EEVblog #1032 Part 5 - John Kenny Keysight Interview

I'd love to know the design phases of this thing you worked on this yeah yep well you designed the front panel you work intimately on managing did you manage the development no no no okay so I'm a technology manager one of the things I do is I work with teams to create common building blocks that get shoes between all the projects I've got and I'm often brought in to help them go over a particular difficult technical challenge so for example on the high end function generator they had some horrendous problems with the switching power supplies to power up all the high power Asics creating Spurs yeah yeah into the output of the function because the 120 mega oh yeah it doesn't take much dbid noise output and I brought in some power supply people to help them redesign all the power supplies to reduce the noise because our power supply group is really good at low noise switching they were not good at low noise switching but they had tons of power requirements cuz they're giant a giant fpg I should say and some really high current DAX and eight of these they're not DVDs but you know just high power circuits so the the switching circuits they put in were like larger than they had ever designed before and they're not experts on switching pallets the folks in New Jersey are so I got involved and I got another guy involved and we redesigned their entire power subsystem for them so they could stay on track on the project there's when they hit a wall rather and say you guys better fix it and work 12-hour days for the next three months no that's not how we're gonna do it we're gonna bring in people who know how to do it and get it right quickly so you don't delay the project three months and don't burn out we get it done right and get it done once so I'll do that I'll the folks who developed RSM use didn't know how to do this kind of graphics interface so I had the the team that developed the the firmware for this and our hardware design for this and we transferred that stuff to our Japanese team that did the SMU so the mountain the Mount Fuji project was our new SMU to compete with her folks in Cleveland and they leveraged all that stuff from the other designs they use the same same circuit designs they took the schematics and transfer them and then they built it up from there that's what I do is I help make those connections so the development process you need a replacement mm-hmm for the buddies at 3400 401 34401a Juke um up was it what shipped was that the original concept how many ever little wasn't a evolution did you go through did you go like we're gonna have a graph Alou screen it's gonna have the it's gonna have the graphene and it's gonna have this and how I already had the function generator at that point oh my captioner was pronounced the first major product at the counter and the function ever developed more than us in parallel so we had the basic structure but it didn't meet the cost target so that's when we had hi and crush it and reduce the cost dramatically and that's where I got heavily involved but the DMM technology believer not was actually taking the we had done a research project on replacing the 3458 which we ended up putting aside and that team stopped working on that and then we had them go and do him dramatic cost reduction to fit it in these kind of price point interesting and you strip it down and simplify it and then the team you know in Loveland Colorado site came up with the devote meter to make it meet the cost targets I came up with a new low-cost front panel and we merged the whole thing together but this architecture is dramatically different than the 34401a s the 3441 that the fundamental problem with the volt meter is that it's isolated and the 34401a the entire processing core next to the measurement engine yeah and it directly controls the measurement engine and the entire front panel is floating right next to at the main potential that worked back in the day of GPIB and rs-232 because we had isolate GPIB and rs-232 over very simple you arts as a chip called the mouth protocol which basically takes and isolates that and brings it across isolation into a custom ASIC and that's great for GPIB and rs-232 by the USB and LAN it can't be done yep you need processing right next to the LAN and USB because it's all grounded you can't isolate must be in land very easily at the high speeds so you change the architecture entirely to put the processing grant which to control the screen and did all the processing right at the back end so to speak that's correct that that we're all the ground potential yep so the front panel display also because this is would be floating that was a problem isolating this what's safe with a flat panel display is a little bit more challenging than with a vacuum fluorescent display ah a little bit easier to keep the cracks from giving you you know shock hazard so yeah this also the LCD display has to be directly connected to the processor yes all right 24 lines I mean it's a very wide bus so all that had to get moved to ground so they had to develop a whole new protocol to a much smaller set of hardware that controlled the actual measurement engine yes they developed a whole new system to do that for this project and that's what enabled these products to do what they did using the same processor platform for the most part that's in the counter the function generator and our power products as well karna sowwy they ended up using FPGA for the measurement there's an EPG and a single chip and a single what's the single chip the single chip does the communications across isolation yep does the slower state machine if then the FPGA does the high-speed processing of the delta-sigma A to D so it does the big fir filter that basically takes this this A to D converter no doing high-speed sampling the multiple bit and then turning that into a slow very high-resolution reading so cost was paramount was it the entire driving factor behind this said well there were two one would be complete compatibility with the existing product because it was so ubiquitous are divided into intestine by Mississippi largest selling product in tester matters your company is like a stream probably in any other tested measurement test images to we're the big but the 34401a tits peak sold over 25,000 yeah we're not back up to that with this but we head in there with some of the disk competitors have come in and taken some way but and it's growing this is just the rolling quite nicely but uh so there was compatibility and there was cost seemed to be the - absolutely driving factors nothing yeah well and also differentiation we have a have nuclei abilities we have do the new graphic system yeah yeah we've added data logging we did histograms things like that which you know today they're very useful exactly but the point is you got a graphic explain damnit use it yeah you know yeah and that's it's not just a different way to control the thing because the menuing system on the 34 for one is not as easy to use nice but it's hard to justify switching all that just for that right so you might get more data and there's more things we'd like to do in the display but you know and they did some newer things I think with the 65 and 70 if not mistaken they have dual parameter information yes yes yes I do there is a Judas light right that wasn't in the first one we didn't have enough time to fit that in right they wanted to but when we got to 65 and 70 day today and frankly with a 34 for 10 and 11 they had the dual parameter in a vacuum fluorescent and when we first introduced the the more basic models we didn't have it and we got feedback that was a problem they liked the dual display all right so we fixed it and they rolled it up into 65 and 70 the biggest challenge in this was when they developed the 34401a early they were on the second floor of a building and the manufacturing center was on the floor good so it was very easy to optimize cost and learn what how you're gonna cost us with the rules for engagement work to optimize in the right way now this thing is manufactured in SEM not even interpreting operation Oh like I had our the supply chain of the Transformers made in Kuantan Malaysia right okay and back then we bought it from a local vendor all right and therefore it optimizing it the perfect shield India at the common mode noise all those things was much easier you've got in your car you drove to their site in an hour you have meaning you get the thing then you gotta move the shield this way you know no problem he's now it's a third party through our penetration back to the States and yeah and getting the cost of it and getting the performance of it is much more difficult so one of the other things I do is I help I go to Penang fairly frequently and they had an issue with the vendor we were using they weren't meaning the cost I actually traveled to China to visit another new company too because I do a lot of magnetics materials work for people to power supply background and I bet it visited the news cup this new company and vetted them out and made sure they could do a quality job and and help communicate our technical needs and they're now meeting cost on that much better than they were before is the reason you went to a contract manufacturer cm for though you were using the terminology there is it is that purely a price point thing or a bit like why don't you mean you've got your own huge manufacturing facilities we do it's an interesting story some of it I think in retrospect we might not have done but let me play the whole thing out first of all keeping up with the latest surface-mount board loading it's very expensive and it's constant changing by you're not going high density and stuff like this we're not as high density as some products but we're certainly using smaller and smaller parts everywhere and find pimp each BJ's and I know for oh twos and even up to Oh ones in some cases and right D offends which there are almost as bad as BGA's yeah and you know we're moving toward you know three and four thousands traces and spaces oh okay things get denser only because that's what the packages are coming out and they're smaller and you try to fit more functionality and less space you have to but the point was our company made a decision get out of board manufacturing cuz we this was back we got back out after just went surface mount and equipment because already starting to evolve and change and keeping up with it and doing it in a quality way and as rojas happened to yeah we didn't want to make that investment leave it to the you could buy experts yeah you combine that service so we moved the boards out and when we moved the boards out we still did the assembly of the final product in our own site but then we were still part of the this is kind of a related thing is that we were still part of colyn at the time agilent was still building their life sciences products in the states in fact in delaware and a few other locations and they looked at our cost of manufacturing much lower in Oh neither about to invite well our factories only so big yeah and they said you guys got to move out to make space for them to move in Oh so we basically started we went to the CMS we're building our boards and said we want you to build our products and they said there's no problem I have your business and yeah we started to move a lot of the products out of our local and the idea to us that was to lower our indirect overhead they were more efficient at manufacturing because they do a whole lot more of it than we do and it was even less expensive and also one of the other benefits is that they often have small shops in the US that we can work with for prototypes even at the final product level and for some products where the volume is are very small we would just stay in the US yeah we could we could do more localized manufacturing and get quicker turnaround and faster response time for low volume specials and high-value products so we made a decision to move our manufacturing out of our factories to local vendors right in the Penang area north down in Johor and places like that that created some challenges some trained challenges we're not completely through that but we're mostly through most of the stuff that's moved out some of the very old products we did MOOCs we're not going to keep them for that much longer so if you go in in Penang now you'll see production lines for our products but they're mostly the old through-hole old interesting that we don't want to spend the money to move because it costs a fair amount of money to transfer all the test sets and training and all that kind of stuff to other people so one of the challenges for this was this was introduced as we were starting to make some of those transitions right and we were still able to keep we tried to do the new products in our factory and then move it at first now we do direct to CM the new products go right to the CM first product this product was still built in Penang operation they moved out after it went into production but the newer versions of the product they're going right into that I noticed that I've got one of the original made in Penang once yeah but everything's made in Malaysia yeah it's toughened made of our factory it's it's an interesting challenge and I think the cioms have been exceptional frankly at doing a good job it still takes a lot of scrutiny on our part in review and keeping the quality high but these are big big corporate companies they know how to do a good job and they we think the flexibility long term is going to be good for us thank you able to move things around you know customer needs are special and they have a facility one of the three I'm not going to name all that seems we use but he's really good at transferring between the different sites there that's one of their skills and we're seeing that as a potential way for us to respond quickly for modifications and specials and needs so we'll see how that pans out some of these things you know I wouldn't do myself mm-hmm I'm not saying this one is risen but if this one's worked out okay and our goal is long term is to lower cost and get more responsiveness back do you know what the issue there was absolutely yeah you do tell us very interesting Stella Oh tell us the interesting story about how this was respect it came out of the thousand volts spec and then did somebody realize oh did you guys realize that there was an issue there wasn't an issue of it wasn't an issue I said he's blown out of proportion did it blown out okay do-do tell us anyway so what happened was we had a customer who was very interested in our voltmeter to reduce the amount of calibration they had to do and they wanted us to create a custom version of it they've built into their system oh and you would do that for a big enough customer for a big enough customer that would pay us enough sure right and it was also strategic in that we wanted to learn more about their needs so sometimes we justify things by saying this is a door into learn more about their needs and in certain growth areas we'd like to do that and the team that was gonna do that was in one location and they took the design of this and they were gonna reform factor it into a card that went in the customer system so they're going over the design effectively doing a complete design review and they notice that one of the parts in the design the the voltage across the part was above the rating on the part which pop do you know is it not gonna say okay and what happened was they said this is unacceptable and we can't do this and they they said it's a safety issue which it wasn't and they alerted some people and and panic ensued and panic ensued and the knee-jerk reaction was to make a change to safety first that pie change to the spec into the spec and don't do that because yeah we said this could be a safety issue and the reality was that our further review it was not a safety issue but but you did publish the spec you did revise it and published it and hence why it was noticed by people on the forum or big time sure yeah yeah it's not a safety issue right and the part is accepted to use in that application that way and since then it pushed it back to a thousand volts and so you've changed the spec back and everything's hunky-dory yes so there was no changes made no changes made new people looked at it and said this isn't right right and it turns out it was right but people is one of these things we've done it that way for so long no one remembered anything about but it turned out one of the things that got it back to the right was we took our competitors products they did the same way and finally the people who had not looked at this in a long time said you know maybe we need to rethink is this really wrong and figured out it wasn't what's okay and it's back on track you know we don't get everything perfect every time oh there's such a complex set of situations safety is we take very very very seriously so I say if we're gonna have any jerk reaction I'll take it there right okay okay and unfortunately did upset people it was a mistake and we undid it yep and everyone's happy with it now everyone accepts your explanation well maybe they were I don't know if they're looking at this but that's that is what happened in it was a it was unfortunate that it caused the disruption in ways that our customers time wasted our internal time but the fact is you get judged by when you get it right yep Dan yes we're sorry that we accidentally thought something wasn't as safe as it really was yep but that's not on the cautious side that's right yeah a lot of companies would have just don't worry about that we probably could have gotten this one right sooner and not have the option you okay I think it was a little knee jerk on our part but if we're gonna do a knee jerk I'll take it here yeah okay so that's what interesting I was actually involved in that oh I can't because our supplies have a lot of safety issues especially switching off line power supply so I was pretty familiar with it and I was one of the people that knew it wasn't a problem but the certain people thought there were too many opinions in the room and told me to stand down and I say right okay okay and finally came around you know I'm not saying I'm that I was right I'm glad they got it right that's what matters it wasn't one of our best days look honest about it all worked out well yes excellent site um compliance you know the cat testing and stuff like that not so much the bench meters but hand very important in handheld meters and things like that have you ever sent something away that's you know like failed and they've come back and say hey or hey this is not gonna pass or something like that or you do you guys always get it right first time so excited it right the first time not it's a challenge and frankly the new IEC 61 oh one yes you comply with it's much more challenging yes one of the subtle things what's what's the major change for those one of the biggest changes is that you have to anticipate likely mistakes the customer might make all right that course the safety hazard it's a problem as in operational mistakes probes in the wrong everything frozen there is so for example on our new power analyzer which can measure a thousand volts all right and isolated 50 amps of current ice laughs and a thousand volts we have BNC s for the the current input if you want to put an external transducer on the high input and we went to put the triggers that are at ground reference we can see because they could plug the beam on today's show two different connectors yeah no you can just put a label on the back saying don't do that Robin I anticipate likely mistakes and it still has to be safe we could have had a hook up to there if we made it a 25 amp dancer that's the kind of thing I know that's gonna start shifting us into yahweh's you can't just put a label saying if you're stupid you're gonna get hurt it's not accept it won't pass compliance that's correct it's relatively new and but that's not a bad thing no it's offensive it's a smart thing but you know certainly something that you see examples where that was the that was the way you handled it you just said you know Oh can't we just put a label you can't you don't do that right you can't just put a label anymore yeah we're having an issue in some certain the power supplies where the the binding post can only handle 20 amps all right and we're gonna we're gonna have 40 amps so you know what do you do well they pull 40 amps out of a 20 amp Bunny post it melts yep and it can create a hazard it's hot to touch my bridge games so we can't do that anymore used to just you know we put a sense leads on the front panel that's a you know 20 EPS max oh and then it throttles it no no there's no throttle it just says don't do that Oh hold more the design is like that's not acceptable anymore huh it was a purple I see the new the referee of the IEC 61 I won oh so those are some of the standard challenges and one of the other things we did many many years ago for our power products was we switched from an offline system to a 48 volt grounded system we use a distributed DC system and that's made safety a lot easier to meet for us shorten development times because even though our output floats a lot off grounds with a 240 our input is only 48 volts and it's grounded on it as opposed to being 400 plus with a PFC front end so that's actually simplified safety for us a lot in shortened development times and made our performance higher and not directly because we found that the time was taken to all that primary related circuitry the past safety was was increasing test times and compliance testing and certification dramatically now we buy a pre certified brick you know but handles all that thing and you're seeing that with a lot other products I use they brought these with choke snake things or they you know the phone charger type solutions for low power and all the safeties inside that yes sure that's right and that makes things a lot simpler you're going to see more of that for low cost you can see it for a low-cost voltmeter Oh interesting okay because it's just to do all the compliance testing and yep the cost to do it is it's quite a bit higher and those choke snake things and the little plugs in the wall for the phone charges are getting higher and higher in power yep you know there nowadays you can watch crazy your phone your phone can put out can take in like 60 watts yes it's mask for a quick charger you can take it in like 12 volts at 5 amps or something like that so you can get little tiny plugs that can put out a hundred watts you know it's one unit and that's more than enough to power a voltmeter oh it's crazy speaking of power is just standby power consumption you mentioned power factor correction all that sort of stuff is that a big deal in test and measurement is it a requirement like do you got a requirement it's not it is something we care about I know the new one we just came out with people are a little upset because the the power when it's off is not really off and we're actually working on the newer ones and we're gonna make it lower then you will want which product I can't tell you oh oh this is an internal still under development ah rock I need to refine that and somebody internally went hi this is drew an x amount of water well they don't the fan was spinning very slowly oh right you know it's the famous spinning because we don't nobody turn power off and it turns out that we use a PWM fan and PWM fans can make the fan go much slower so it's quite but they never stopped completely those new designs we have to get a switch to turn the fan off so you won't have notices yes it's not so just a little nuisance if you will so there's no requirement if a scope tight you know has a power factor of 0.5 or something you don't know nowadays that once you power five full power PS fees you know even low power the PFCs are available down to what less than 100 watts so you will have a power factor almost any product today will have a power factor of 0.98 0.99 all right it used to be PFC was only used at 600 watts and up now it's even a 45 what brick has right yeah because the chips have gotten so cheap and it's actually above a certain power level it's actually cheaper to have PFC than none ah the components the amount of lost savings and the parts makes up for the extra cost of a circuit I'm sure see the whole PFC fits in a single little hybrid that they make for 50 Cent's oh really yeah it's Wow that's why you still need to choke and think yeah a little bit at your input filter but uh no we we do care about power efficiency at full power I thought you were asking about that pice hey chief a question we are interested in getting the power they were on the bench people do not like when it sits there and consumes a lot of power yeah but there's kind of a threshold about 10 watts energy stars and a 10 watt number in the u.s. that means equivalent in Europe that they like it to be below something but it's not regulatory for test and measurement equipment it is from Zoomer products is the law for consumer God and God yep right so test and measurement gets a free ride well PSC we got a free ride but we start attention to it because yes it's a free ride for us to get yay all right but we do we do have to pay attention to it for standby power because people don't like that those big ugly push switches Oh what do you mean big ugly I like them mmm big clunk and pass but it takes a lot of space up yeah I know yeah you gotta have a plastic rod to go back break you wouldn't believe how easy those things break yes yes they do I've broken a few doing today yeah well not even know it's break during a some cheer and assembly ah right oh maybe then get a bit brittle and then break in the field that's right hi so everything it so we do towards we're going to more soft switching something it also helps boot times because we pre boom guys it's it's sort of waiting it's halfway through the boot process so you hit the switch it comes right back up right but it does draw slight amount of power and we're working to get that lower and lower over time how low can you get like what they do you have like a tile like anything under a walk because it's a nice round number or well we're like limits it is because we're using bricks applies here on this is using a transformer which yes the Transformers engaged and the the power supplies on the output of that are engaged so this is actually there's a limit to how low we can get this but would be the little supply the brick switching supplies they actually are designed for consumer products so they can often spec they go less than a lot when you haven't standby and then they have a small power supply that comes out that keeps a little bit of circuitry alive but why on your TV yeah has this a small amount of circuit then they infrared the infrared to detect you want to turn it on yeah so that that architecture is slowly seeping into our products to minimize the power and you can get those down to like very very low like Condor milliwatts right as to that one button is separate from all the other buttons here yeah speaking of the competitors yes like especially in the power supply market I mean it's a ridiculous what you can get for like on you know you can get a power supply for 50 bucks on eBay mm-hm including delivery probably right no maybe not about that cuz there it's still the box it's just crazy are you at what price point at what price point of instruments will you not go on is there a price point where you'll go look we won't bother today there's probably a price point that's 500 I see the new scope yep but that's just a temporary situation our low ankle long-term goal is to go long go where we need to go to be effective interest if someone makes something that we don't think customers should have on their bench because it's gonna cause more problems than it's worth right we won't go there obviously just cuz someone else makes it doesn't mean we want to put that product out so that $25.00 product may be fine if you're powering up a non-electronic for example but it's gonna have tons of ripple it's gonna be a slow it's not gonna necessary be reliable yeah be easy to use you know we're never gonna make that product yeah but there's there's no reason we can't do a 500 our product as evidenced by YESCO of course we can we can address those spaces right empower in everywhere it takes time for us to turn our organization around to really get the efficiencies and the infrastructure we need to be effective but we're on a path to do it awesome is there enough margin still up that low-end there is to make it or do you have to make it up in volume like well you obviously volume by volume is the point and vine makes a big difference common design common parts and mmm excellent engineering got it it really requires some of the best engineering it's actually to me just like you're doing a six and a half or even in a tin have DMM requires some of the best engineering in the world low cost product design is an art unto itself component obsolescence you clean in about meeting the but its total wasted effort you know you already got a design that works you can ship it and you know spend money that make it continue to ship and the problem is anything you change in that digital domain you have to repeat all the testing all the environmental testing are five years ISD because the new chip could be sensitive when the old ship wasn't of course that's of course big bucks yeah and strangely enough you may have tested it last time and gotten lucky oh I didn't know and then you test my second time and had nothing to do with the changing god yes because last time maybe they change the test setup slightly or the unit's you test last time a little less sensitive and you fail and had nothing so you end up fixing and something it wasn't what you started out to do Wow I suppose common parts they go obsolete ramen ramen ramen ramen before because de di has registers that you're gonna write specific DDR has registered you have to configure so you have to change the code sometimes right change the gr right yeah because I want to thought dollars just pin-compatible right they sorry jelly sometimes it is when you start to do all the testing right and then sometimes it is and you have to change the configuration for the DDR that's one of the once we run into the another big one that's been a big headache recently was and I think it's more of a glitch in time was FPGAs alright yeah one of the two big FPGA vendors moved their foundry from one company in Taiwan to another company in Taiwan yeah the first company I guess wasn't working for very well wasn't on the latest the greatest technology and they decided they didn't want to build anything at the old company so they started obsoleting a lot of the parts that we're making an old company well it's extremely painful 9i um they older FPGAs weren't done with full HD le where schematic base now new tools don't want that except the schematics and the documentation on that pj is never good enough to you just hand it to someone else you give me an original engineer involved if they're still around and it's just we've spent a lot of money in time doing it we're getting good at it but do you have to get I've worked in in industry where we have to get like a guaranteed written guarantee signed by the CEO that they will keep this part alive for 10 15 20 years okay we buy large quantities when they pass through oh I can't lend you a last point we do a bridge bar yeah probably is our products you know the test and measurement history and power supplies in particular have the longest product line yep yeah of course it 20 30 years it's right yeah now the newer product families for this particular FPGA vendor since they've moved to the new foundry they're moving out to deliver the lifetime date they're gonna make it forward because it's working very well for them and they're not trying to get out of it but you know once bitten twice shy yeah so we're taking a more full documentation approach we're doing more training so people can take over and move move the design more easily in the future it's moving completely away from schematic based designs yeah as much as possible to feel full HDL which read compiles in any of the tools all right so that's gonna help us but hopefully this was a one-time glitch the analog stuff there has been some interesting parts that have gone away especially in like a power supply control you know regulators and stuff like that have going away because companies do it for consumer products like you know they've done a really high volume part for a consumer product and then you make it better probably they don't sell anymore the old ones they don't want to make it anymore we designed it in foolishly yes we have to fix it those are not as big of an issue we still have to retest the whole product yep you know it's a bit of a pain of course it's bad if they're bigger but most of the time the switching chips are smaller the next time all right so it's easier and they're more efficient and they work better and so there's advantages to do I sometimes you got to remember an event is doing it in the sense that the products shipping the way it was we didn't need to change your arthritis been one size yeah yeah the fact that it's a little bit smaller well the space was there for the bigger part it doesn't help us right certainly in a new design will move to the newer chip but we're forced to get rid of it in an old design it kind of stinks frankly it's it's a problem that's never gonna go away really is it it goes away by doing more modular ization okay and when you do modernization and you put things in the right groupings we've made some relatively foolish mistakes in some of our products we group things that should have been grouped together so for example this product was a good example this product the ADA DS and das are on the same board with the power supply so that means we have five different versions of this boards so when that A to D changes they arrive we gotta do we have to babble yeah so we've done in the newer products it's all the digital control is on a little plug-in board a memory module right and that snaps in and we actually use that to upgrade to performance the product we have different versions but also when that part goes obsolete that board is common to all the family got it okay and that's helping us we have to do it once and then we just pop it in and do the retest so making some smart decision about partitioning helps to the part of obsolescence quite a bit but sometimes it's just bad you

Channel: EEVblog

Views: 15,459

Rating: 4.8781724 out of 5

Keywords: eevblog, video, john kenny, keysight, low cost product design, product design, how to, soft power buttons, power switches, power factor correction, IEC61010, IEC61010 rev3, 34460a, 34470a, 34410a, cost targets, dc-dc converter

Id: aVBmnVKDJQg

Length: 30min 42sec (1842 seconds)

Published: Fri Oct 20 2017