RoadTest: Enroll to Review the Analog Devices Digital Isolator Eval Kit - MAX2256XAEVKIT#
Author: strb
Creation date:
Evaluation Type: Evaluation Boards
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?: Other digital isolators like ISO776x series and Si866x series. From ADI, there is also the MAX2266x series which is very similar
What were the biggest problems encountered?: Soldering the IC to the evaluation board was a bit tricky, but it's available a complete evaluation board with a MAX22565C if you don't feel confident about doing that.
Detailed Review:
Isolation between different circuits is a key factor for safety and noise control.
However most of the times is not as simple as placing an optocoupler and calling it done due to several requirements such as speed and delay constrains, power consumption, reliability over the long term and much more.
That’s where digital isolators come in, providing best in class performances for connecting digital signals in isolated systems.
In this review I will explore functionalities and performances for the MAX22565C using the MAX2256XAEVKIT#
Before starting I want to thank element14 and Analog Devices for giving me the opportunity to be part of this RoadTest.
The MAX2256 series is a family of 6-channel digital galvanic isolators featuring ADI’s reinforced, fast, low power technology. Each isolation channel features a dual isolation barrier in series meeting VDE and UL reinforced standards.
Image from Analog Devices introductory video.
Main features of this IC family are:
Another feature that I like is the possibility to set a default state, allowing to get a known value in case something on the input side goes wrong (like a microcontroller reset) and improving system reliability.
By comparison, the most similar isolator I was able to find is the Si866 series. Still, it falls short reaching a maximum data rate of 150Mbps with a higher power consumption and with a slightly diminished supply voltage acceptance. Other specifications, such as common mode transient immunity are comparable.
Another close match is the ISO776 series which can reach half the maximum bit rate of the MAX2256 series. It has higher current consumption and it can go down only to 2.25V supply compared to 1.71V of the MAX2256. On the other side, the ISO776 has a better CMTI and can tolerate temperatures down to -55°C compared to -40°C.
Overall, the MAX2256 series shines as the fastest, lower power consumption digital isolator I was able to find.
The RoadTest kit comprises two components:
The board came into a standard cardboard box and inside it was packaged with an ESD safe bag further wrapped with pink packing material, presumably also ESD safe. It all fits inside the box nicely and it’s not able to rattle too much during shipping.
The MAX22565CAAP+ IC has been shipped inside an ESD safe box and kept in place with two pieces of ESD foam. Great care has been taken to protect the MAX22565 from ESD, both the foam and the cardboard are conductive!
When I received it, the seal was already broken. Probably the customs was interested in the MAX22565CAAP+ and wanted to have a closer look
{gallery}MAX22565C |
---|
The MAX2256XAEVKIT# board looks very nice. I like the soldermask color and the silkscreen is nice and sharp.
I was pleasantly surprised by the ease of reading the QR code. Being printed on the silkscreen, I was initially worried about its readability but the phone scanned it instantly, showing a link to the MAX22565 IC where you can find all the general information required, the datasheet, evaluation boards and a brief introductory video. It’s also available an Ibis model to carry out signal integrity simulations if needed.
On the MAX22565CAEVKIT page there is an ODB++ file with the board design information (oddly, this file is not available from the MAX2256XAEVKIT# page). Other than the soldermask that apparently has a custom color, one noticeable thing is the board stackup, precisely the dielectric between top layer and the first inner layer. ADI decided to go with ROGERS 4003C, which is a special material with tight control on dielectric constant and low loss. Pretty fancy!
I initially planned to try soldering the MAX22565 to the board using different techniques, but once I realized that they shipped to me only one IC I decided that was too much of a risky move soldering and desoldering multiple times the same IC, so I decided to use other similarly packaged IC to start with and only at the end solder the “main star of the show”.
I’m sure the board will withstand a bit of abuse without any problem because both dielectric types used to manufacture this PCB are high quality with high Tg.
All the unpopulated board pads are pre-soldered, which I think is nice and hopefully will help a bit. If you have good eyesight magnification is not strictly necessary, but I will advise it nonetheless. I used a maglamp I bought a while ago at a local store.
The soldering setup
First test: soldering iron
For the first soldering test I used a WE1010 equipped with a 0.8mm chisel tip (model ET-HL). The temperature was initially set to 300°C but I had to bump it up a little bit due to the small tip size and the heavy presence of ground planes that like to suck heat away.
The soldering was uneventful. The space around the MAX2256 is more than enough to freely maneuver the soldering iron and I accidentally bump test jumpers only once with the soldering iron body. Make sure to use some flux to help avoid bridging.
Second test: hot air
After cleaning all up from previous tests, it’s time to try soldering with hot air. I used a MP740784 set to 350°C, 25% airflow and equipped with an 8mm nozzle.
The initial plan was as follows:
Unfortunately, it did not work for me, as I was not able to apply just the right quantity of solder paste. I applied too much paste and due to pins being quite close one to another I bridged a couple together. The fact that my paste is old probably doesn’t help either .
I ended up tinning pads with the soldering iron and then soldering the IC using the hot air station and that worked well enough. During the heating process, I had to keep the IC steady with a pair of tweezers to avoid blowing it away but other than that no particular issues.
Soldering the MAX22565C
Given the experience just made soldering with the MAX2256XAEVKIT#, I decided to go for the hot air, as it can yield a more “aesthetically pleasing” solder joint. To make my life a little bit easier, I initially thought about using a hot plate to pre-heat the whole board and then reflow the MAX22565 using hot air. I then tried to remove the standoffs from the board but I wasn't able to get them off easily and I didn't want to ruin those trying too hard, so I opted to not use the hot plate and stick to hot air only.
So, after cleaning all the mess I’ve made until now, I pre-soldered once again all pads with the soldering iron. Looking at footprint dimensions and placing the MAX22565 on top, I think that this footprint is not though to be hand soldered but rather they have optimized it for an automatic assembly line. I found aligning the MAX22565 with the footprint a little difficult due to the small margin available on the pad length.
A bit of patience and a healthy dose of flux took care of it and the end result I think it’s more than acceptable.
{gallery}MAX22565C soldered on MAX2256XAEVKIT# |
---|
Before starting with some measurements, why would anybody swap the old trusty optocoupler with a digital isolator, given that the latter has usually a higher price tag? Well, there could be many reasons.
The first one that can be easily seen by eyes is the reduction of needed board space. The MAX22565 series fits 6 channels inside a 20-SSOP package. There are some multi-channel packages for optocouplers as well but they are not as small as ones used for digital isolators and furthermore they usually contain only up to 4 channels.
Another great difference is the temperature stability and reliability over the long term. Optocouplers as core components use an LED and a phototransistor as a receiver. Looking only at the diode, they suffer from IV curve shift due to temperature, causing the LED “turn on point” to be shifted.
ch1: optocoupler input signal, ch2: output signal. output variation between ambient temperature and hot environment
Furthermore, the LED loses brightness over time degrading the current transfer ratio. These facts can limit optocouplers in applications where high reliability in harsh environments over a prolonged period is necessary.
Speed is another important factor as well. Optocouplers usually are quite slow, allowing only data transfer rate up to tenths of Mbps (the best I could find reached 50Mbps) while digital isolators can easily reach and exceed that speed.
Common mode transient immunity (CMTI) is another strong point for digital isolators. Due to internal construction, unwanted capacitances are minimized thus allowing greater transient immunity even when compared with specialized high-speed faraday shielded optocouplers and the usage of a differential structure like in the MAX2256 series further improves its performance.
Last but not least, all these performance improvements come with a reduced current consumption, allowing a better fit for power critical applications as well.
As every good evaluation kit, the MAX2256XAEVKIT# has an user guide with a brief procedure to get started, good! The hardware is well described, showing how to set jumper link positions and how to use the board to get the most accurate propagation delay measurement using the additional calibration traces they provided on the board.
A brief calculation gave me the expected delay for the board traces: between 400ps and 500ps… yeah, it’s not something I expect to be able to measure with my current equipment (Rigol DS1054Z).
Reference trace delay: not measurable with my scope
As predicted, with a minimum timescale of 5ns/div, my scope is barely able to see this kind of delay.
As a personal comment, I was initially surprised that ADI did not say to deskew probes but after reading through all the propagation delay measure procedure they provide it makes sense. By subtracting from the total delay the one measured with the calibration trace you are nulling out any probe skew, provided that you must keep the “probe order” the same. Still, as good practice, I checked my probes for skew especially for attempting to measure the absolute value for propagation delay of calibration traces.
With that out of the way and knowing better my instrument limitations, I tested the behavior of DEFx jumpers that, according to the guide, should set the default output state. I moved the default state from high to low, powered up the board and… it seemed to not working properly. I checked things a little bit just to discover that among all pins the only one that decided to not properly solder to the board was one of the DEF pins… but that’s down to me, probably not pre-tinning enough the pad after all my initial experimentation (if you’re not messing up too much at the beginning as I did you should not have such problem ). A quick touch up with the soldering iron and all started working as expected!
Let’s move to some more interesting measures.
As you can see, for my first set of tests I didn’t stress too much for perfect probing techniques. Knowing that my equipment is the limiting factor and not looking (for now) at signal integrity but only at timings I preferred a simpler probing setup. In the next images, the slight ringing you will eventually see is due to probe ground leads and not to MAX22565C IC.
Timing performances
Starting with propagation delay, as expected it proved to be aligned with the specs given by ADI (testing with 1.8V supplies). I would not stress too much about the absolute value I measured knowing I’m not nulling out trace delay simply because I’m not able to reliably measure it.
{gallery}Propagation delay with Vdd = 1.8V |
---|
Channel to channel skew at ambient temperature proved to be much better than the maximum provided in the datasheet, being hardly noticeable in my case.
ch1: input; ch2 and ch3 outputs, no visible skew
Glitch rejection is unfortunately something I’m not able to verify but I somewhat managed to check pulse width distortion. Accounting for measurement errors, it’s aligned with the typical value provided in the datasheet.
{gallery}Pulse distortion |
---|
ch1: input signal; ch2: output signal. Pulse distortion estimated around 500ps
After improving a bit my probing set up, I tried obtaining some sort of eye diagram. Unfortunately, I ended up testing my scope performances . Jokes aside, the eye is as open as my scope can display and it is in fact limited by my system bandwidth and not by the MAX22565C. The image below shows the rise time limitations.
Improved probing for the "eye diagram"
{gallery}Eye diagram attempt and datasheet comparison |
---|
Unfortunate eye diagram attempt. My scope bandwidth (50MHz) is the limiting factor. Comparing it to what you should see
Slow input behavior
Being rated down to DC, I wanted to explore the input/output relationship for a slow input signal. In my test, I opted for a 1Hz triangular wave. The result didn’t disappoint me, obtaining a fast clean bounce-free transition on the output thanks to a small input hysteresis (about 80mV). If you have an extremely noisy slow changing input signal, you may want to look at the slower B option, which has a greater input hysteresis. This “hard switching” behavior is another advantage the MAX2256 shows against optocouplers, which have a wide transition zone making them less than ideal for digital signal transmission.
{gallery}Slow input behavior |
---|
About 90mV of hysteresis |
Zoom on the transition: nice and clean |
Power on/off and undervoltage behavior
The power on/off and undervoltage behavior is well documented into the MAX2256 datasheet as it shows input/output relations with different cases and different default output states with actual scope captures. I measured one case as well and as expected it matches perfectly to what is reported in the datasheet.
Datasheet curves for power on/off
ch1: Vdd B; ch: output waveform (set to default high)
Looking at the evaluation kit guide, it is stated that DEFA and DEFB jumpers must be set coherently one to another (both defaults high or low). Out of curiosity, I decided to test this prohibited setting.
I set DEFA for a default low and left DEFB for default high, no input signals. Vdd A is connected to my power supply and Vdd B is connected to my function generator. While Vdd B where cycling, I turned on the Vdd A supply and observed the output behavior on the first channel.
While side A was unpowered, the output behaved seemingly following the DEFB default high state. After powering up side A as well, DEFA setting (low state) apparently took over and the output went low. The transition between these two states however is probably not well controlled (that’s only my supposition, it may be wrong) and that’s why it's advised to set DEFA and DEFB with the same setting.
ch1: Vdd B; ch2: outup (side B); ch3: Vdd A. Side B set to default high and side A set to default low
I asked ADI tech support a question about it, taking the opportunity to try their responsiveness. They promptly responded after a couple of days and this is their answer: "DEF A and DEF B must be set to the same default state not different. It is a must. Both must either be tied high or low."
So I guess there are two takeaways here:
Power consumption vs frequency
Low current consumption is one of the major strengths point for the MAX2256 compared to competitors and I was curious to verify it. Looking closely at some datasheet curves seems like there is some peculiar stuff going on especially for B rated parts (25Mbps) at low input/output speed while nothing special seems occurring on the output.
{gallery}Current consumption datasheet curves |
---|
Starting with “output side” current consumption I decided to not use any explicit load capacitance. Any other additional current consumption can be estimated with the following formula:
I = f * Vdd * Cl
Where f is the switching frequency, Vdd is the output side supply voltage and Cl is the load capacitance.
Side B current consumption
Seeing that at low frequency my results are identical to datasheet ones but not at higher data rates, I was dubious about my results, so I tried to dig a little deeper. Using provided fabrication files I measured the output trace length, which is around 50mm. Using a calculator, that specific trace should have about 5.5pF of capacitance. Using the above formula this load would provide an additional 400uA current demand at 40MHz and 1.8V, which is almost the difference I have (about 600uA at 40MHz and 1.8V) between my measure and datasheet information.
To prove my theory, I decided to desolder R7 to disconnect as much parasitic capacitance as possible from the MAX22565C output and I repeated the measure.
I was quite impressed how much difference there is in power consumption only due to a small trace and a pin header, but considering the high data rate it’s able to support it makes sense. At high speed even small details can make a substantial difference. This time, my measures are spot on with datasheet’s graphs!
Side B current consumption with (old plot) and without trace connected to the MAX22565C output (new plot). The difference is huge!
Luckily on the input side current measures went smoother. There are one/two points where oddly the power consumption isn’t monotonic and I repeated the measure multiple times, even closing up sample spacing but the little jump was still there. Nothing to worry about but I found it interesting. Maybe there’s some magic going on inside the IC.
{gallery}Input current consumption |
---|
Transient immunity check
The last thing I planned to test is the transient immunity check, or at least I wanted to see how further can I push my instrumentation to test the MAX22565C.
I connected my signal generator between gndA and gndB (ground reference on side A, B is floating) and a couple of probes hoping to use the math channel to get the output signal referenced to the floating side, but it didn’t work. Connecting both probes on the same signal shows as much as 1V difference between channels and that completely invalidates any results.
Using math function is not good enough: up to 1V mismatch between scope channels during high speed transient
I don’t own a differential probe but I wanted to carry out at least a half meaningful test. Taking advantage of the one channel in the opposite direction, I decided to fold back the signal from ch1 to ch6 thus being able to directly read the digital signal without needing a differential probe or closely matched probes/channels on the whole frequency range.
I maxed out my signal generator producing a pulse with a slew rate around 1.1kV/us and checked the signal at ch6. From far enough it isn’t visible anything, however zooming in I was able to capture a tiny 20mV signal superimposed on the digital data. I’m more than an order of magnitude away from the typical CMTI value that the MAX22565 should be able to handle but still this is a remarkable result to me.
{gallery}CMTI |
---|
ch1: transient between gndA and gndB; ch2: output signal on side A ch6
This kit is for everybody who wants to try itself performances of the MAX2256 digital isolator series and maybe refine some datasheet data (although I found it comprehensive and well explained). For some proof of concept designs it can be easily inserted between two isolated systems thanks to ease of connection with provided SMA connectors or through pin headers.
The MAX2256 series digital isolator aims at high end designs that need best in class level of speed and power consumption, with all the added benefits and general performance increase that a digital isolator can offer.
Product performed to expectation: 5/5. The MAX22565CAAP+ performed as expected, confirming information available in the datasheet and often exceeding my measurement capabilities.
Specifications were sufficient to design with: 5/5. The datasheet covers all the information I can think of and the evaluation kit guide is well written and shows how to get started.
Demo software was of good quality: 5/5. Not applicable for this RoadTest.
Product was easy to use: 4.5/5. Soldering was just a bit trickier than expected, but to overcome that ADI provides an evaluation board with pre-soldered MAX22565CAAP+. Test points and connectors provide all you need to test this digital isolator.
Support materials were available: 4.5/5. While there is all the material needed, it is not shared on all product/evkit page variants. Other than this small detail, I found all I needed and even an Ibis model to verify and simulate signal integrity.
The price to performance ratio was good: 5/5. The MAX2256 is not a cheap IC. That being said, it rewards you with the fastest data rate and lowest power consumption in the market, so if you need this kind of performance I think the price is fair. The board isn’t cheap either and maybe even a bit overkill, but it’s extremely well built to allow everybody to verify MAX2256 digital isolator performances with minimum external uncertainty. I enjoyed using it.