[FT] xDevs.com KX LTZ1000A fairy-tale, or the story of little jumper.

[VK5RC]

I love the solder (leaded of course) with just a little bit of Silver, a percent or so - seems to flow much better - I used .3mm when building my LTZs. I managed to make a solder bridge on two occasions due to the closeness of the guard trace (and my shake!!!) . Found before power on fortunately. 

[TiN]

You guys make me work... Received KX B03 PCB from SvanGool. Looks quite nice, and substrate is lighter color, which is usually for higher Tg substrates. I'd expect these boards are bit more pricey than mine.



Couldn't resist putting few caps already before taking photos, sorry. Now the "problem" is that I'm out of LTZ chips, so might just order one, likely ACH-version.
Also this one will have Edwin's PWW resistors, as all my others are with BMF/MF.

Are you excited? Which opamp shall I use, ADA4522-1 or usual LTC2057? (yes, LT1013 is enough, but that's not the point  )

[Kleinstein]

The LT1013 is not only good enough, it is about the best suited OP available. There is a very similar version from Ti (OPA2234). If there is a negative supply available (or in the negative voltage circuit) there are a few more choices.

The problem with AZ OPs is, that they are usually more sensitive to EMI and also produce EMI. So I see no good reason to choose an AZ OP in a circuit that really don't need it. An AZ OP can be a good choice for a stage behind the reference, e.g. amplification to 10 V.

If one would really like to go beyond the standard circuit, the more obvious modifications are not different OPs, but more like better RF tolerance and maybe a linearized and power limited heater circuit or even a second layer of temperature stabilization. A different OP might be needed if for some reason a lower noise in the 10-100 kHz frequency range is needed - but this would not be an AZ OP.

[Andreas]

Hello Illya,

try the ADA4522 then we will have an option when AD drops the LTC2057 
On the other side: the ADA4522 seems to have a capacitive load problem (stable up to 250 pF) 
So perhaps that is not the best option (together with the lower open loop gain).

with best regards

Andreas

[TiN]

Perhaps I've used wrong word regarding 1013. The board design use single-amp package only, so 1013 is not a direct option here.
Indeed if one need proven design, there is no reason to go for anything other than 1013, which is what we see in almost every commercial device with LTZ1000 ref.

ADA4522 was already used in module built at very beginning of this thread. No visible difference from other 2057 or A9 3458A ref in terms of tempco or output noise, based on limited testing.

[David Hess]

Perhaps I've used wrong word regarding 1013. The board design use single-amp package only, so 1013 is not a direct option here.

The LT1006 is a single LT1013/LT1014 but has offset null and programmable supply current for low power or high speed (low noise) operation.

I forgot; what makes the LT1013 uniquely useful in this design?

[Kleinstein]

The LT1013 is a precision single supply OP with about the right balance of input bias / current noise and voltage noise.
Single supply is convenient as the common mode voltage is about 0.5 V in the positive ref circuit.
Voltage drift and low frequency noise is not that important, as the transistor in the LTZ1000 already provides a gain of about 200. The source impedance seen by the OP is a little below 70 K (of whatever resistors are used to set the transistor current).

There are a few rather similar alternative, like OPAx234 or maybe even OPAx171.

[Alex Nikitin]

The LT1013 is a precision single supply OP with about the right balance of input bias / current noise and voltage noise.
Single supply is convenient as the common mode voltage is about 0.5 V in the positive ref circuit.
Voltage drift and low frequency noise is not that important, as the transistor in the LTZ1000 already provides a gain of about 200. The source impedance seen by the OP is a little below 70 K (of whatever resistors are used to set the transistor current).

There are a few rather similar alternative, like OPAx234 or maybe even OPAx171.

Definitely not the OPAx234, as its common mode range doesn't include the negative supply. The OPAx140 may work rather well though.

Cheers

Alex

[TiN]

Assembled the board. 



But likely got LTZ's temperature sense transistor killed by ESD. 
Dropped chip on the floor by accident. Now heater section does not work, pin 8 stuck at 0.11V and output is +7.50x V. Removed chip, moved on my other module with socket, same stuff.

Moral of the story : don't drop your LTZ's and wear ESD-protection.   

Now need to get a replacement. Digikey will be happy (need to get some other parts, so i'll just drop in fresh chip there).

[ManateeMafia]

At least you didn't do it live with witnesses like someone I know hooked up battery terminals backwards while a friend was watching remotely. 

[Andreas]

But likely got LTZ's temperature sense transistor killed by ESD. 

Shure you didnt kill the resistors?
(bending the wires without strain relief directly at the case).

With best regards

Andreas

[Cerebus]

At least you didn't do it live with witnesses like someone I know hooked up battery terminals backwards while a friend was watching remotely. 

I've done that often enough that breadboards and circuit boards that are likely to be 'messed about with on the bench' get reverse biased 1N4007's or similar slapped across the supply rails on the board before I do anything else. It doesn't help that my favourite bench PSU (favourite because it's fan free and also electrically quiet) only has a fixed current limit.

[plesa]

But likely got LTZ's temperature sense transistor killed by ESD. 

Shure you didnt kill the resistors?
(bending the wires without strain relief directly at the case).

With best regards

Andreas

what me worry is long term stability because of stress in bobbin.

[mimmus78]

At least you didn't do it live with witnesses like someone I know hooked up battery terminals backwards while a friend was watching remotely. 

Mostly is bad luck.

I tortured one of my ltz reference badly.

Hocked power supply backwards on my fist KX reference build because Siglent and Agilent power supplies have swapped positive and negative binding posts.

Than I soldered backwards because I had used "mirrored" footprint on my design.

Soldered/desoldered it something like 10 times when testing many KX and my design.

You won't believe me if I tell you is still working and stable. It's just a little bit noisier than normal, mostly cured by using 100 ohm resistor instead of 120.

Inviato dal mio Nexus 6P utilizzando Tapatalk

[MisterDiodes]

TiN <respectfully, suggestions just based on long experience>:

For all the time, money and effort you're going to, I'd just order some new resistors directly from Edwin.  When you re-install, -don't- kink the leads against the resistor body like you've done.  Not only does that create brand-new unstable resistors in the leads where you've cold-worked the alloy and strained it with high compression and tension zones, but that can also literally jack-pull the lead out of the body internally - which can tug on the weld point.   In some cases you also destroy the resistor lead's ability to carry thermal heat flow in and out of the resistor wire - which in a PWW is where a lot of the heat flow must to take place if you want stability.

Otherwise any tests you do from now on are just checking damaged resistors, not known good production quality PWW's.

In our lab if we saw a precision resistor mis-handled like that it would immediately be rejected & tossed - just too many potential problems later on. Only use a gentle lead bend radius when assembling and never expose the body to lead stress - be it mechanical or high heat.  Better yet: use a proper lead former jig, and follow good minimum bend radius specs (min radius is at least 4x lead wire dia or more) and as Andreas pointed out you want length of untouched exposed lead at each end of the body as strain relief.  We use a heat sink properly when assembling -any- precision resistor, whatever type.

If possible, for a more accurate test use the correct PC board layout pad spacing to accommodate the resistors also.  Precision circuits call for good power in / power out stability, and that means stable heat flow.

[ManateeMafia]

@MisterDiodes

Would you recommend a simple bend or one that includes a strain relief bend in the leads?
I have looked at some of the tools available but I cannot decide if it is worth the added expense. Personally, the simple plastic lead forming tools would be adequate for me. As long as I get repeatable bends I can make the right component in my CAD software.

[ap]

You could use standard tools, just make sure when the leads are bent, the bend is between two tools (not a tool applying force on one side of the bend, and the resistor body being to other part that applies a force to that wire bend area, because then stress is applied to the resistor body).
As always, good workmanship re. parts preparation (as well as soldering...)  is specified in NASA/ESA or MIL documents.

[Cerebus]

If you don't have a purpose designed lead forming tool to hand may I recommend  the use of locking surgical forceps (also known as haemostats) for bending leads and for soldering. Such as these:



They let you get a very positive grip on a lead without damaging it (they are designed to clamp arteries without damage) and you can bend it without transmitting any force to the component body. Use one pair to hold the lead and another as an anvil to get a nice radius in the bend.

In my experience they make much better soldering heatsinks than most of the purpose designed tools. They are also great for picking up that tiny screw you just dropped into the depths of a chassis while reassembling something.

They come in a range of sizes and, best of all, they are cheap, you can pick up a set of 4 on flea-bay for under £2 each. It's one of those tools that, once you have some, you'll wonder how you lived without them all these years.

[VK5RC]

+1 re forceps - that is what I used in forming the leads of the 'good' resistors of the LTZ KZ reference - pictures in reply no 257 of this thread - I also used them as a heat sink when soldering the resistors and the LTZ to reduce heat stress - but with the LTZ I think you may replace the heat stress with a mechanical one. You also end up with the LTZ 4mm or so above the PCB. Swings and roundabouts.

[Cerebus]

+1 re forceps - that is what I used in forming the leads of the 'good' resistors of the LTZ KZ reference - pictures in reply no 257 of this thread - I also used them as a heat sink when soldering the resistors and the LTZ to reduce heat stress - but with the LTZ I think you may replace the heat stress with a mechanical one. You also end up with the LTZ 4mm or so above the PCB. Swings and roundabouts.

The various types of forceps have names, the commonest one you'll find are called Kelly forceps. Mosquito forceps have, as you might expect, narrower jaws and are a better option if you're concerned about introducing a standoff between a device and board when using them as soldering heatsinks.

[MisterDiodes]

@MisterDiodes

Would you recommend a simple bend or one that includes a strain relief bend in the leads?
I have looked at some of the tools available but I cannot decide if it is worth the added expense. Personally, the simple plastic lead forming tools would be adequate for me. As long as I get repeatable bends I can make the right component in my CAD software.

Generally as few bends as possible, but don't over-kink the leads.  We define a kink here as any curve with a radius < 2X wire diameter - that can really inhibit heat flow.  Keep the bends as gentle as will fit your application.

We use an in-house built jig that forms several components at once and acts as an assembly heat-sink, but there are commercial adjustable component formers that are required for military / aerospace / medical use - and for prototypes we also have some good round-nose pliers we can use as a lead bend former (dia around 0.200").

A well-equipped assembly contractor will have a thru-hole board stuffer / prepper with a handling head that puts on a smooth adjustable-radius bend to your precision components - you specify the pad spacing and minimum allowed curve radius, and they handle the rest.  Sometime they will build a component former block that fits into the machine custom-made for your project.

The plastic lead former tools are better than nothing, and can work well if you're careful but those can make an over-sharp corner on larger lead wire.  You might have to modify the cheap lead former so that it puts on a larger-radius curve.

The general idea is when you bend the component lead, keep all stress away from the resistor body - so that means you really want to grab the lead in two places and guide it around the former corner - don't make the resistor body take any bending stresses, and don't extrude or cold-form the lead as it goes around the corner. 

Don't let the soldering heat get directly into the body either - if that epoxy or body softens you can get into trouble and ruin a good component.   Keep the soldering time short and precise - and make use of a heat sink.  Even some stainless steel wool, de-solder braid, coax braid, etc. held against the component lead while soldering can act as an emergency heat sink if you don't have a clip-on style.

[ManateeMafia]

Thanks to everyone for their suggestions. 

It seems all details cannot be ignored when working at the ppm level.

[TheSteve]

SvanGool was kind enough to send me a PCB so I am joining the club.

I plan to build it using resistors sourced off of ebay. This does limit the values I can acquire a little bit. I will be using 12.5K for R5 and am wondering if people recommend keeping R4 at 1K or would 900 ohms be better to run the LTZ1000A a little hotter?

[TiN]

That really depends on your max ambient. Hotter gives you little higher long-term drift rate in theory, but in practice you wouldn't be able to measure, unless you has banks of 732A/B. 

[MisterDiodes]

SvanGool was kind enough to send me a PCB so I am joining the club.

I plan to build it using resistors sourced off of ebay. This does limit the values I can acquire a little bit. I will be using 12.5K for R5 and am wondering if people recommend keeping R4 at 1K or would 900 ohms be better to run the LTZ1000A a little hotter?

12.5k at the very low end (based on experience), if your ambient is pretty much always on the cooler side.  You'll have to see how your board works in your enclosure, insulation, etc. We use recommended 13k as lowest value, but on a warmer production line situation we'll go with 14k~15k, as used by HP in 3458a meters.

If your circuit is running too cool it will stop being stable at higher ambient temperatures.  You need that LTZ to be above ambient (ambient temp around the LTZ) at all times so it can stay at a stable point...It only has an on chip heater, not a cooler.

TiN is correct:  The change in drift rate at higher temps is not significant for a hobbyist to worry about and unless you have several 732a/b's to compare against.  And maybe not even then.  The bigger contributor to overall drift rate will be your unique LTZ die and the luck of the draw - beyond your control.  It is not uncommon to see LTZ's that really don't have much measurable change in drift rate at higher temps... but some are more sensitive of course.

I can also point out several 3458a's that we've used that run at their standard 15k / 1k that don't drift a ppm / year, if that makes you feel any better.  They are perfectly happy at that heater ratio, and have been running for decades that way.

I'd assemble your circuit, fire it up, get it in your enclosure and let it run a while - and then you can test what effect ambient temp has on -your- circuit situation.  Let it run several months to settle in and establish a reference baseline.  LTZ circuits take some patience to really study and if you've done it right it's really about the best reference you can build without a government-sized budget.