reference that I could identify, driven by another STM32. The front-panel keypad uses membrane switches with carbon-based tracks, while the LCD shows a moderately thick glass panel glued to the front for robustness, glued into a carrier so the original manufacturer could not be identified. On the whole, it seems sensible, however, the construction quality did leave something to be desired with some flux/washing residue and discolouration as a result.
The AWG follows a similar case design but instead uses a switching power supply (FJ-SW916-A7) which is shared with the DSO. This supply has conformal coating on the rear, covering some (probably non-functional) spark gaps and mediocre construction quality with diodes mounted askew. The supply also uses a mixture of ChengX and OKCAP capacitors which are not highly-reputable brands, which is perhaps the cost of being price-conscious. However, the design of the mainboard itself shows quite a bit of care, using Nexem (formerly NEC-Tokin) signal relays from Japan and being based around an Intel (formerly Altera) Cyclone 10 and an Analog Devices TxDAC. The fan used comes from HengLiXin, while the CR2032 coin cell for the real-time clock comes from Chao Chuang, both seemingly preferred suppliers.
The DSO also shows relatively good construction quality (with the RTC crystal being slightly askew) and uses a mezzanine board named “BLT”. The board houses a hefty ST SPEAr600 dual-core communications processor, a Xilinx Spartan-3 FPGA and two hefty chips under heatsinks, presumably their own ASIC and maybe a hefty FPGA. Each channel’s front-end is encased behind a screening can – I didn’t dare remove it in case it affected the calibration. While it uses the same power supply and fan, the fan direction is reversed compared to the AWG.
Finally, the PSU is a heavy and chunky beast, containing a sophisticated over-sized toroidal transformer from Eaglerise Electric & Electronic with seven secondaries and a screened primary. The power supply seems to be a linear design, with each output controlled by its own STM32 and with another STM32 driving the whole show. Capacitors in this device all appear to be from JiangHai, which is yet another less-reputable brand, which appears to be a compromise for price. Similarly, the 12cm rear fan comes from HengLiXin.
In the end, no instruments were harmed in this peek under the covers – all instruments pass their self-tests post reassembly and all screws were accounted for in the process. There seem to be compromises made in order to reach the low price, namely the use of many more China-sourced components in the design. Given sufficient quality control, this in-itself would not be a problem, and I feel inclined to trust Keysight given the fact that there seems to be preferred suppliers for parts which remain consistent across these products, so there seems to be some care to make deliberate selections of these components. However, the construction quality of some parts seemed to show slightly less attention to detail than I would have expected, but this shouldn’t affect the operation of the instrument at the end of the day.
---
This post is part of the Keysight Smart Bench Essentials RoadTest Review.
Direct links to detailed blogs:
- Keysight SBE In-Depth – Ch1: The Need for Smarter Benches?
- Keysight SBE In-Depth – Ch2: Unboxing^4 & Design Features
- Keysight SBE In-Depth – Ch3: Initial Setup & Documentation
- Keysight SBE In-Depth – Ch4: On-the-Bench User Experience
- Keysight SBE In-Depth – Ch5: Connected to the LAN
- Keysight SBE In-Depth – Ch6: PathWave BenchVue Oscilloscope, Power Supply, Digital Multimeter & Function Generator
- Keysight SBE In-Depth – Ch7: Keysight BenchVue Test Flow Automation
- Keysight SBE In-Depth – Ch8: Instrument Performance Tests
- Keysight SBE In-Depth – Ch9: Peeking Under the Covers
