RoadTest: Step Down Converter EVM
Evaluation Type: Development Boards & Tools
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?: This EVM is as unique as any other product and has it own performance characteristics. It is therefore a bit too difficult to make a comparison.
What were the biggest problems encountered?: From the EVM aspects the test points on the TPS55424EVM should be replaced with something different. The TPS54A20EVM on the other hand is excellent. The only other challenge was certain test parameters I carried were very time consuming, but fortunately had plenty of time to carry them out. Finally the discrepancy issues I found with output ripple. This has set me back, but fortunately only slightly.
In engineering developing a product requires several blocks to work with. These blocks can be dealt with one person or a team of people. In electronic engineering these blocks will require speed and efficiency to develop a system in as short as time as possible and this is where Evaluation Modules (EVM) come into play.
One option is to design a circuit from the ground up, the IC is selected, a schematic is created, sent off the a PCB house and then returns for diagnosis. Fault are cleared and board is modified and a new version go back to the board house to be remade. This takes an enormous amount of time and the costs are extremely prohibitive in today's product schedule. Now companies such as Texas Instruments create Evaluation modules EVM's to allow engineers to bypass many of the modes mention previously. In fact the cost of several EVM's is usually less than trying to create your own design. From here we investigate an engineers preferred method of designing.
For this demonstration I will be looking at two products, the Texas Instruments
These products shipped by Element14 as part of a Road Test review. For this I would like to thank Element14 and Texas Instruments for providing the opportunity to test the EVM’s.
For this road test to be so comprehensive there was a great deal of support and I wish to acknowledge them:
The items arrive I good time and is quite normal and very well done by Element14, I have had no real issues in the past and have not been let down again.
Instead I would like to comment on the packaging.
Texas Instruments – The use of a box to deliver the EVM’s and the large size only to end up with two single evaluation modules of a small comparative size.
I have an active mind in relation to sustainability and find the use of these boxes to be a waste of space. As an technician/engineer I have no interest in presentation on supply and the use of these boxes, I personally cannot justify this waste. To the left are the boxes supplied and shown in the image below is what I would call a more efficient supply packaging which can look presentable in most circumstances. The packing a Padded Mailer was all I had at the time, but any padded container in small stature could make a difference.
There are significant benefits, one is size reducing transport costs and subsequently either lower sale prices or increasing profits. The other benefit is the small reduction in footprint, although boxes can be recycled, the recycle incentive being good is still not 100% efficient, therefore any small reduction of discarded products is a small positive, with large quantities this could lead to a significant impact on the environment and costs.
This packaging is only a suggestion. The packing though is excellent with regards to static and protection ensuring the items arrived in excellent condition.
Products at a Glance
First observation quickly shows the products benefit, it is small, and that is just the packing. A typical constraint with development boards or EVM’s is that for engineers there is a need space to place probes and as expected Texas Instrument have not let us down.
With plenty of necessary test point both boards are well laid out and easily accessible to monitor and measure voltage, current and other necessary test points to satisfactorily characterised the EVM’s core components for my application.
The TPS54A20EVM-770 has excellent layout, containing the active components in the centre of the EVM, while placing and strategically locating test points throughout the EVM, colour coded to easily identify +ve (Red), -ve(Black) and other test points (various other colours), which can be easily identified at a glance
(Pictured to the Left).
The TPS54424EVM-779 although has a similar layout does not include the coloured test points, I find this slightly more difficult to navigate and visualise the probes position. The test points used in this evaluation module are soldered half rings, and from the Texas Instruments website image look like SMD resistors or capacitors. If I were to evaluate which is the better for testing the TPS54A20EVM would win in this respect. I could easily make test clips to match the colours and use these to identify easily during my characterisation testing.
The layout though is still very good and as far as general layout the TPS54A20EVM gets my vote as the better of the two. Being mindful that in no way the central active components in both EVM’s are there in a specific way to match the optimal design layout.
One notable placement is the input/output capacitors, these boards have been extensively designed and created to provide optimal performance and this is easily seen as the input and output capacitors placed very close to the chip to reduce loop inductance in the design.
For any engineer developing a product the need to come up with a design that is constrained by the specification set. Webench is a tool I use a lot. With a very simple starting section plugging some data into Webench front page then submitting opens an online GUI with numerous suggestions. After getting the open selection screen, I wanted the Integrated solution for relative design ease and then finally narrowing down necessary option the the TPS54424 came on top. To understand I wish to create a power supply circuit to power a Raspberry Pi with voltage monitoring and an enable pin to switch it on and off when necessary.
The beauty of Webench is the simplicity to start with, just a few parameters to get started and a stream of option come up. The process is simple and for the project sorting the most important specification for the project narrowed down several components that would be more relevant.
Initially Webench can be daunting at the beginning but after few attempts getting the best solution was really easy to narrow down the preferred parts with ease.
All the negative aspect with ease of use initially, Webench is a fabulous tool that allowed me to precisely narrow down my search to only a few parts with TPS54424 being one of them. The result selection was a competition between a few TI components. As I was ready at the time to order parts the Road Test came along at the right time.
For more experienced users of Webench the narrowing down of the right product will become easier and in my case the review allowed me to spend some time on manipulating information to get the best product in minimal time. Further more I hope in the near future use some more advanced tools to improve the design process. This is an essential tool for engineering a solution, I highly recommend it.
Webench design tools
Once the search has been completed the next stage for me is to ensure the most important factors are there. One my major design constraints for my project is efficiency. This will be a purely off-grid project. So I require to get the most efficiency for the project while maintaining very high regulation for the Raspberry Pi.
I could have spent a long time carry out simulations within LTSpice or TINA, but instead WEBENCH offers a much quicker simulation bench than I would have expected to do outside the reference tool. A good indication initially was the sorting for efficiency in the initial search.
Therefore WEBENCH offered me to simulate efficiency shown in the image to the left, and thermal performance shown in the right. I do not discuss the thermal performance till almost the end of this review but the similarities to the real life bench results I measured are astonishing.
It is important to remember that the simulation was based on my parameters for 5V 4A output using the TPS54424A chip.
In short Webench is a tool for me that adds to the popularity of selecting Ti Parts over others, mainly because it narrows down a selection very close in minimal time.
A recently selection process I needed as a example was a simple RF amp module where the original unit has been discontinued. Ti does not have a part that will suit so I had to use external sites to narrow down the search. Unfortunately I along with my boss are still searching.
Characterisation for the design engineer is a testing method to determine if a product is going to work for the task that is being carried. Characterisation may only need a few tests or a complete range of tests. I am fortunate enough to use the TPS54424EVM specifically for my project I am working on and will require full characterisation of the EVM for this purpose. I was fortunate that the Road Test coincided with my project and the TPS54424 came up as one component specified for the power supply section of my project. An extreme co-incidence.
For most new engineers that are cutting their teeth in engineering practices many manufacturers have application notes or instructions on how to carry out testing. There are 2 places I usually look, one is a test equipment manufacturer for details on testing and methods to carry out testing, while the component manufacturers have their own application notes on testing procedures. As I am using a TI part, Texas Instrument has an exhaustive number of application notes available, including live training modules. For this instance, a section called “Engineer It” is available from TI and has many training video and materials. For my project although I have some experience in power supply testing I still decided to go through the tutorials as there may be some new techniques to be learnt that I may not be aware of.
Engineer It – is well laid out providing step by step into methods of testing power supplies and is a good start. The videos are not too long and contain relevant information on the testing procedures. The link Engineer It - How to Test Power Supplies is a set of 4 training modules to characterise power supply designs. There was a couple of items of interest to myself although in general was relatively straight forward for my experience. For young engineers, site like Engineer It are valuable to maintain professional development ongoing in an engineer’s life.
Test and Equipment used.
Now onto the tests. Each of the tests are carried out in order of importance for my application:
Tests that I carried out are:
More can be gleaned from the results but most importantly confirming the above is more important for my project.
The test equipment I was using are:
Fluke TiR27 Thermal Camera - One of the best fault finding devices I have used in my time as a technician.
The problem I had was there was no way collect the data in the current setup and the requirement for licences for Agilent Benchvue Software were not available so I was unable to record the data automatically. I really didn’t want to write it down as I was taking notes, so instead took reams of video then later sitting next to my computer able to quickly punch in numbers. It is fortunate that I will be using the video later to edit a video version of the review later.
Still a very laborious task but found a method where I could combine multiple test into one video.
Along with some extra tests most of which were very quick including oscilloscope capture capability. The data came up with was amazing and decided to take it onto myself to try a new display method of showing plots.
Primarily these are 3D plots based on numerous reading taken I was able to get a reasonable display that showed some interesting data.
For myself, I decided to take on a different approach to taking the results for simulation, In this instance I have used 3D plotting method. The primary reason is to provide a complete performance analysis. It can sometimes take a bit of a thought process to understand them properly but after doing this for a while 3D plots show a lot of information at a single glance. In my plots the ability to see problems for example the instability issue in the TPS54A20 which can be viewed in the next section. Finally the ability to isolate specific zones to give a realistic operational area to determine the full operating conditions over a complete usage scenario for the project. My project supervisor made a valid point during my explanation of what I was doing and was "Too much information may not be a good thing". Only time will tell?
To the left is a plot showing Load and Line Regulation and to the right is the efficiency plot. As can be seen the plot shows very good with only very slight variation in levels.
For the regulation the voltage O/P starts from 1.208Vdc and ends at 1.218Vdc .The scale of the plot accentuates the variations but a plot starting from 0V O/P voltage has an extremely flat response. Also shown in the plot is an anomaly that I found very interesting. Wondering how you can get nearly 100% efficiency. The possible cause is instability and am sure a component change could rectify this.
It shows clearly why an oscilloscope is such an important tool in characterisation. This would quite possibly have gone unnoticed in purely meter-based testing. With 3D plotting the symptom may not have appeared if only 3 or 4 fixed voltages where taken as usually found in a datasheet.
Is this a manufacturing issue? possibly not. It may have been best to take a better characteristic when applying specs to a board.
Another test using the DC power analyser was the turn on a turn off voltage. The DC Analyser is comprehensive and has ramp capability that allowed a continues sweep from a low voltage to high voltage and taking the video I could easily see the exact switch on/off voltage. They were slightly lower. This pricked my ears up to the problem with the slight instability experienced in that the transfer characteristic may be a bit lower than manufactured.
All in all, the performance was satisfactory but suffered slightly in the efficiency and stability. As an engineer I would never spec my system to respond at maximum level instead apply a minimum 10 – 20% tolerance from maximum.
It is very clear the results fully passed my expectations and would be very happy to use this product in my equipment.
Though the next board is the one I am mainly interested in.
Further to this I also carried out a thermal test and shown to the far right is a thermal image taken using the Fluke TiR27 thermal camera. This image was taken during the converters maximum input voltage and output current. For an integrated FET DC/DC converter the temperature is very low considering. This would be relative to it efficiency.
The TPS54A20 is currently not the major board that I will be using for my project but this is a good supplement board for cascading in processor so most of my concentration will be no the next board, but further test were carried where there is a lot of images but not enough room to show in this review.
Also tested is the turn on time. The MDO scope image shown unfortunately I was unable to capture timing measurements instead can be derived with approx 100uS power supply rise switch on and a 1.2mS delay to start of the output. More than enough settling time allocated.
More trade offs can be sought usually with some performance caviets.
In this case I first took measurements as opened from the box with no modifications. 4.5 - 17Vdc I/P and 1.8V O/P. As with the TPS54A20 I reached slightly less efficiency. This would be a fairly accurate assumption, a couple of reasons I could think of but mostly due to the necessity of a large swing Vout would be fairly accurate, otherwise there would be no requirement for the TPS54A20 for low to mid current applications.
The Load / Line Regulation plot though is nothing short of fantastic.
It clearly shows a flat plane with excellent stability over all current and voltage input ranges.
By changing the z axis from 0 - 2V to 1.74 - 1.796V the plot looks at lot messier but does show what I feel an optimal design. I clearly shows a curved response with load regulation optimising the worst case in the central region. Any other method of design may have caused a downward trend in load regulation that could have started at an optimal point down to a much poorer regulation at higher voltages. This did confuse the theoretical calculation of load regulation at 100mA to 0, whereas the real regulation is 0.00085. Still impressive though but can hide many sins.
For the initiated Load regulation is calculated as
V(out-min load)-V(out-max load). This calculation can be very tricky to deal with a non linear regulation as shown here. For my project the way I would isolate this is taking the V(out-min) - V(out-max), at least would provide a figure as mentioned earlier.
For this configuration I also carried some ripple measurements.
Starting at 31mV output at 200mA through to 163mV output at 4A. Unfortunately the ripple did not match the specification sheet for 10mV at I(out) = 4A, Although I am concerned about this as the I/P ripple remained constant over the whole range at 19mV. There is noise at input and so I will need to investigate this further. I have also tracked back through my videos that were taken and confirmed the scope probes are correctly placed. What is interesting is the I/P and O/P Ripple is opposite to the datasheet and will require further investigation. Theory also suggests that my data is incorrect so further investigation will have to continue.
5V output test.
The advantage with the EVM's are they can be modified. In general the design layout is quite strict and can suit many methods of regulation.
There are though as I found the need to change more than one component, but my supervisor suggested I was putting a bit too much effort on a single aspect of the project. The test procedure I used was fairly primitive as I couldn't capture the data effectively.
The next series of tests dealt with changing the out regulation to 5V. As I would have expected requires a minimum E96 component selection to achieve a more accurate output voltage. I have purchased a resistor book to cover the E96 range, although not sure when it will arrive. I have in the meantime set the output to 5.2V, or with such an accuracy in regulation 5.2768Vdc.
As I was hoping the efficiency skyrocketed up to an unbelievable 98% far higher than I expected. This I expect is due to me changing only 1 component the reference voltage divider R8 from 12.1k to 42k.
Here I arrived at the first problem that didn't seem right. After changing the resistor I expect the output voltage to reach slightly over the 5V calculated to 40.266k, instead my output voltage was 4.28Vdc, this was a significant under estimation. To combat this as I knew the one of the errors of this magnitude could be from an incorrect voltage reference. I then recalculated the new voltage reference from the derived value and came up with a reference voltage of 0.556V. I then went back calculating the resistance to and came to 43.453k resistor.
The closest in the E24 range was a 47k and replacing R8 with 47k left me with 5.2768V from the calculated 5.32V. It was very close much better than the data sheet. There is obviously more errors to take into consideration and with time restriction I was unable to take it further. My supervisor was expecting me to take on the RF measurements for the project. So instead of looking into it further I carried out the same tests at the 5.3V.
The efficiency is great and very consistent as expected across the board. Although the efficiency was higher than expected curve matches the data sheet well. As mention previously I should be able to manipulate the X-Y axis to provide a efficiency and regulation trend for my project. Unfortunately Excel is limited in its option in this regards. I will though be looking at MATLAB or some other math program to look at the data in a more interactive way. In the meantime I feel my experiment was a good success. I have data now that I can easily manipulate to use in other projects in the future.
I do not need to show the Line/Load Regulation plot for the 0-5V range as it is exactly the same as the load/line shown above in the 2V test. I was not expecting any more. Although to the right is a plot showing a regulation output voltage of 5.26 - 5.28V and is still remarkably flat.
Other images are available showing On/Off timing and below is a ripple still image. All of which is the requirement of changing other components which will be done later. I will continue to post blogs on the update of my project in relation to anything regarding the power supply and keep the community informed.
This is only mean't to be a review on the product and I feel that I have gone a little further but I have learn't a lot on my methods for testing and record keeping. I understand some of the mistakes I have made and the resolution that I require to do to speed up the process and provide a more accurate result. There will be follow up posts on my progress and maybe if allowed some of the implementation.
In saying that this road test has been especially important as it relevance is very real for me. My experience was mixed, learning and applying my experience with new testing, but the product itself did not really let me down other than the output ripple which will be an investigation over time to determine where the real problem will be. It is hopefully I will locate the issue within the deadline of the Road towards the end of the month.
The DC/DC Converter though is excellent and number one in my situation when designing new systems. I feel confident that any issues will be resolved and more so will have excellent product to add to my project.
To add, it took a while to create this review as I was struggling with the web interface, not something I have used much but now have an idea for future blogs I hope to use.
Excellent review with a great mixture of content. Well done.
A very nice review. It looks like to you captured a lot of data and generated a much of interesting plots. I hope that you have good luck in narrowing down the issues that you were having with the ripple…
Excellent review with a great mixture of content. Well done.
I have a video camera facing the test equipment and read of the screen into excel. I fortunately have Dragon Naturally Speaking V12 and could talk it through. The first time by typing took on/off 3 days(5 hours) to fill in the information. The second using Dragon took on/off 1.5 days(2.5 hours).
That is the reason why I do not recommend it. My case was so that I can devise a solution and compare it my improved concept in the future. I have already a plan, but will take some time before I can post.
I knew 3D plots would invoke some interest, my method of gather information though I do not recommend as it was extremely laborious. In saying that I will be investigating a new method of gathering the required information with the hope to log the data more efficiently. The DC Analyser fitted with the right options could have produced a better data collection but the option was not fitted.
The package is simply Excel, but it is extremely limited. I will be posting more blogs on this subject in the future because it is very interesting.
A very nice review. It looks like to you captured a lot of data and generated a much of interesting plots. I hope that you have good luck in narrowing down the issues that you were having with the ripple on the supply.
To add, it took a while to create this review as I was struggling with the web interface, not something I have used much but now have an idea for future blogs I hope to use.