RoadTest: Tektronix AFG31052 Signal ARB/Function Generator
Author: three-phase
Creation date:
Evaluation Type: Test Equipment
Did you receive all parts the manufacturer stated would be included in the package?: True
What other parts do you consider comparable to this product?: The AFG41052 was directly compared to the Rigol DG1022 and the Uni-T UT962E for basic performance parameters. A secondary comparison was made to the built in waveform generator of the MDO34 for generating partial discharge patterns.
What were the biggest problems encountered?: I could not get the built in PWM function to simulate an insulation tester output as planned. I am pretty sure though that it is something I am doing and not an issue with the AFG31052. The InstaView technology does not seem to perform consistently for me. I found the control to be split across the keypad and touchscreen, requiring both to achieve some tasks.
Detailed Review:
Introduction
This is my RoadTest of the AFG31052 Arbitrary Waveform Generator from Tektronix. I primarily applied to cary out the RoadTest based upon my experiences with my RoadTest on the MDO34 and using the built in waveform generator to simulate modified partial discharge pulses for verifying the operation of test apparatus.
Due to the limitations I found with the MDO34 for simulating partial discharge pulses, I identified the AFG3000 series as a potential alternative that would be more successful for carrying out the tests.
To run alongside these tests, I planned to compare the AFG31052 to four other arbitrary waveform generators that I have access to. This centred around carrying out the verification tests from the service manual for the AFG31052, but I also included tests from other manufacturers that were not covered in the manual.
The final element to the testing is to using the double pulse functionality of the AF31052 to measure the performance of various MOSFETs found within the insulation testers I have reviewed in other blogs on this forum.
Arrival
The AFG31052 arrived in a Tektronix branded cardboard box with foam inserts for mechanical protection, accessories are included within a separate cardboard box sitting within the foam inserts.
Accessories wise, the unit arrives with two 50 Ohm BNC to BNC leads, USB lead, mains lead, safety leaflet and calibration certificate.
The AFG31052 is a sizeable unit, situated next to the MDO34, it almost takes up the complete desk. This is something that I found quite noticeable with this type of higher specification apparatus from Tektronix. I tend to find that I get quite a congested desktop when using the AFG31052 alongside the MDO34 and does not leave room for much else. Other manufacturer’s apparatus is more easily stackable to reduce desk footprint. The MDO34 can be stacked on the AFG31052, but I wasn’t brave enough to leave it there for too long. Neither instrument can be stacked with any of the other test apparatus I have in my possession.
This isn’t a major problem, but I feel is worth considering for those that have minimal work space.
Build Quality
Having had issues with the build of the MDO34, I took a good look over the AFG31052. I am happy to say that I found no issues surrounding the construction of the instrument. Both halves of the instrument were clamped together tightly, with no signs of movement between the two. All screws seemed to be in place and tight.
Unlike the MDO34, that I am still reluctant to take a look at the internal build quality, I do not have this issue with the AFG31052 as I have a good means to test out its performance and can verify that I have not caused any issues by opening up the instrument.
Four torx head screws hold the rear cover in place. Once removed the rear cover can be eased off the body to reveal the inner housing. This appears to be a multi-piece aluminium housing, made up of three main parts, two at the rear held onto a front section torx screws.
The top rear section is unscrewed and must then be unplugged from the two sections left in place.
The removed section houses two power supply units and the cooling fan. The first power supply on the left is a 12 V unit from Lamda and the one on the right looks to be a multi-voltage unit from Artesyn. These are both open frame power supplies and look to be well made. I did not a 3.15A HRC fuse for protection on the Lambda supply. The other supply looks to be unfused, but there may be some electronic protection employed. The cooling fan is from Sanyo Denki. The picture below is within the internal shield over the PSUs removed.
The use of pre-manufactured power supply units, makes it relatively simple to replace them, should issues arise at a later date. Personally, I am not to fond of internal fuses hidden away inside an instrument.
The bottom rear section is held in place with Philips screws into some posts at the bottom and a torx screw at each side. The PCB plugs into the front board in two places and two ribbon cables come from this PCB and disappear underneath the front board.
Once removed, this section turns out to house a 32 bit 800 MHz Arm Cotrex A9 microprocessor from NXP alongside the IEEE 488 controller, some memory and a CR2032 backup battery. Further along to the right, there the rear BNC connectors for the ‘Add In’, External Modulation and Reference functions.
To the left of the section is a small mains input board covered by a mains filter unit. As the filter unit is unswitched, it looks like this is what was giving me the small load current measured whilst the AFG31052 was switched off. There is a glass 3.15A quick blow fuse installed on this section of the PCB before the mains is taken away to the PSUs in the upper rear section.
There appears to be a power supply section in the upper left corner with regulators mounted on heatsinks.
The rest of the board, then appears to be split into the two channels, housing the main components under heatsinks. In between the two sets of heatsinks is a single Analogue Devices A1851, 16 bit audio DAC and two further Analogue Devices AD9739BBC, 14 bit, 2.5 GSPS, RF DAC units. To the right of the black heatsinks are six DDR3L SDRAM chips.
| {gallery} Front PCB |
|---|
DC regulators installed on heatsinks |
AD9739BBC 14 bit DAC for each channel |
Internal memory banks |
There is certainly some good quality to the build of the AFG31052. All the boards were clean with no signs of debris or solder residue. The boards are neatly laid out and there no ‘calibration modifications’ present.
Powering Up
The AFG31052 has a standard IEC C13 connector to the rear for connection to the mains. A mechanically latching front mounted on/off switch is installed for turning the instrument on and off.
The AFG31052 goes through a boot up sequence that lasts around one minute. The start-up tests are displayed on the screen as it goes through them. On finishing start-up, the user is left at a home screen, from which the auxiliary software installed, basic operation of advanced operation mode can be selected. Selection of the function could only be achieved using the touch screen, there did not appear to be any option to utilise the mechanical controls to the right of the screen.
The AFG31052 has a nominal power consumption rating of 120 Watts and I used a Tektronix PA1000 analyser to verify this. Interestingly, whilst plugged onto the mains, but not switched on, a slight power consumption of 3.5 mW was measured, that I had not expected, given the type of mains switch that was installed.
When switched on, with no outputs enabled, a nominal load of 57 Watts was measured. Two load tests were then conducted, the first with a 1 Vpp output into 50 Ohms on both channels and then a 10 Vpp output into 50 Ohms. The load measured was 59.24 Watts and 62.92 Watts respectively, well below the nominal rating.
| {gallery} AFG31052 Power Consumption |
|---|
Power drawn with both output channels off |
Power drawn with 1 Vpp output on both channels |
Power drawn with 10 Vpp output on both channels |
Throughout the power measurements, the power factor was measured at circa 0.88. I then took a look at the voltage and current traces on the analyser. The voltage, seen as the red trace, is fairly symmetrical, but a high level of distortion was seen in the current trace, seen in green.
Seeing this high level of distortion, I decided to enable the total harmonics distortion (THD) measurement into the analyser. The voltage harmonic distortion was 2.7 % and the current harmonic distortion, quite significant at 34.9 %.
Back in my RoadTest of the MDO34, I carried out a test in the field utilising a battery power cube as a power source and suffered from noise on the traces captured on the scope. I carried out three test configurations, with the power cube supplying the AFG31052 and mains supplying the MDO34, main supplying the AFG31052 and the power cube supplying the MDO34 and finally the power cube supplying both of the instruments.
In all cases, the display on the screen was steady with no signs of rise or distortion in the traces. Turning on InstaView on the AFG31052 did show a little distortion in the output signal when the frequency was taken above 1 MHz.
Verification Tests
The complete set of tests carried out and the results obtained can be found in the separate blogs;
AFG31052 Verification Checks Part 1
AFG31052 Verification Checks Part 2
This section of the RoadTest will just detail the issues found during the tests. The majority of the comparisons were made against a Rigol DG1022 and a Uni-T UTG962. The Rigol is circa 20 years old and is a basic 20 MHz waveform generator, but I have used it to produce air gap search coil signals in the past. However, I was not expecting it to perform anywhere near the AFG31052. The UTG962 is a 60 MHz device that is only a few months old, but coming in at £100, I did not expect great things from it.
Where possible, I also added in data from the built-in waveform generators of the MDO34 and Picoscope 3404A. Unlike the other three instruments, these are single channel devices.
The main issues surrounding the verification tests were with the AC flatness, Sine wave Distortion, Total Harmonic Distortion and Spurious Output tests, where all three of the main instruments produced out of tolerance results. I was convinced that the AC flatness test fails were due to the instrumentation I was using. The Sine Wave Distortion, THD and Spurious tests fails were not so easy to diagnose. They were initially all measured with the spectrum analyser built in to the MDO34 and the majority of the issues were surrounding the peak value of the second harmonic.
A further measurement was taken using the spectrum analyser built into the Pico 3404A and a drop was seen for all the instruments, that actually brought the DG1022 into specification. This seems to suggest that the issue may be due more to the test setup rather than an actual instrument fault.
In an attempt to overcome this, I tested the AFG31052 in a different area to try and determine if there was something in the local electrical system that was causing an issue with the second harmonics being detected. I moved the AFG31052 and scope to my garage that has its own separate earthing system to the house.
The tests carried out on Channel 1 of the AFG31052 did not really show any significant improvement. The calculated THD was 0.3713 % against the original measurement of 0.3794 %. An improvement of 2.1 %.
As the screen capture of the oscilloscope shows, the main culprit is again the second harmonic that is circa 49 dBc.
This is still above the manufacturer’s specification. I will have to do some more research to try and ascertain any improvements I can make over the test methodology. Given the reading on the second harmonic, I will not seen any improvement in the Sine wave distortion and Spurious output tests either.
I intended to improve the results of the AC Flatness test by utilising an AE20401 counter / RF power meter from Ascel Electronics. The power meter function is specified to work from DC to 500 MHz.
https://www.ascel-electronic.de/kits/14/ae20401-5.8-ghz-frequency-counter-/-rf-power-meter
In theory, this meter should be capable of giving comparable readings against those captured on the 8846A. In practice, this did not quite work out.
At the 1 kHz reference test, the 8846A read 3.97 dBm, where as the AE20401 read between -12.5 to -14.2 dBm across both the channels. In fact, the AE20401, did not start to pick up stable readings until around 70 kHz. This was much better than the previous attempt, and whilst it does not appear to operate to the specifications from the manufacturer, I was able to get sensible readings between 100 kHz and 5 MHz, where the original meter failed to read correctly. The new readings obtained for the AFG31052 can be seen in the table below alongside the original test data.
The 100 and 300 kHz levels are slightly out of tolerance, but the rest of the readings are acceptable. Since the two lower readings are in tolerance when measured with the 8846A, I think it is fair to say that the AFG31052 has passed the AC Flatness tests and the issue lies with the RF power unit in the AE20401.
If anything, these tests have shown that measuring parameters at RF has a lot of complications. I did find a an error regarding the calculation example that Tektronix provide in their verification manual, that they will hopefully address. I did find a little inconsistency in the recommended measurement points for the AFG31052 I was testing.
Generating Partial Discharge Pulses
This was the main aim of my experiments using the AFG31052. There were a number of aspects that I believed would improve the ability to generate partial discharge pulses for testing the measuring apparatus over the built in arbitrary waveform generator of the MDO34. The full details of this testing can be found in my third blog Tektronix AFG31052 PD Partial Discharge Pulse Generation.
The Rigol DG1022 has limited memory depth for generating arbitrary waveforms. I have created waveforms for air flux probes and used the arbitrary function of the DG1022 to play them back to reasonable effect, but it just does not have the capability for simulating partial discharge pulses along with the reference waveform.
So how do the two paths for creating arbitrary waveform generator in the DG1022 and the AFG31052 compare?
The major difference between the AFG31052 and the DG1022 is with the built-in ability of the former to merge the pulse and reference signals. With the DG1022, this has to be done externally, and to do properly, really requires further hardware.
It is also much simpler and easier to generate the required waveforms with the Tektronix unit. Where as working with the Rigol, really requires the use of a computer running the Ultrawave software, to convert files or draw signals manually, both of those functions can be completed on the AFG31052 directly, without the need of a computer.
Some of this will undoubtedly be due to the 20 year age difference between the two units, it would be interesting to obtain a more up to date Rigol unit to see if they have made improvements.
Using the ArbBuilder to manually draw a partial discharge pulse worked quite well after a few attempts. Each pulse was saved to the internal memory and I then used the Advanced Mode of the AFG31052 to generate a pattern from the individual pulses.
This proved to be a very flexible way to generate a partial discharge pulse pattern. A 50 Hz reference waveform was a bit more of a challenge to add to the pattern as the ‘Add In’ function only appears to work in the Basic Mode of the AFG31052. An extra few steps are therefore required to capture the partial discharge pattern on an oscilloscope, so that it can be played back on Channel 2 as an arbitrary waveform in Basic Mode and then the ‘Add In’ function becomes available for Channel 1, and a combined pattern and reference waveform is successfully generated.
There is an added benefit of the AFG31052 in that the phase relationship between the partial discharge pattern and the reference waveform, is easily adjusted using the phase control on either of the channels. This adds a further dimension to the flexibility of the system.
I did look at the use of InstaView when playing back the partial discharge patterns to see how the InstaView screen would compare with the oscilloscope screen captures.
Sadly, this did not do too well. It looks like InstaView will only work with the AFG31052 operating in basic mode and not advanced. Whilst the InstaView system will display the signal generated by Channel 1 in Basic Mode, it would not recognise the pulse pattern being applied to the ‘Add In’ facility of Channel 1.
Therefore whilst the oscilloscope displayed the combined waveform, InstaView only displayed the 50 Hz reference waveform.
I came back to the partial discharge pulse creation after finishing my other tests and I then managed to get the Instaview to work and display a combined waveform. The solution to this was to reverse the function of the Channels and playback the pulse pattern on Channel 1 and add in the 50 Hz reference generated on Channel 2.
Whilst I say it was a combined waveform, there was a significant element of distortion in comparison to the screen capture of the oscilloscope, as seen above. An element of the pulse pattern can be seen as a little block sitting on the reference waveform in the positive pulse, but the reference waveform is misshapen at the start of the cycle. The block of pulses is not repeated in the negative cycle, that is more representative of a sine wave.
MOSFET Testing with Double Pulse Application
My experiment with the Double Pulse application for the AFG31052 is to test out the performance of some the MOSFETs I have identified inside the insulation testers I have reviewed. Overall, I managed to identify and obtain five MOSFETs and one IGBT.
Tektronix AFG31052 MOSFET Tests
The test setup followed the guidance within the Tektronix Double Pulse Application Note with a couple of departures. The application note is written in terms of testing the performance of the MOSFET, where as I was more interested in the performance of the MOSFET in a specific application. I therefore carried out the tests using a 9 V supply, typical of the battery pack for an insulation tester and instead of an inductor as the switching load, I used a pulse transformer from an insulation tester.
The Double Pulse software needed to be downloaded from the Tektronix website and installed onto the AFG31052 via a USB drive. This was a quick and trouble free process. The result is an application icon on the home screen next to the ArbBuilder application. Unlike the ArbBuilder I can only access the Double Pulse software from the home screen using the touch interface.
As I am using a differential probe and current clamp, I need to synchronise these to the other channels on the oscilloscope. The AFG31052 produces a good sharp pulse with a 4.5 ns rise time to enable this to be carried out.
Channel 1 and Channel 2, that utilise Tektronix probes are naturally synchronised. Channel 4, using the voltage differential probe, needs a 3.9 ns adjustment of the deskew to align the pulses. The rise time at just over 8 ns, is twice the Channel rise time and I can live with that.
The issue comes with Channel 3 that not only requires 175 ns of deskew adjustment, but has a rise time around 100 ns. Given the nature of the operating time of the MOSFETs, I feel that this is just too much and therefore utilised a 1 Ohm shunt resistor to provide the current measurement as a voltage drop across it.
I cobbled together an arrangement from a BNC adapter, 50 Ohm resistor some wire and 4 mm test sockets to carry out the channel synchronisation with. This is something I would like to improve over time and will look to get a PCB made to aid with the setup.
Testing of the MOSFETs is relatively straight forward and the data can be extracted to make the necessary calculations on the performance of the MOSFETs. By testing with a 9 V supply, very low levels of energy dissipation are observed. A lot of the timing measurements do not meet the manufacturer’s specifications, again likely to be due to the different test arrangement I used. A lot of the data sheets give a test arrangement for measuring the timing characteristics of the MOSFET.
I was caught out when testing the IGBT, and this device did not have a body diode within it, and responded with a high voltage spike in between the two operating pulses. This was resolved by temporarily installing a Schottky diode across the top IGBT.
Current Clamp Testing
Having a powerful waveform generator available, I decided to look at testing the performance of current clamps using a clamp table and my current amplifier.
Testing using a standard 50 Hz sine wave is an obvious choice, but I want to experiment a bit further using the arbitrary waveform capabilities of the AFG31052 to load in a captured magnetising current from a transformer.
As with the partial discharge patterns, the captured waveform is converted to a tfwx format that the AFG31052 can interpret. It then loads without any issue.
I used both the boxed amplifier and the board mounted parallel amplifiers to carry out the measurements. I had no problems driving either of the amplifier arrangements with the AFG31052.
This test scenario worked well, and I was able to compare the accuracy of an averaging clamp meter against an RMS clamp meter in terms of frequency response and measurement of non-sinusoidal currents.
I will look to advance the testing of clamp meters to verify the inrush measurement capabilities of some of my other clamp meters.
Thoughts on Instaview
I tested the InstaView function at various stages throughout the RoadTest with mixed results. It is a clever piece of technology, although it does seem a little rough at the moment. It can only be instigated from the Keypad, I could not find a way of enabling it from the touchscreen.
It does not appear to be always available. It did not work whilst in advanced mode, nor did it work when using the modulation function with the outputs.
There are inconsistencies when it is available. It certainly did not like the affect of an attenuator on the cable. It would not record a propagation delay and the voltage measurement is inaccurate. There are times when the data reading is covered by the trace. There seemed to be some issue with using it with the ‘Add In’ function for Channel 1. Using Channel 2 to add in the standard sine wave produced some results, but still with distortion. Using Channel 2 to add in the arbitrary waveform pulses caused InstaView to miss them out completely and just display the Channel 1 waveform.
This leads to frustration and confusion whilst using the function.
To my mind I am not sure what Tektronix are hoping to achieve with InstaView from a practical perspective. It does not have anywhere near the same functionality to that of an oscilloscope. Given the cost and nature of this instrument, I think it is reasonable to assume that anyone owning the AFG31052 would be working at a bench and likely to have a reasonable oscilloscope to hand.
So why would they not just use the oscilloscope to measure the output directly?
If the technology was more reliable and installed on a arbitrary function generator designed to be more portable, with a field based work-scope, I could understand the desire for it to do away with having to take an oscilloscope along with the function generator. However, my work-scope very rarely requires the use of a waveform generator out in the field. It may be that some do use such an instrument out in the field, and that experience is missing from my synopsis.
Conclusions
This has been an interesting RoadTest and I have learnt a lot working through my test proposal. The AFG31052 is undoubtedly a piece of good quality test apparatus with plenty of functionality to hand.
The build quality is excellent on this unit and the controls have a good feel to them and are intuitive to use. I did find that the operation is split across the touchscreen and keypad, with some functions only available on the touchscreen and some only on the keypad.
In that sense it seems curious as to why they installed a button to turn the touchscreen off. For this RoadTest, it would have been better to have had that button as a screen save function.
I did enjoy using the advanced mode and its ability to create complex waveforms. The ease of use of the touchscreen with the ArbBuilder application was very much appreciated. With this facility, I can dive deeper into simulating partial discharge activity without the need for a few thousand volts floating around.
The Double Pulse application seemed to work very well. Personally, I would benefit from carrying out the tests again utilising the test setup specified by the MOSFET manufacturers, to give some confidence in my test results specifically made around the functionality of the MOSFET within an insulation tester.
I still need to get to grips with the PWM functionality, to see what the issue is and why my experiment to test the pulse transformer from an insulation tester failed.
I struggled with the InstaView system. For me it seemed a bit inconsistent and did not add much value to the unit. Towards the end of the RoadTest, I did not find myself using it, and did not find myself thinking that I should of used it with an experiment or two. The oscilloscope always seems the better option.
Comparing the AFG31052 to some other brands of function generators was very interesting. It is quite incredible how well a basic function generator can perform when put against this unit from Tektronix.
Whilst they can match some aspects of the AFG31052, they start to struggle with the more finer points. They could not match the AFG31052 for rise and fall times. Quite often the control of them is not as intuitive as the Tektronix unit. The harmonic distortion seemed lower and the output more consistent on the AFG31052 than on the other units, that displayed more noise in their output waveform.
This completes my RoadTest of the Tektronix AFG31052 Arbitrary Waveform Generator and I hope that the element14 community find it useful and informative.
Many thanks to element14 and Tektronix for giving me this opportunity and I am sure I will have many more years service and experiments along with the AFG31052.
Top Comments
Nice road test review Donald. Good work!
Nice work Donald, Really helpful!
three-phase Great review. Enjoyed it a lot.