UPDATE: The ADA4528 with a nominal 2.5uV offsett actually looks like it has a tighter spread than the MAX4239, and unity gain stable. But lower bandwidth.
Rochester electronics are legit. They have deals with all the big manufacturers managing low demand, EOL consignments and aftermarket manufacture to original manufacturer data. Anyone involved in obsolescence management should be familiar with them.
I am happy you have found OPA189. Using it you can get rid of the second stage depending on bandwidth required. Or you can make use of the high dynamic range. Have a look at its offset voltage distribution. Also, INA228-Q1 is interesting for precision current measurement. It provides 2 uV max offset and you get digitized result with I2C right away.
Perhaps TI, Maxim and Analog use different sigma values (e.g. 3, 6, or even 7) to spec those min/max values. Wise idea to look at the graphs. Have you also considered using a different 2nd opamp? If the first stage has a gain of 5 with 0.1uV offset, then effectively the 2nd stage can have an offset voltage of 0.5uV without adding more offset than the 1st stage. To achieve the same gain/BW spec, the 2nd opamp needs to have more GBW, but with 0.5uV offset that is alot easier to find. Maybe the ADA4528 for 5x gain (600kHz) and then a OPA189 at 20x gain (700kHz). Haven't looked at noise though, because usually it's best to have the first stage with most gain for the lowest noise figure.. (however the MAX4239 is quite noisy, so might still be good)
@@nlhans1990 I was told to only trust tabulated values as they would be measured and to test everything outside that. Graphs might be useful to know what to expect but can't be sure what may have been extrapolated.
Hi Dave, I retired as the managing director of product definition for standard products at Maxim a few years ago. My group and I defined thousands of products like this for over 20 years and I had a hand in the 4239 parts as well as many, many others. There were a couple issues that came up in this video that I can shed a bit of light on from the chip company side. Although Philbrick made precsion amps in 50's with tubes, Intersil , the spiritual father of Maxim invented this category in CMOS chopper amps in the 70's with the ICL7652. Maxim's second sourced these parts and made improved second sources early in its life. Here are a few points that may give you a better handle on how chip companies approach high performance analog. Because of the way that these parts work, the nominal offset is zero. What gets in the way of this is real third order effect like thermocouple action in the leads and package stress. Maxim is somewhat unique in that it can and does do post package trim (zener zapping) vs. just laser trim a the wafer level. Parts that are perfect at wafer end up being a couple of uV post package due to stresses. The other thing about very high spec parts like this (100 nV) is testing. In order to get good yields, test max limits are generally set higher, we called this the threshold of pain. The typicial values and histograms are there to give you a feel for what you're really going to get. In difficult to test specs, you're balancing how many seconds you're spending on a million dollar tester- the cost get significant. Some specs llike leakage currents in analog switches will often have a max spec of 10 uA though will generally be in the femto amp range. We would release a different external part number that guaranteed a spec like this. You can get an idea of the real spread by asking a vendor to make you 10,000 (a common minimum) with some spec tightly tested and they don't balk, it means the parts will yield if the customer is willing to pay the delta in test cost (plus NRE).. Maxim routinely does this for big customers with precision requirements like a test equipment company. The histograms in the data sheet is taken from the first few wafer lots as the test guys and the design guys dial things in- this process is called correlation. On a little amplifier parts on big wafers, there be 5000 die per wafer so you can generate a lot of test data pretty quickly. The voltage range/dynamic range issue is driven by modern processes. Maxim has a lot of boutique processes but most analog part in the market are made on kind of vanilla cmos processes which are generally digital and low voltage. These are generally 100-200 micron processes. If you see cmos parts with +-5V or +-15v supplies, they are fabbed on old 2 micron plus processes or a specialty analogy process. Maxim, TI and Analog and a very few others keep these old processes running for precision analog. The economics of doing 130 nm analog on 300 mm wafers makes parts really cheap even though it takes some real design chops to do it. Your comments about automotive are somewhat on target, they can be robust but mainly this designation has to do to with consistency of supply, change notices and paper. Sometimes specs are relaxed to improve yields also. Automotive guys can work with anything but they don't like surprises. Hope this helps a little.
Daniel- fancy processes usually cost more and are on smaller wafers. It all adds up to cost. There are several companies that have the processes and even foundries have them but you don't want to use a cadillac process unless you have to. A 200 mm wafer that could produce 5000 amplifier die per wafer might cost $500 in a cheap CMOS process. Fi you want dielectrically isolated or SOI, you get 150 mm wafers for $1000 that only produce 1250 amps (4x less area). The trick is to make stuff on the biggest, cheapest wafers you can if you want to make money. The engineering that it takes to do all these clever things on bulk CMOS is what seperates the men from the boys. There are times when you need SOI but they are avoided strongly.
In a project we had to measure a massively resistive sensor (>10GOhm) with a very small output (single digits uVolt range), which meant that we had to use the opamps with the lowest current offset available. After testing several of them trying to find a balance between current bias (which also had some effects on the sensor), current offset and voltage offset, we opted for choosing the ones with lowest current metrics and placing a mosfet across the input. Before measuring we would short the inputs and measure the offset. Goes to show that sometimes it’s better to rethink your architecture than to try to find the specs you wanted.
did you test the LMC662 or one of it's relatives? it's got fA level bias and offset currents with decent offset voltage (1mV ballpark) Used it to great effect on a photo-diode amplifier with a 10Gohm feedback resistor. Noise was 30fArms at ~30Hz. Also added a discrete transistor high voltage output stage for added dynamic range and bandwidth.
Here are a couple tips for good amplifier design. 1. It's always best to use opamps as inverting amplifiers if you can. That way you don't get common-mode input voltage. Also, Vos sometimes varies over the common-mode input range which adds distortion. 2. You could improve accuracy and lower cost by moving the second stage inside the feedback loop of the first stage. That way the second stage doesn't need to be precise. You could use a high GBWP second stage with relatively high Vos and cheaper 1% resistors. Those errors will essentially be integrated out by the first stage. This approach is called a composite amplifier.
@@AMalas Depends on the opamps, and how you use them. Typically you would use a precision opamp with relatively low BW for your input stage, and a high-BW low-precision opamp inside the loop. This 2nd opamp would normally be used as a high-gain noninverting amp using low precision feedback resistors. The only high precision resistors you need are for the outer loop, which feeds the output back to the first stage. But if you don't understand classical control theory, you could easily make the circuit unstable. BTW, every analog circuit designer really needs to know classical control theory; it's one of the most powerful tools in your tool belt.
Just did a quick SPICE sim. I think Dave could reduce his BOM cost significantly while improving accuracy by making the following changes: 1. Replace U4 with TSV911AIDBVR. This looks much cheaper than MAX4239AUT in similar qty. 2. Reduce cost of R13-R15 by using 1% or 2% instead of 0.05%. 3. Increase values of R3 and R11 to total 99k instead of 9k. Could use 0.1% instead of 0.05%. 4. Disconnect C3 and R11 from the output of U1 and reconnect them to the output of U4. 5. Tweek compensation caps as needed. My sim shows C4 can stay at 10pF if C3 is about 3.8pF. I haven't breadboarded this, or done a more complete sim using the manufacturer's SPICE models, so I can't guarantee it works, but I'll bet the overall noise and stability are the same. But accuracy should be better.
@@steverobbins4872 I am still in university, we have done two control courses but it felt like I was learning how to analyze one single circuit and nothing else. Any resource you'd trust and recommend to learn more?
When you searched in the mouser catalog you had the "op amp" search key active. That's why you've found so few parts. There's the AD LTC1050 that may suit (Vos 0.1uV typ, 5uV max,2.5MHz, 18V)
Rochester Electronics works with major semiconductor manufacturers to continue to supply their obsolete and end-of-life parts that the original manufacturers don't want to be bothered with. They stock billions of original chips and wafers and package the chips as needed. When the original chips are no longer available, they can make more - for a price - using the original masks or by recreating the design via reverse engineering. They've recently become a Digi-Key Marketplace supplier, but they've been around forever and are headquartered in a lovely resort town not far from me.
Actually, I made it a Friday morning routine once a month to have a look at the 4-5 biggest manufacturers to see what new flavors they have. Cost you maybe an hour or two. I'm actually blown away how little knowledge people have on a professional level what's actually available. Sometimes come with horrible old and close to obsolete parts. I even had whole discussions about certain ideas that people thought it wasn't possible, yet there was just an easy of the shelf part available.
However, the problem is often that in many companies the inventories are full of old components purchased in large quantities and at low prices. Developers are encouraged to use them if possible, which leads to their use even though there are a dozen better ones on the market. This is unfortunately the market economy.
I'm really surprised nothing new has come around. I was actually getting a bit worried that something had, because I had just purchased one of your uCurrents a few weeks ago.
I think the reason you're noticing you're missing parts is because you first search for "opamp" and then click the category - maybe try clearing the "opamp" search criteria. I noticed you did it with both mouser and digikey.
I would not put my attention on a low offset voltage with such a product. This can be calibrated later. The drift is probably more critical here. I would even go so far as to say that another very important criterion for such a product is the isolation (between input and output). I have been working on this topic for quite some time and came to the conclusion that it doesn't even make sense to use an isolation amplifier or high precision optocouplers but to go the way through digitalization (24 Bit ADC). In this way you could include various calibrations which is very difficult or costly to achieve with pure analog technology. Of course the question arises whether one wants to output the values as voltage (DAC) - which would be possible, or whether one can live with the digitized value.
Your time for programming is free? This is by far the main reason not to go for any microcontroller. It takes time to program them. Yes, obviously this can be automated. But that's only interesting when selling enough of them and in the end will cost you still a little more.
P_ Mouse flexibility is the other, I much prefer a digital system, with minimal analog decisions deferred to digital, digital power supply for example, with a fully digital feedback loop, SDR is the other example
5:24 I disagree too. It's fun to find new and improved ICs. I remember Woz telling the story of how he improved the floppy driver on the original Apple by replacing 3-4 old TTLs with a single new chip.
I always loved reading new product announcements and poring over new datasheets and app notes. I learned a lot that way. Now, searching for parts often seems like 99% of the job and is definitely more tedious and less rewarding. Digi-Key's and Mouser's parametric searches, while better than most, are really not that well designed. Now I spend hours looking for parts that need to meet multiple environmental specs and conform to multiple standards. Try finding a waterproof panel-mount USB-C jack or a power cable that meets both UL and CSA standards for use in low-voltage LED lights suitable for wet locations or . . . You get the idea. Not nearly as much fun as reading the first RCA datasheets for 4000-series CMOS or the first datasheets for the Intel 4004, 8008, and 8080, the Motorola 6800, the AMD 2900 series, etc. Those were the days. I miss my bookcases full of trade magazines and data books. And I miss designing and building circuits instead of writing code that does the same thing.
Rochester Electronics is one of those suppliers that tend to buy up parts that are going obsolete, yet are notoriously difficult to deal with. An option to buy from them through digikey would have been a god send 5 or so years back (I mean, "email us a scanned credit card, both sides" after agreeing to direct bank transfer and two months of back and forth of various checks, nda's and so on, what a waste of time that was)
Odd.. I was watching this video and, before I knew it, I was reading this: "Goat meat can be prepared in a variety of ways, including stewing, baking, grilling, barbecuing, canning, and frying; it can be minced, curried, or made into sausage. Because of its low fat content, the meat can toughen at high temperatures if cooked without additional moisture. One of the most popular goats farmed for meat is the South African Boer, introduced into the United States in the early 1990s."
In the industry an engineer can simply contact his distributor and ask him to look for a newer and better part; you just tell him which specs that interests you. In that way I often got parts that you cannot find with catalog distributors like DigiKey or parts that are so new that they are not officially available yet.
5:16 I also disagree. Finding and deciding the parts is imho one of the most fun things about desiging. I can spend hours using parameteric search. But I may be biased, just got my boards in on my first design in over 20 years. I forgot how much fun and rewarding the whole process can be.
At 46:08 you missed the "general purpose" (non-automotive) opamp section on Ablic's website. There are another dozen or so parts there, but none look relevant to your search.
ADA4522 is simply where the state of the art is :) 2.7Mhz bandwidth, 22nV/C drift, 100nV noise, this is what we use for LTZ1000 buffers or scaling; ADA4523 and ADA4528 would do too, one is lower voltage, and both uses twice as much power than ADA4522; heating might be a problem in a precision circuit like this input offset is not important, you can cal it out, or zero it out externally; more important is the offset drift with temperature I prefer lower bandwidth to noise, MAX4239 is quite noisy
The AEC-Q 100, 101, 200 standards describe, how electronic components have to be qualified /tested by the manufacturers before they are allowed to sell their parts to the Automotive Electronics industry, like Continental, Bosch, Valeo, Visteon, and so on. The tests include environmental, reliability and electrical tests. These are the de facto automotive industry standards, i.e. very important. Go to www.aecouncil.com/ for details. Specially for Automotive, manufacturers specify and test for -40.. +85°C for Interior, and -40 .. +125°C for Power Train, Body Control and Safety applications, which you found in all of the chopper specifications. Use of the standard Industrial range of 0..+70°C is mostly not allowed in Automotive.
I have always wondered if its maybe a good idea to leave some pads on the µC for those that want to add a few parts to get rid of the least bit of offset...
I love searching for parts on sites like digikey eventhough their parametric search is a bit crap just like every other site. Even if i dont need a part i enjoy looking for fun parts i could use or just looking at them knowing i cannot afford it.
With regards to the distributions presented in the datasheets I have two questions: 1) is it plausible that some manufacturers would sacrifice yield to allow them to cut off the lower and upper tolerance intervals, allowing them to present tighter tolerances than their competitors? Of course, this would imply testing and scrapping during manufacturering, but I imagine that this added cost would then simply be added to the sales cost of the device. 2) Does the distributions take batch-to-batch variations into account, or is it possible that the tested devices are all from the same batch? This would certainly yield different results and would look bad in the datasheet. Has anyone experienced these kinds of creative statistics? Any particular brands?
I love to have a manufacturer's or distributor's parametric search open in one window, and LTSpice (or whatever simulator) open in another window. Meaning I only shop for opamps with models available, which is a great filter for avoiding the junk.
wow 2014, i have used a microcurrent countless times since then, all the gold is basically rubbed off i recently needed current amplification in a project and i used an AD8554 with 2 sets of 2 series amplifiers, it worked wonderfully but was a bit overkill, could cal out in software
Dave, how do you know tht the max4238/9 are indeed chopper amps? Well I am from a time where the µA725 (Fairchild) was the top of the top. Sure there are much better choices today, but even back then there were laser trimmed dual chip solutions available. Also self calibrating op amps I think I have heard of. Now chopper amps are a different category alltogether. Just wondering how you conclude its a chopper amp? Does the spec sheet specify a copper frequency?
It's a patanted "auto-zero" design. Usually they mix the two. Yes, the switching frequency is pseudorandom from 10kHz to 15kHz, and you can see this in the response.
Concerning the bandwidth of an non inverting opamp amplifier, WHY the bandwith decreases with increasing gain? There is any reason that's not the slew rate limitation on its output, or it's the only reason?
Nope the slew rate is only a property of the last stage. Op amps have multiple stages. To ensure a constant gain bandwidth product they need to be compensated as every stage inserts a pole in the overall response. General use opamps have compensation caps on the chip. High bandwidth ones can be compensated externally. Now the reason for the forced 20dB per decade for the general purpose amps falloff is stability. In the Nyquist plot of any four-pole stability is guaranteed by avoiding the -1,j0 point. There are special applications that get more bandwidth by getting the plot closer to instability by placing the poles of each stage at different frequencies. But then you no longer have the maximal flat response and depending on the location of these poles you get a different phase behavior which may lead to ringing on transients. By means of L2 optimization you can find compensations that fulfill your needs (e.g. Bessel, Butterworth, Cauer etc. Responses). Way back in the 80s this was part of my work.
49:09 "What's the collective term for Diodes"? Everybody: "FULL BRIDGE RECTIFIER" LOL For people that don't do fun - Yea I know they are for protection.
I don't think you've done a video on overload recovery time :) You should make a list of topics that you think you could do a video on and cross out the ones you've already done :)
On the first schematic, what exactly R12 (in series with the op amp input) and R8 (in series with the output) do? Is the first one for overvoltage protection, and second one for stability?
I’m ignorant of the importance of bandwidth. Can anyone explain why a higher gain bandwidth product like 6MHz is important and how it makes a difference compared to a lower one like 1MHz, for this application? Thanks
With the high voltage versions that still have a low offset voltage, but a far higher dynamic range, I know for your microcurrent project that's 3V powered it wouldn't be good, but if you had the same project with 36V power, wouldn't that still be better by allowing you to use lets say the 20V range on your DMM, and then you'd be well above the least significant digit on it?
Anyone have recommendations for high speed (slew rate) current sense or instrumentation amplifiers with very low input offset? I'm looking to measure lowside phase currents with a minimum sampling window of 1.5uS. I was originally using the INA240A1, but its 2V/uS slew rate was too slow for the high duty cycles required by my application. According to the graph and my measurements, it wouldn't reach 90% of the target value until just after 1uS. I've been looking at the MCP6N11-010 which is less accurate, has much higher input offset and offset drift, and a greatly reduced negative common mode voltage, but has a very fast settling time in comparison to the INA240, additionally it requires me to externally drive the reference voltage with a dedicated op amp due to the low feedback resistor values that the datasheet recommends. I need something that's rated down to -1V common mode operational because I need to be able to measure at least 150A, and parasitic resistances can very easily pull the common mode below -300mV. I can't do inline phase sensing because I need to support up to 100V operation. Isolated sensing topologies will price my product out of competition, and even then, I implement a signal injection scheme to resolve motor position at zero speed with no sensors. This requires extremely accurate measurements and a high fidelity data acquisition stage. Don't ask, my customers basically want a motor controller with the power density of a nuclear reactor that comes with a set of black magic features at chinese level pricing. .__.
Disagree with your statements about guaranteed by design vs guaranteed by testing. Guaranteed by design means assuming process tolerances are within spec then the specification is guaranteed based on physical principles. They can then easily test the process tolerances, or similarly guarantee them. Guarantee by testing just means that they never found a bad part.
Rule 1: Vendors want your money Rule 2: Vendors will lie, see rule 1. Rule 3: Vendors are not your friends, see rule 2. Rule 4: Vendors are inescapable
EEVblog2 is a general interest channel and not appropriate for this type of content. I don't see this as boring, rather it is a very informative video that serves as a lesson to those that aren't familiar with how to spec out and find parts, and as an insight into how another designer will go about the part spec determination and searching process. I've been guilty in many cases of locking myself into a particular manufacturer or supplier as a matter of preference and/or bias toward that manufacturer (like TI). The EEVblog channel has (I assume) helped many to become better engineers (and, I'm sure, helped many to become engineers) with this type of video.
@@EEVblog your content has matured and refined my electrical knowledge. What you teach cannot be learned in university. It is the real world experience, and how circuits actually work. Don't listen to the haters. Also have an old analog Avometer 3kv AC/DC, fully functional. Will send it if you do not have one. With original manual and leather case. Reply if you want it.
UPDATE: The ADA4528 with a nominal 2.5uV offsett actually looks like it has a tighter spread than the MAX4239, and unity gain stable. But lower bandwidth.
Rochester electronics are legit. They have deals with all the big manufacturers managing low demand, EOL consignments and aftermarket manufacture to original manufacturer data. Anyone involved in obsolescence management should be familiar with them.
I am happy you have found OPA189. Using it you can get rid of the second stage depending on bandwidth required. Or you can make use of the high dynamic range. Have a look at its offset voltage distribution.
Also, INA228-Q1 is interesting for precision current measurement. It provides 2 uV max offset and you get digitized result with I2C right away.
Perhaps TI, Maxim and Analog use different sigma values (e.g. 3, 6, or even 7) to spec those min/max values. Wise idea to look at the graphs.
Have you also considered using a different 2nd opamp? If the first stage has a gain of 5 with 0.1uV offset, then effectively the 2nd stage can have an offset voltage of 0.5uV without adding more offset than the 1st stage. To achieve the same gain/BW spec, the 2nd opamp needs to have more GBW, but with 0.5uV offset that is alot easier to find. Maybe the ADA4528 for 5x gain (600kHz) and then a OPA189 at 20x gain (700kHz). Haven't looked at noise though, because usually it's best to have the first stage with most gain for the lowest noise figure.. (however the MAX4239 is quite noisy, so might still be good)
@@nlhans1990 I was told to only trust tabulated values as they would be measured and to test everything outside that. Graphs might be useful to know what to expect but can't be sure what may have been extrapolated.
Texas Instruments TLC2652AC has the tightest spec i could find. 1uV max. at 25C lower GBW and higher typical value than the MAX tho
Hi Dave, I retired as the managing director of product definition for standard products at Maxim a few years ago. My group and I defined thousands of products like this for over 20 years and I had a hand in the 4239 parts as well as many, many others. There were a couple issues that came up in this video that I can shed a bit of light on from the chip company side. Although Philbrick made precsion amps in 50's with tubes, Intersil , the spiritual father of Maxim invented this category in CMOS chopper amps in the 70's with the ICL7652. Maxim's second sourced these parts and made improved second sources early in its life. Here are a few points that may give you a better handle on how chip companies approach high performance analog. Because of the way that these parts work, the nominal offset is zero. What gets in the way of this is real third order effect like thermocouple action in the leads and package stress. Maxim is somewhat unique in that it can and does do post package trim (zener zapping) vs. just laser trim a the wafer level. Parts that are perfect at wafer end up being a couple of uV post package due to stresses. The other thing about very high spec parts like this (100 nV) is testing. In order to get good yields, test max limits are generally set higher, we called this the threshold of pain. The typicial values and histograms are there to give you a feel for what you're really going to get. In difficult to test specs, you're balancing how many seconds you're spending on a million dollar tester- the cost get significant. Some specs llike leakage currents in analog switches will often have a max spec of 10 uA though will generally be in the femto amp range. We would release a different external part number that guaranteed a spec like this. You can get an idea of the real spread by asking a vendor to make you 10,000 (a common minimum) with some spec tightly tested and they don't balk, it means the parts will yield if the customer is willing to pay the delta in test cost (plus NRE).. Maxim routinely does this for big customers with precision requirements like a test equipment company. The histograms in the data sheet is taken from the first few wafer lots as the test guys and the design guys dial things in- this process is called correlation. On a little amplifier parts on big wafers, there be 5000 die per wafer so you can generate a lot of test data pretty quickly. The voltage range/dynamic range issue is driven by modern processes. Maxim has a lot of boutique processes but most analog part in the market are made on kind of vanilla cmos processes which are generally digital and low voltage. These are generally 100-200 micron processes. If you see cmos parts with +-5V or +-15v supplies, they are fabbed on old 2 micron plus processes or a specialty analogy process. Maxim, TI and Analog and a very few others keep these old processes running for precision analog. The economics of doing 130 nm analog on 300 mm wafers makes parts really cheap even though it takes some real design chops to do it. Your comments about automotive are somewhat on target, they can be robust but mainly this designation has to do to with consistency of supply, change notices and paper. Sometimes specs are relaxed to improve yields also. Automotive guys can work with anything but they don't like surprises. Hope this helps a little.
This is truly fascinating!
Daniel- fancy processes usually cost more and are on smaller wafers. It all adds up to cost. There are several companies that have the processes and even foundries have them but you don't want to use a cadillac process unless you have to. A 200 mm wafer that could produce 5000 amplifier die per wafer might cost $500 in a cheap CMOS process. Fi you want dielectrically isolated or SOI, you get 150 mm wafers for $1000 that only produce 1250 amps (4x less area). The trick is to make stuff on the biggest, cheapest wafers you can if you want to make money. The engineering that it takes to do all these clever things on bulk CMOS is what seperates the men from the boys. There are times when you need SOI but they are avoided strongly.
Really interesting knowledge here. Hope you could write more blog posts like this somewhere.
This is really useful for us youngster.
In a project we had to measure a massively resistive sensor (>10GOhm) with a very small output (single digits uVolt range), which meant that we had to use the opamps with the lowest current offset available.
After testing several of them trying to find a balance between current bias (which also had some effects on the sensor), current offset and voltage offset, we opted for choosing the ones with lowest current metrics and placing a mosfet across the input. Before measuring we would short the inputs and measure the offset.
Goes to show that sometimes it’s better to rethink your architecture than to try to find the specs you wanted.
did you test the LMC662 or one of it's relatives? it's got fA level bias and offset currents with decent offset voltage (1mV ballpark) Used it to great effect on a photo-diode amplifier with a 10Gohm feedback resistor. Noise was 30fArms at ~30Hz. Also added a discrete transistor high voltage output stage for added dynamic range and bandwidth.
Here are a couple tips for good amplifier design.
1. It's always best to use opamps as inverting amplifiers if you can. That way you don't get common-mode input voltage. Also, Vos sometimes varies over the common-mode input range which adds distortion.
2. You could improve accuracy and lower cost by moving the second stage inside the feedback loop of the first stage. That way the second stage doesn't need to be precise. You could use a high GBWP second stage with relatively high Vos and cheaper 1% resistors. Those errors will essentially be integrated out by the first stage. This approach is called a composite amplifier.
You can do number 2 while maintaining the bandwidth?
@@AMalas Depends on the opamps, and how you use them. Typically you would use a precision opamp with relatively low BW for your input stage, and a high-BW low-precision opamp inside the loop. This 2nd opamp would normally be used as a high-gain noninverting amp using low precision feedback resistors. The only high precision resistors you need are for the outer loop, which feeds the output back to the first stage.
But if you don't understand classical control theory, you could easily make the circuit unstable. BTW, every analog circuit designer really needs to know classical control theory; it's one of the most powerful tools in your tool belt.
Just did a quick SPICE sim. I think Dave could reduce his BOM cost significantly while improving accuracy by making the following changes:
1. Replace U4 with TSV911AIDBVR. This looks much cheaper than MAX4239AUT in similar qty.
2. Reduce cost of R13-R15 by using 1% or 2% instead of 0.05%.
3. Increase values of R3 and R11 to total 99k instead of 9k. Could use 0.1% instead of 0.05%.
4. Disconnect C3 and R11 from the output of U1 and reconnect them to the output of U4.
5. Tweek compensation caps as needed. My sim shows C4 can stay at 10pF if C3 is about 3.8pF.
I haven't breadboarded this, or done a more complete sim using the manufacturer's SPICE models, so I can't guarantee it works, but I'll bet the overall noise and stability are the same. But accuracy should be better.
@@steverobbins4872 I am still in university, we have done two control courses but it felt like I was learning how to analyze one single circuit and nothing else. Any resource you'd trust and recommend to learn more?
@@steverobbins4872 I'd love to see Daves response to that comment. Really didactic, thanks!
5:42 Dave perfectly explains why I never finish my projects... I agree 105% :D
My worst words:
"Whoa, I got an idea!"
Explains my unfinished projects...
Join the club ;)
I have so many projects on my todo list im afraid i dont have enough time in my life to do them all, even if i stopped adding to it. :(
I am in the seemingly never ending quest to finishing my ultrasound to human audible converter from 20KHz to 150KHz... i can't decide on the mike...
@@soulrobotics That sound really cool, I want to do something similar to detect rf and whatnot from outer space.
7c3c72602f7054696b ....hahaha YES, it used to be called Radio...
When you searched in the mouser catalog you had the "op amp" search key active. That's why you've found so few parts. There's the AD LTC1050 that may suit (Vos 0.1uV typ, 5uV max,2.5MHz, 18V)
Exactly right, I always make sure to clear that.
He did the same for Digikey
Rochester Electronics works with major semiconductor manufacturers to continue to supply their obsolete and end-of-life parts that the original manufacturers don't want to be bothered with. They stock billions of original chips and wafers and package the chips as needed. When the original chips are no longer available, they can make more - for a price - using the original masks or by recreating the design via reverse engineering. They've recently become a Digi-Key Marketplace supplier, but they've been around forever and are headquartered in a lovely resort town not far from me.
Rochester Electronics is a US based firm and have been around for a long time. They tend to specialise in buying up obsolete stock and EOL products.
They bought up a ton of old Maxim stock recently
EXACTLY!!! I get sucked into the rabbit hole every time I go look for parts too, lol
Same here. I have to force myself to stay focused to get the project spec’d out.
The MAX4288 with 0.1µV is a database entry issue. According to datasheet, Input offset voltage is Typ 0.1 MILLI volt, not micro.
Kamikaze
Actually, I made it a Friday morning routine once a month to have a look at the 4-5 biggest manufacturers to see what new flavors they have. Cost you maybe an hour or two.
I'm actually blown away how little knowledge people have on a professional level what's actually available. Sometimes come with horrible old and close to obsolete parts.
I even had whole discussions about certain ideas that people thought it wasn't possible, yet there was just an easy of the shelf part available.
However, the problem is often that in many companies the inventories are full of old components purchased in large quantities and at low prices. Developers are encouraged to use them if possible, which leads to their use even though there are a dozen better ones on the market. This is unfortunately the market economy.
@EEBblog I've been waiting for a video like this for a long time, thank you so much!
I'm really surprised nothing new has come around. I was actually getting a bit worried that something had, because I had just purchased one of your uCurrents a few weeks ago.
Bring back fundamental Fridays, we watch more of your old content than the new stuff.
"Generally temperature wouldn't change that much"
Cries in MIL-STDs
I have dreams about 38535
I think the reason you're noticing you're missing parts is because you first search for "opamp" and then click the category - maybe try clearing the "opamp" search criteria. I noticed you did it with both mouser and digikey.
Thank you for how to seek the specs and how to find a information from datasheet's various graphs, numbers, lines etc! 😎👍
I would not put my attention on a low offset voltage with such a product. This can be calibrated later. The drift is probably more critical here. I would even go so far as to say that another very important criterion for such a product is the isolation (between input and output). I have been working on this topic for quite some time and came to the conclusion that it doesn't even make sense to use an isolation amplifier or high precision optocouplers but to go the way through digitalization (24 Bit ADC). In this way you could include various calibrations which is very difficult or costly to achieve with pure analog technology. Of course the question arises whether one wants to output the values as voltage (DAC) - which would be possible, or whether one can live with the digitized value.
Sure. But the entire point was to avoid calibration.
We don't wanna do any " pot tweaking "
@@EEVblog we can't have it all, it's either bandwidth, offset, noise / stability,
Your time for programming is free? This is by far the main reason not to go for any microcontroller. It takes time to program them.
Yes, obviously this can be automated. But that's only interesting when selling enough of them and in the end will cost you still a little more.
P_ Mouse flexibility is the other, I much prefer a digital system, with minimal analog decisions deferred to digital, digital power supply for example, with a fully digital feedback loop, SDR is the other example
Nice search 👍
Thanks for sharing 👍😀
Who the heck thinks this is _boring?_ I agree with you here - looking for parts is the most fun part of design.
Agree, browsing data sheets is fantastic. Then dreaming up applications for special interest devices, consider limitations etc
5:24 I disagree too. It's fun to find new and improved ICs. I remember Woz telling the story of how he improved the floppy driver on the original Apple by replacing 3-4 old TTLs with a single new chip.
I always loved reading new product announcements and poring over new datasheets and app notes. I learned a lot that way. Now, searching for parts often seems like 99% of the job and is definitely more tedious and less rewarding. Digi-Key's and Mouser's parametric searches, while better than most, are really not that well designed. Now I spend hours looking for parts that need to meet multiple environmental specs and conform to multiple standards. Try finding a waterproof panel-mount USB-C jack or a power cable that meets both UL and CSA standards for use in low-voltage LED lights suitable for wet locations or . . . You get the idea. Not nearly as much fun as reading the first RCA datasheets for 4000-series CMOS or the first datasheets for the Intel 4004, 8008, and 8080, the Motorola 6800, the AMD 2900 series, etc. Those were the days. I miss my bookcases full of trade magazines and data books. And I miss designing and building circuits instead of writing code that does the same thing.
Rochester Electronics is one of those suppliers that tend to buy up parts that are going obsolete, yet are notoriously difficult to deal with. An option to buy from them through digikey would have been a god send 5 or so years back (I mean, "email us a scanned credit card, both sides" after agreeing to direct bank transfer and two months of back and forth of various checks, nda's and so on, what a waste of time that was)
Odd.. I was watching this video and, before I knew it, I was reading this: "Goat meat can be prepared in a variety of ways, including stewing, baking, grilling, barbecuing, canning, and frying; it can be minced, curried, or made into sausage. Because of its low fat content, the meat can toughen at high temperatures if cooked without additional moisture. One of the most popular goats farmed for meat is the South African Boer, introduced into the United States in the early 1990s."
"Not as tight as a nun's nasty"
-Dave 2020
... but reasonably tight. - I almost choked on coffee.
In the industry an engineer can simply contact his distributor and ask him to look for a newer and better part; you just tell him which specs that interests you.
In that way I often got parts that you cannot find with catalog distributors like DigiKey or parts that are so new that they are not officially available yet.
5:16 I also disagree. Finding and deciding the parts is imho one of the most fun things about desiging. I can spend hours using parameteric search. But I may be biased, just got my boards in on my first design in over 20 years. I forgot how much fun and rewarding the whole process can be.
At 46:08 you missed the "general purpose" (non-automotive) opamp section on Ablic's website. There are another dozen or so parts there, but none look relevant to your search.
Yes, he was fumbling around but did not see the most relevant category in the middle of the screen
ADA4522 is simply where the state of the art is :) 2.7Mhz bandwidth, 22nV/C drift, 100nV noise, this is what we use for LTZ1000 buffers or scaling; ADA4523 and ADA4528 would do too, one is lower voltage, and both uses twice as much power than ADA4522; heating might be a problem in a precision circuit like this
input offset is not important, you can cal it out, or zero it out externally; more important is the offset drift with temperature
I prefer lower bandwidth to noise, MAX4239 is quite noisy
The AEC-Q 100, 101, 200 standards describe, how electronic components have to be qualified /tested by the manufacturers before they are allowed to sell their parts to the Automotive Electronics industry, like Continental, Bosch, Valeo, Visteon, and so on. The tests include environmental, reliability and electrical tests. These are the de facto automotive industry standards, i.e. very important. Go to www.aecouncil.com/ for details.
Specially for Automotive, manufacturers specify and test for -40.. +85°C for Interior, and -40 .. +125°C for Power Train, Body Control and Safety applications, which you found in all of the chopper specifications.
Use of the standard Industrial range of 0..+70°C is mostly not allowed in Automotive.
I have always wondered if its maybe a good idea to leave some pads on the µC for those that want to add a few parts to get rid of the least bit of offset...
I love searching for parts on sites like digikey eventhough their parametric search is a bit crap just like every other site. Even if i dont need a part i enjoy looking for fun parts i could use or just looking at them knowing i cannot afford it.
I got bit by the input diodes when I was wiring up an op amp as a compactor for a product tester.
With regards to the distributions presented in the datasheets I have two questions: 1) is it plausible that some manufacturers would sacrifice yield to allow them to cut off the lower and upper tolerance intervals, allowing them to present tighter tolerances than their competitors? Of course, this would imply testing and scrapping during manufacturering, but I imagine that this added cost would then simply be added to the sales cost of the device.
2) Does the distributions take batch-to-batch variations into account, or is it possible that the tested devices are all from the same batch? This would certainly yield different results and would look bad in the datasheet.
Has anyone experienced these kinds of creative statistics? Any particular brands?
Wow, that description of Dave as he goes on Wikipedia... That so much resembles my own experiences 😂😀
I love to have a manufacturer's or distributor's parametric search open in one window, and LTSpice (or whatever simulator) open in another window. Meaning I only shop for opamps with models available, which is a great filter for avoiding the junk.
I've always wondered why you use two op-amps with X10 gain each when one gain stage would do the same thing?
No comparison of input bias/offset current?
Nice
The "233 typical units" on the binning graph for the LTC2063 comes from the fact that many JEDEC reliability tests are done with 77pcs/lot, 3 lots.
wow 2014, i have used a microcurrent countless times since then, all the gold is basically rubbed off
i recently needed current amplification in a project and i used an AD8554 with 2 sets of 2 series amplifiers, it worked wonderfully but was a bit overkill, could cal out in software
Searching for ICs is just like searching for ice cream in the beaches
In my day we used CA3140E and CA3240E for everything.
I'm still using them in my projects today, yea thumb the comment down haters!
Nice video
Dave, how do you know tht the max4238/9 are indeed chopper amps? Well I am from a time where the µA725 (Fairchild) was the top of the top. Sure there are much better choices today, but even back then there were laser trimmed dual chip solutions available. Also self calibrating op amps I think I have heard of. Now chopper amps are a different category alltogether. Just wondering how you conclude its a chopper amp? Does the spec sheet specify a copper frequency?
It's a patanted "auto-zero" design. Usually they mix the two. Yes, the switching frequency is pseudorandom
from 10kHz to 15kHz, and you can see this in the response.
Concerning the bandwidth of an non inverting opamp amplifier, WHY the bandwith decreases with increasing gain? There is any reason that's not the slew rate limitation on its output, or it's the only reason?
Nope the slew rate is only a property of the last stage. Op amps have multiple stages. To ensure a constant gain bandwidth product they need to be compensated as every stage inserts a pole in the overall response. General use opamps have compensation caps on the chip. High bandwidth ones can be compensated externally. Now the reason for the forced 20dB per decade for the general purpose amps falloff is stability. In the Nyquist plot of any four-pole stability is guaranteed by avoiding the -1,j0 point. There are special applications that get more bandwidth by getting the plot closer to instability by placing the poles of each stage at different frequencies. But then you no longer have the maximal flat response and depending on the location of these poles you get a different phase behavior which may lead to ringing on transients. By means of L2 optimization you can find compensations that fulfill your needs (e.g. Bessel, Butterworth, Cauer etc. Responses). Way back in the 80s this was part of my work.
49:09 "What's the collective term for Diodes"?
Everybody: "FULL BRIDGE RECTIFIER"
LOL
For people that don't do fun - Yea I know they are for protection.
I don't think you've done a video on overload recovery time :)
You should make a list of topics that you think you could do a video on and cross out the ones you've already done :)
On the first schematic, what exactly R12 (in series with the op amp input) and R8 (in series with the output) do? Is the first one for overvoltage protection, and second one for stability?
R8 is an output short circuit protection. I think R12 is to help input clamp diodes in case of input overvoltage.
I’m ignorant of the importance of bandwidth. Can anyone explain why a higher gain bandwidth product like 6MHz is important and how it makes a difference compared to a lower one like 1MHz, for this application? Thanks
Wonderful ! T Y Dave
why not use "hand-picked" op-amps with "0 µV" offset in your product?
With the high voltage versions that still have a low offset voltage, but a far higher dynamic range, I know for your microcurrent project that's 3V powered it wouldn't be good, but if you had the same project with 36V power, wouldn't that still be better by allowing you to use lets say the 20V range on your DMM, and then you'd be well above the least significant digit on it?
How could they claim that this TLC2652 is 1uV input offset max? From the graph it is clearly 2uV input offset max.
@49:08 Mob of kangaroos. Leap of leopards. Maybe a rectification of diodes?
49:30 17 diodes
Anyone have recommendations for high speed (slew rate) current sense or instrumentation amplifiers with very low input offset? I'm looking to measure lowside phase currents with a minimum sampling window of 1.5uS. I was originally using the INA240A1, but its 2V/uS slew rate was too slow for the high duty cycles required by my application. According to the graph and my measurements, it wouldn't reach 90% of the target value until just after 1uS. I've been looking at the MCP6N11-010 which is less accurate, has much higher input offset and offset drift, and a greatly reduced negative common mode voltage, but has a very fast settling time in comparison to the INA240, additionally it requires me to externally drive the reference voltage with a dedicated op amp due to the low feedback resistor values that the datasheet recommends. I need something that's rated down to -1V common mode operational because I need to be able to measure at least 150A, and parasitic resistances can very easily pull the common mode below -300mV.
I can't do inline phase sensing because I need to support up to 100V operation. Isolated sensing topologies will price my product out of competition, and even then, I implement a signal injection scheme to resolve motor position at zero speed with no sensors. This requires extremely accurate measurements and a high fidelity data acquisition stage.
Don't ask, my customers basically want a motor controller with the power density of a nuclear reactor that comes with a set of black magic features at chinese level pricing. .__.
Wait, Dave has humble opinions?
SW1 ! The current is running through that switch! What if contact resistance isn´t reliable and stable?
Uh oh...
When your employer hands you a requirements spec as follows:
Bandwidth: All of it
Temperature range: Dark side of the moon to surface of sun
Gain: Yes
I wonder if you can null it out using a DAC during startup?
Wow, Quantum theory to Goat farming in 0.01uS
Tight as a nun's nasty. That's a new phrase.
digikey's parametric search is mediocre imo. i like mouser's better. precision op amps are separated on mouser though
Disagree with your statements about guaranteed by design vs guaranteed by testing. Guaranteed by design means assuming process tolerances are within spec then the specification is guaranteed based on physical principles. They can then easily test the process tolerances, or similarly guarantee them. Guarantee by testing just means that they never found a bad part.
Maxim used a 6 sigma spec for GBD specs which is on the same order as test escapes- about 3 ppm
Quantum goat farming!
👍
741 :)
"Nun's nasty" :P
Choppa Oppa
Your channel still alive?
Rule 1: Vendors want your money
Rule 2: Vendors will lie, see rule 1.
Rule 3: Vendors are not your friends, see rule 2.
Rule 4: Vendors are inescapable
Why does Dave hate nuns ? Catholic school upbringing perhaps ?! 🤣
You should upload boring videos like this to your EEVblog2 channel.
I disagree. He is an electrical engineer, this fits the main channel perfectly.
Love your videos Dave, keep it up.
Boring ? really ?...I don't think so, this is one on one design gold.
Nope, this is the kind of content I've always done on the main channel.
EEVblog2 is a general interest channel and not appropriate for this type of content.
I don't see this as boring, rather it is a very informative video that serves as a lesson to those that aren't familiar with how to spec out and find parts, and as an insight into how another designer will go about the part spec determination and searching process. I've been guilty in many cases of locking myself into a particular manufacturer or supplier as a matter of preference and/or bias toward that manufacturer (like TI).
The EEVblog channel has (I assume) helped many to become better engineers (and, I'm sure, helped many to become engineers) with this type of video.
@@EEVblog your content has matured and refined my electrical knowledge. What you teach cannot be learned in university. It is the real world experience, and how circuits actually work.
Don't listen to the haters.
Also have an old analog Avometer 3kv AC/DC, fully functional. Will send it if you do not have one. With original manual and leather case. Reply if you want it.
MAX9922??? Was that 10uV at 100x gain???? I think you may have missed that rather important detail because if so that would make it 0.1uV
The same for bandwidth!