IntroductionAround Christmas my GAS started to act up again and I bought another oscilloscope. Last thing I need is another oscilloscope but this one is special: A Yokogawa DL708E. This is an 8 channel oscilloscope. Or better put: it is an oscilloscope mainframe with room for 8 acquisition modules. It's bigger brother the DL716E can even hold 16 modules. There are several different acquisition modules like standard, isolated, digital, high accuracy, strain guage and thermocouple. The modules typically have a resolution of 12 bit or better. These oscilloscopes are geared towards (long) recording purposes. One of the more unique features it has is a timed trigger which allows to take a snapshot of a signal at regular intervals and save the data.
I found my one on Ebay and lowballing the seller worked. But it isn't without problems. The picture from Ebay shows there is at least a problem with the screen:
Also the shipping costs where quite high. It wasn't clear from the pictures on the listing but it came in a big & heavy duty flight case:
The listing said it included 8 modules but in fact only one was included. It has some nice options though like 16Mpts memory (with 2 channels enabled; less with more channels) and an internal hard drive. Fortunately I managed to find some more modules on Ebay and now I have 4 isolated and 3 normal modules. Don't expect high samplerates or extreme bandwidths though. The maximum sample speed is 10Ms/s. The module's bandwidth maxes out at a few MHz. There where also some additional items like front cover, accessory pouch, BNC cables (too thick for my taste; good for the giveaway bin), some thermocouples, a piezo load cell (probably for vibration measurement) and last but not least a big flight case (Explorer model 5822). Maybe I can sell the flight case.
Take it apart!An initial power on (using a variac to make 110V AC) showed the screen had issues indeed. From earlier experiences with TFT screens my diagnosis is that a couple of banks of the column drivers are broken. So let's take it apart! Fortunately I have found a service manual with a parts list and exploded view. Nevertheless it wasn't easy to take it apart; the hidden screws under the glued-in rubber feet wheren't helping!
Front panel assembly (without screen):
A look inside the chassis at the side where the PSU and the main board live:
The mainboard:
Note that the mainboard uses standard PC style memory (72 pin SIMM modules) for the acquisition memory. Yokogawa lists several part numbers for different memory sizes. The modules fitted are 32Mbyte each. No wonder since the aqcuisition modules produce up to 16 bits (2 bytes) for each sample that needs to be stored. Furthermore it has several ASICs to deal with the acquisition modules.
Power supply:
This seems to be a module that is bought from a firm that specialises in power supplies. More about this later!
The hard drive:
A good old Quantum fireball but unfortunately dead as a Dodo. Doesn't even spin up. I tried to use several different hard drives but none is recognised by the oscilloscope. Which leads me to the next item:
Hard drive interface:
I never thought I'd lay my eyes on something like this but it is a custom made SCSI to IDE interface. Appearantly the people at Yokogawa got the message that SCSI wasn't the best way forward to put a hard drive in their system. However my assumption is that this one is broken for some reason. I wish they had put an IDE interface on the mainboard though. I'm willing to bet that would have still worked. It would have been nice to put an SSD into this oscilloscope for storing data but I can certainly live without that.
Module bay:
Not much to see here...
Isolated module:
This module is specified for measurements with a common voltage of 48V from ground. The construction is quite straightforward. Isolated DC-DC converter, some optocouplers for digital control and a 'linear' optocoupler.
Standard module:
The PCB layout looks pretty similar to that of the isolated module.
Fixing the screenAfter taking the LCD panel apart (this is really easy to do!) it became clear that it likely has some water damage:
Unfortunately this has corroded some vias on the column driver board. This means that one of the column drivers isn't getting power. Fortunately it was easy to put a wire in place:
After the fix there is still a vertical stripe left so not 100% fixed but good enough for now:
I have looked at buying a 'new' Sharp LQ10D367 screen. These are available but I don't like the price so I'll leave it for now. Maybe a screen with a better price pops up at some point.
Buttons...It turned out one of the buttons along the screen wasn't working so that also needed replacement. I had some buttons leftover from my DL1740 restoration project so that was a quick fix:
Not the nicest job but it will do.
Power supplyDid I mention I bought this from US? I guess not but since they have 110V in the US my DL708E comes with a 110V PSU. I kind of betted that converting it to 230V would be very easy to do.
This gave me high hopes:
The board supports both 100V and 200V. But I got a nasty surprise. From the filters the mains goes straight into a rectifier which feeds 2 200V capacitors on parallel. No way to connect it in half bridge and no other signs that a conversion to 220V is easy. I guess the conclusion is that the 200V version has a completely different bill of materials. That sucks big time! Besides needing a transformer in one form or another, I'm going to plug it into a 220V outlet sooner or later. So what to do... using it with the variac all the time? Buy a stepdown transformer and bolt it onto the scope? Or even more idiotic: make a new power supply? After a good night's sleep making a new power supply didn't seem that idiotic. I recalled that nowadays you can buy small AC/DC modules which I have already used for a project. At first glance the AC/DC modules look very affordable so let's see where this idea goes.
As a first step I measured all the voltages and line trigger signal (!). Some of the voltages are marked as unregulated and I wanted to know how unregulated they are. 14V turned out to be 15.5V and -7V more like -8.5V. Further investigation showed that the -7V is feeding 2 7905 (-5V) regulators. 5Vac is the line trigger signal which isn't AC but a square-ish wave obtained through an optocoupler. So onto Farnell to select a bunch of AC/DC conversion modules that would fit the bill without breaking the bank. Soon I ended up with a shopping list for all the parts including connectors, resistors, filter components, etc.
After carefully measuring the board sizes I created a PCB outline that should fit including the mounting holes. The end result is this:
I kept the mains input the same as much as possible but I did replace the right-angle Molex connector with a straight one due to keep the space behind the ventilator open. I used the same connector on my replacement board and made a new cable including a ferrite core. I can put the original PSU back if I want.
After receiving the board, I mounted the connector and aluminium frame from the original PSU to see if it fits... and it does! Perfect!
To limit the inrush current and improve filtering (according to an application note for the Multicomp / Mornsun) modules there is a 10 Ohm 4W resistor in series with each module. This is a wirewound resistor with a high surge capability.
A class Y capacitor connects the mains ground to the 0V of the power supplies. An common mode filter, MOV and class X2 capacitors keep the nasty stuff in or out (depending on how you look at it). Each supply voltage is filtered using an inductor and extra electrolytic. Except for the unregulated -7V. For that supply I used two 1.5 Ohm resistors for two reasons: 1) two lower current 9V modules where cheaper than a single 9V module so I needed a way to share the current. 2) I wanted to dissipate some power in the PSU so the 7905 regulators don't get cooked.
Some probably already noted a red bodge wire. Before inserting the new PSU, I checked all the supply voltages and it turned out I swapped +12V and -12V so that needed to be fixed.
An optocoupler circuit with a capacitive dropper (again class X2 capacitor) creates the line trigger signal. The optocoupler I used has a anti-parallel LED input so I had to add an extra diode to make it work on one half of the mains cycle only.
If you wonder why there is so much space behind the fan: my original plan was to put a silent Noctua fan in there but those aren't available in a 15mm thickness (only 25mm) so I needed more room to fit a fan. Since there is only room for a 15mm fan in the air exhaust I had to move the fan inwards and thus I created the cutout in the board. However, after fitting the Noctua fan the scopes started beeping and throwing error messages:
Appearantly it not just checks if there is a tacho signal but also the frequency. -Sigh-
But at least I can mount the fan further away from the air outlet and reduce the noise a little bit:
Another picture of the new PSU tucked nicely into the DL708E:
The mounting pillar lines up with the hole in the cover so all the original mounting points for the PSU are used.
BTW: the foreground shows the extra PSU to power the hard drive.
Testing & remote accessOne of the things I like about a DSO is to make screenshots. Initially I had the idea of making a board which takes the signals from the display and can store an image on an SD card (through a CPLD + memory to grab the screen and a microcontroller to convert the data to PNG and write it onto an SD card). However, the new PSU was a big enough project in itself and I didn't want to spend even more money.
Using the disk drive has proven to be flaky in the past. The USB disk drive I bought a long time ago doesn't work very well. So I turned to a device I acquired earlier: A Prologix GPIB ethernet adapter. Another interesting option for pulling images from a GPIB device is the Kiss-488 printer to webserver emulator. But since I already have the Prologix I decided to use that.
I had already created (can't call copy & pasting code from internet writing code) a Python script that pulls an image from my Tektronix frequency counter. I changed that script to work with DL708E:
#! /usr/bin/python3
#DL708E image grabber
# Make sure the Prologix is set to auto mode using ++auto 1 command through telnet to port 1234
from PIL import Image, ImageFile
import datetime
#Bring in the VISA library
import visa
#Create a resource manager
resources = visa.ResourceManager('@py')
lib = resources.visalib
#Open the FCA3100
fca3100 = resources.open_resource('TCPIP::192.168.2.20::1234::SOCKET', read_termination='\n')
fca3100.timeout = 3000
bufferSize = 2000000
fileSaveLocation = r'/tmp/dl708e.bmp'
#Return the scope's string to tell us it's there
print(fca3100.query('*IDN?'))
fca3100.write('IMAGE:FORMAT BMP')
#fca3100.write('IMAGE:TONE REV') #set reverse colors to get white background
fca3100.write('IMAGE:TONE COLOR')
fca3100.write("IMAGE:SEND?")
header = lib.read(fca3100.session, 8)[0]
lengthBytes = header[2:8]
fileSize = int(lengthBytes)
bytesLeft = fileSize
print('Length: ', fileSize, ' ', bufferSize)
imgBytes = bytearray()
# Read the rest of the image
while len(imgBytes) < fileSize:
imgBytes = imgBytes + lib.read(fca3100.session, bufferSize)[0]
print('rx length: ', len(imgBytes))
# Save the bytes to a file
imgFile = open(fileSaveLocation, "wb")
imgFile.write(imgBytes[0:fileSize])
imgFile.close()
ts= str(datetime.datetime.now())
img = Image.open(fileSaveLocation)
#img = img.convert(mode = 'RGB', palette=0, colors=256)
#This resize is ugly.
#img.resize((int(img.size[0] * 1.6), int(img.size[1] * 1.6))).save('dl708e_' + ts[0:19] + '.png', 'png')
img.save('dl708e_' + ts[0:19] + '.png', 'png')
#back to local operation
#T.B.D. ...
#Close the connection
fca3100.close()
No, I didn't bother to change fca3100 to dl708... The script uses Pyvisa which is a standalone Visa implementation so the horrible NI Visa package isn't needed. The DL708E allows to grab images in greyscale, normal colors or reverse colors. The script converts the BMP image to PNG for better compatibility with modern software.
Color grading test image:
And on to some more testing to see if all the modules work:
Crap... trouble in paradise. That looks like the ADC for channel 1 has issues. After taking the module apart, reseating and cleaning connectors it dawned to me that pushing the BNC connector on the module in place instead of the module itself resolved the problem. It turned out to be a flaky BNC T-piece (which is now being recycled). Likely some noise / current coming from the isolated DC-DC converter in the module get injected into the input by the way I setup the test.
Time for some testing...After some pondering on what would be something simple to test I came up with measuring the voltage across a DC load and the current through it using a current shunt resistor. The power source is a PSU set to 2V with a current limit of 10A. I used the floating channels on the DL708E for the measurement so I can put the current shunt in the positibe line coming from the PSU (high side measurement). The voltage is measured using a banana to BNC converter which goes straight into the DL708E.
The current shunt resistor is connected to channel 1 and has a sensitivity of 10mV/A. So I set the linear channel scaling to 100x (1e2) and the unit to A (the unit is a text; if I want to have the trace show 'MHz' as a unit I just enter it as a text).
As a test I want to keep it simple: what happens when I press the 'short' button on the DC load? Well... this:
Let's zoom in a little:
Or with just the zoomed area:
The DL708E allows to have 2 zoom windows defined besides the (regular) main window. It is possible to show various mixes of main and zoom windows to get an overview of a signal at different locations and zoom levels.
Ofcourse vertical zoom is also possible. Let's take a closer look at the wiggle when the DC load is about to kick in:
The 12 bit ADC resolution really shines here giving 16 times more resolution compared to an 8 bit DSO. Note how the vertical zoom is set to 12.5x !
I see this oscilloscope becoming handy for floating measurements. A while ago I had to do some testing/debugging on a power-over-ethernet supply which has two seperate grounds at the primary side.
Well, so far for my Christmas holiday project!
Home
Help
Search
Login
Register
