Author Topic: Ultra Precision Reference LTZ1000  (Read 1333395 times)

0 Members and 1 Guest are viewing this topic.

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #1875 on: October 07, 2017, 09:24:04 pm »
The filters around the voltage regulator will likely make it oscillate. The regulator needs the caps directly at the input and output. The filter at the output of the regulator does not make any sense. Extra inductive filtering might be a good idea at the output and maybe for the negative supply side.

The R18 / R19 resistors should probably be higher value. The ADA4552 does not need such a low impedance.

Usually there is no need to have separate OPs for the 7 to 10 V amplification and the output driver with drive and sense for the positive side. Both can be done with a single precision OP and maybe a second lower grade one to do current compensation for the sense input. Provisions for missing sense connection is a good idea - this could be a resistor with an additional diode for protection.
The output amplifier should be made tolerant to capacitive loads (e.g. extra capacitive feedback).

The noise of the driving transistors (2n3904) is not a problem as they are inside an OPs loop.

An output with separate force and sense only makes sense if there are force and sense for the negative side too. This also includes the ground point of the divider that sets the gain. One might want to use the second half of the 4522 for the negative side - however this would need a different supply (e.g. a negative).

The guard lines make more sense if closer to the right voltage. So either R3 or R16 should be adjusted.
 
The following users thanked this post: TiN, Mickle T.

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
Re: Ultra Precision Reference LTZ1000
« Reply #1876 on: October 07, 2017, 10:11:51 pm »
TiN, if you are open to some observations on your FX reference.
I would apply EMI filtering and bypass capacitors differently. Unless I am misunderstanding things which frequently happens. Very difficult to know all subtleties and wisdom applied.

Having bypass caps grounded to multiple places is not what I would do, because they end up effectively in series, connecting nodes at high frequencies.
Ground for C1, C3 not at U2A pin 4; FB3 would ensure U2A rails are noisy at RF.
Ground for C10, C11, C12 different than C8, C16; shunting FB2.
FA2 looks bypassed; C17 comes first after the Vreg. for stability. If worried about RF from the Vreg, I would place FA2 downstream of C17.

I would break up rails+gnd with FB or inductors into islands, so no intermingling bypass cap grounds, and PVIN filtered to U6, Q2.
I'd draw a pic but it's Thanksgiving long weekend up here.

Running off a battery, I would add a polyfuse so D28 and PCB traces do not burn up if wrong polarity, or perhaps use a series Schottky.
 

Offline Cerebus

  • Super Contributor
  • ***
  • Posts: 10576
  • Country: gb
Re: Ultra Precision Reference LTZ1000
« Reply #1877 on: October 07, 2017, 11:05:02 pm »
1) The 2N3904 is a bit noisy for the output transistor (6dB specification). You could use a low-noise part, typically around 2dB, instead. Or maybe even a series pass FET, as per the Fluek 732.

I really wouldn't worry about it. A typical 2N3904 has an en of 1.35 nVrms /sqrt(Hz) at 10 mA, realistically the bandwidth the reference is going to be used in is probably only 25Hz, so that's a 6.75 nV rms contribution to a 10V output. Even taken as peak that's a mere 1.05 ppb, or if you prefer -167 dB. It's going to be swamped by the intrinsic reference noise at around 2uVptp/sqrt(Hz), 48 dB higher.
« Last Edit: October 07, 2017, 11:09:30 pm by Cerebus »
Anybody got a syringe I can use to squeeze the magic smoke back into this?
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Ultra Precision Reference LTZ1000
« Reply #1878 on: October 08, 2017, 12:24:08 pm »
I wanted to get started on PCB, hence excessive different ground nets, to see how to separate on routing first. Sorry if that cause confusion.
Added first idea for placement. LTZ top left, LDO and heater transistor bottom left, 7V->10V top right and output pass transistor bottom right.
Board size arbitrary atm, need to check metal cases that can be reused. Fluke SIP is bend to right angle, to save on Z-height.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #1879 on: October 08, 2017, 12:46:14 pm »
The 7 to 10 V amplifier part and output buffer need a major change. Especially getting separate force and sense wires for the negative side will likely need a big change to the circuit (maybe including the power supply). So it is still way to early for layout.

Getting the ground right is also important - nothing to decide afterwards how it best fits the layout. It also depends on the output stage - one might even end up better having the LTZ1000 circuit for a negative polarity.

The heater transistor is not such a bad heat source. The only really bad thing about it is that it is so localized, as a distributed heat source it also has positive effects - up to the point of stabilizing the temperature of the whole board. 
 

Offline hwj-d

  • Frequent Contributor
  • **
  • Posts: 676
  • Country: de
  • save the children - chase the cabal
Re: Ultra Precision Reference LTZ1000
« Reply #1880 on: October 08, 2017, 03:37:53 pm »
If this is a really new project, would it be better outside this LTZ1000 mega threat, even maybe projects?
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Ultra Precision Reference LTZ1000
« Reply #1881 on: October 08, 2017, 03:44:05 pm »
It's in good place here, as just another iteration of LTZ module with 10V output.
Also I'm bit lost why negative sense would be helpful, as it's essentially battery powered source.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline ap

  • Frequent Contributor
  • **
  • Posts: 282
  • Country: de
    • ab-precision
Re: Ultra Precision Reference LTZ1000
« Reply #1882 on: October 08, 2017, 06:45:13 pm »
Negative sense compensates for the drop in voltage accros the negative output. While the output amplifier senses at the positive output post, the negative output post is not sensed.
Assume the negative line is 10mOhms and you draw 1mA then you have 1ppm of error through the load. If the load is always high impedance (3458A), then this load offset error would of course not occure.
« Last Edit: October 08, 2017, 06:48:41 pm by ap »
Metrology and test gear and other stuff: www.ab-precision.com
 
The following users thanked this post: TiN, Free_WiFi

Offline kj7e

  • Frequent Contributor
  • **
  • Posts: 911
  • Country: us
  • Damon Stewart
Re: Ultra Precision Reference LTZ1000
« Reply #1883 on: October 09, 2017, 06:33:18 pm »
Quick and simple question, when buffering the LTZ1000 reference, with say an LTC1150, should a series resistor be added to prevent excessive loading should there be a problem with the buffer?  Like 10K or so?  Or is this generally not necessary or desirable?

I have a LTC1150/LT1010 buffer built that works quite well, but I'm not sure if I should worry about the possible loading of the LTC1150 in the off or no-power state.  Such as turn on or if there was some fault with the buffer.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #1884 on: October 09, 2017, 07:23:22 pm »
A series resistor and a capacitor at the input side of the buffer to ground is a good idea for several reasons:
It works as a filter and thus would reduce the higher frequency noise a little. The LT1013 and LTZ zener are usually more noisy than a 10 K resistor. The added noise of the 10 K are not that bad at low frequency. One could ague that maybe 5 K might be better, but 10 K is a reasonable value.

Not powered the buffer circuit might be so much of a load (at least until the supply caps are charged) and might even get damaged from a powerful voltage source without a resistor or similar. The load might be to much for the LTZ circuit and could cause it to turn the heater too high.

The AZ amplifier has some current spikes that contain RF frequencies and thus might upset the reference a little. If paranoid even more than just simple RC filtering might be good.

A resistor (especially if suitable for sufficient voltage) at the input of the buffer might also offer some protection against ESD.
 
The following users thanked this post: kj7e

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Ultra Precision Reference LTZ1000
« Reply #1885 on: October 09, 2017, 07:42:58 pm »
Kleinstein, ap, floobydust, martinr33
Thank you for feedback. I've updated output stage as discussed.

PNG-schematics FX Rev.02

Also I've dig into stashes and got enough parts to build some crude prototype. Sanity check in LTSpice before we go blowing stuff up:



Few hours later...to my surprise it even worked right off the bed.
Critical resistors used are custom VPG VHD200 (20K/7.663K, measured <1ppm/K TCR), and trim S102K 330R + PTF 30R.
Rest are random PTF's. KX LTZ output connected to opamp ADA4522-4 input via 3.01K-1uF-3.01K network. Current drawn from reference is <300nA.
Output transistors are NPN+PNP pair STD830CP40.



Currently I've hooked it to 3458A to log some stability/noise ballpark.

FX output stage with ADA4522-4 and KX channel 6 ref, which is +7.13665V output. Simulation values very well match to real output value.

Live measurement data:



Last year data from Fluke 732B vs another 3458A, 24 hour, NPLC 100 to comparison:



Also output is trimmed (R3 on LTSpice schematics) using S102K 330ohm + PTF56 30 ohm 50ppm/K. Per calculation shift of this resistance network to 100ppm will cause output to change 1.3ppm.

I'm pretty happy with the result, given it has zero shielding, no air drafts protection.

Both LTZ reference and prototype amp powered from Fluke 792A battery pack (+12.5/-12.5V SLA).
This is where this build will eventually go, to make it a transportable 10V standard to use for 3458A/5700A calibrations.
« Last Edit: October 09, 2017, 07:45:29 pm by TiN »
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #1886 on: October 09, 2017, 08:03:17 pm »
The shown output stage should get extra compensation / local feedback at the OPs. This would be a small capacitor (e.g. 100 pF range) from the OPs output to the inverting input. R8 is should be a little larger for this. This should reduce ringing and sensitivity to capacitive load.

With a different output stage at the negative side one might be able to get away with a much lower negative supply, maybe down to 0.5 V.
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Ultra Precision Reference LTZ1000
« Reply #1887 on: October 09, 2017, 08:13:16 pm »
I've tried load output with 5Kohm and 47uF cap, that shown just as 0.4ppm shift. 1uF load worked fine as well.

Opamp and BJTs powered directly from battery rails, no LDO. Will see how much margin is there before if go out of regulation. Per opamp DS with 1mA load it should be better than 100mV from rail voltage.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline chuckb

  • Frequent Contributor
  • **
  • Posts: 339
  • Country: us
Re: Ultra Precision Reference LTZ1000
« Reply #1888 on: October 09, 2017, 08:14:32 pm »
I usually design in one other component to limit potential damage to the LTZ1000A chip. Depending on expected operating temperature, I add a series voltage dropping Zener at the collector of the heater transistor. This limits the peak chip temperature rise. Other people have mentioned that if the LTZ1000A chip gets somewhere over 100 deg C it will affect aging. Over 150 deg C can cause permanent damage.

My last design needed to work in a warm environment so I ran the chip at 95 deg C. After it warmed up (in a cool environment on the bench) I think I had 8 volts to the heater. With a 15 V power supply I had a 4V Zener in series with the collector of the heater transistor. The warm up time was a longer than normal. However, when I was checking operation on the bench I did not have to worry about my test probe slipping, shorting out the wrong node and overheating the chip.

When starting up a new design I use a separate PS for the heater voltage and I start with something like 5V till I verify that I have the PCB design correct. I increase the 1k ohm heater transistor base resistor to 10k on some low temperature designs to prevent a sneak path for heater current during warm up.

Adding a series voltage dropping Zener to the collector of the heater transistor is good insurance. The voltage dropping Zener may also help aging in a reference that has frequent power cycles. There will be less peak stress on the chip as it warms up. With 13V across a 300 ohm heater there will be 560 mW of heat on the chip until it reaches operating temperature where it needs around 100mW (65 deg C chip, LTZ1000A) to maintain temperature.

The non A version of the chip may require full heater voltage for correct operation.
 
The following users thanked this post: TiN

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
Re: Ultra Precision Reference LTZ1000
« Reply #1889 on: October 10, 2017, 01:41:30 am »
I think it's a good idea to limit the heater output.

The LTZ heater resistor has a spread of more than 2:1; 200-420ohms with 300 ohms nominal.
Then factor in the "A" needs 1/5 the heater power of the non-A and you could have way too much... gain...heat...

Could you move the zener (led?) between from LT1013 op-amp (-) input to heater transistor emitter, to voltage-limit the LT1013's output.
It's just zener's heat has to be kept away from the ref amp and LTZ. Or use current-sense perhaps.


note the datasheet heater graphs, unknown physical configuration- free air, on a pcb etc.
« Last Edit: October 10, 2017, 01:43:12 am by floobydust »
 

Offline chuckb

  • Frequent Contributor
  • **
  • Posts: 339
  • Country: us
Re: Ultra Precision Reference LTZ1000
« Reply #1890 on: October 10, 2017, 02:12:04 am »
You could use a device between the heater transistor emitter and the (-) input of the op amp but I don't like to do it that way.  I try to limit the wiring going directly to the op amp inputs. They are just another EMI antenna connected into a very very sensitive node. An LED or a glass Zener may generate current when exposed to light or have leakage currents that will disrupt this sensitive circuit node.

A resistor or an Zener on the heater transistor collector can be located far away in the corner of the pcb where the heat does not influence operation.
 

Offline MK

  • Regular Contributor
  • *
  • Posts: 233
  • Country: gb
Re: Ultra Precision Reference LTZ1000
« Reply #1891 on: October 10, 2017, 06:43:14 am »
You could use a device between the heater transistor emitter and the (-) input of the op amp but I don't like to do it that way.  I try to limit the wiring going directly to the op amp inputs. They are just another EMI antenna connected into a very very sensitive node. An LED or a glass Zener may generate current when exposed to light or have leakage currents that will disrupt this sensitive circuit node.

A resistor or an Zener on the heater transistor collector can be located far away in the corner of the pcb where the heat does not influence operation.

Whilst agreeing with the heat flow issue, surely a raised potential at the heater emitter will reduce the parasitic diode leakage?
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #1892 on: October 10, 2017, 11:33:49 am »
Zener diodes have leakage well below the nominal voltage. So it is not a good idea to modify anything at the OPs input side. Limiting with a zener diode in series with the transistors collector is good (if the supply is reasonable stable). This will also spread out the power that otherwise would be all at the transistor.

Another option would be at the transistors base towards ground, so limiting the maximum output voltage to the heater. Here a VBE multiplier (transistor + pot)  might be an option. A power limit decreasing with high temperature is nothing bad.

Due to the spread in heater resistance and different thermal insulation / set temperature one might have to adjust the maximum heater voltage individually.

The higher temperature gain for the heater in the A version is not a real problem. Due to the square law with the resistor, the change in power for a change in voltage is also lower. So the gain is not that much higher.

@TiN: the ouput driver looks like it is only marginally stable. Even if it does not oscillate with 47 µF at the output this is not really guarantied. Capacitance from the outputs towards the circuit ground could cause trouble. Small extra caps at the OPs should help so that caps at the outputs could be used as a measure against EMI.
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Ultra Precision Reference LTZ1000
« Reply #1893 on: October 10, 2017, 12:11:50 pm »
I'd say that large capacitance is rather extreme and unusual. Added provisions for C39,C40,R32 if that what you mean.

Also overnight span with deliberate temperature changes from +25.6 to +31 didn't reveal visible correlations, and output still in 0.4ppm window.  :=\
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Offline Kleinstein

  • Super Contributor
  • ***
  • Posts: 14076
  • Country: de
Re: Ultra Precision Reference LTZ1000
« Reply #1894 on: October 10, 2017, 12:46:43 pm »
C39,C40 and R32 is what I meant with extra compensation.

The caps C41-C44 look a little unusual, just two caps from the transistors collector toward the power ground would be the more normal way.

One point someone already mentioned before is that is might be a good idea to have a resistor or diode between the force and sense pins. This way the voltage would not increase very much if the sense connection is lost for some reason.

less than 0.4 ppm variation over a 6 K temperature range is very good, given that some of this is likely noise and maybe still some initial drift. It won't be super good for the LTZ1000 alone, but it is for the amplifier.

One could get a small negative supply from just a diodes drop or two. This could save you the second battery pack. With RR Ops a single 12 V supply might be just sufficient, though a little more (like 14 V) would make things easier.
 

Offline chuckb

  • Frequent Contributor
  • **
  • Posts: 339
  • Country: us
Re: Ultra Precision Reference LTZ1000
« Reply #1895 on: October 10, 2017, 01:14:18 pm »
Whilst agreeing with the heat flow issue, surely a raised potential at the heater emitter will reduce the parasitic diode leakage?

I measured the leakage current of the parasitic diodes inside the LTZ chip several years ago. The leakage was less than 1 pa at 15 V bias. So the internal parasitic diodes shown on the LTZ data sheet are very good, low leakage diodes.

External Zeners or normal diodes have to be carefully designed in to prevent leakage currents or EMI demodulation issues.
 

Offline TiN

  • Super Contributor
  • ***
  • Posts: 4543
  • Country: ua
    • xDevs.com
Re: Ultra Precision Reference LTZ1000
« Reply #1896 on: October 10, 2017, 01:31:26 pm »
Second battery pack is available regardless, as I still intend to keep original function of the power pack to provide +11/-11V to F792A, once such need occur. This LTZ design is effort to make it more useful, instead of having just a large box with batteries sitting around, doing nothing. Also floating 10V ref in lab can be useful for bridge/null metering purposes, as I have no means to invest into real 732A/732B. Another challenge is how to fanout this circuit output to the external world. Adding 4 binding posts would be non-trivial due to triple shielding shells.

I'll also have to implement under-voltage shut-off circuit, to prevent LTZ+amplifier draining batteries too much, but that's outside of this thread scope.
YouTube | Metrology IRC Chat room | Let's share T&M documentation? Upload! No upload limits for firmwares, photos, files.
 

Online floobydust

  • Super Contributor
  • ***
  • Posts: 6924
  • Country: ca
Re: Ultra Precision Reference LTZ1000
« Reply #1897 on: October 11, 2017, 12:26:02 am »
Zener diodes have leakage well below the nominal voltage. So it is not a good idea to modify anything at the OPs input side...

Yes the heater control op-amp inputs are very sensitive; definitely no zener there to limit the heater power.
The impedances at about 10k and 70k I thought moderate, but the gain is extremely high.
1uA there causes about 7-17mW/500mW power shift.

I used KO4BB SPICE model for the LTZ1000A heater.  The heater transistor is run common emitter though.
 

Offline branadic

  • Super Contributor
  • ***
  • Posts: 2378
  • Country: de
  • Sounds like noise
Re: Ultra Precision Reference LTZ1000
« Reply #1898 on: October 19, 2017, 04:04:40 pm »
Interesting to know, for some reason Rhopoint has changed the specification for their 8G16D resistors from ±3ppm/°C to ±5ppm/°C.

-branadic-
Computers exist to solve problems that we wouldn't have without them. AI exists to answer questions, we wouldn't ask without it.
 

Offline Vtile

  • Super Contributor
  • ***
  • Posts: 1144
  • Country: fi
  • Ingineer
Re: Ultra Precision Reference LTZ1000
« Reply #1899 on: October 19, 2017, 11:20:10 pm »
Since the heater/oven seems to be one of the achilles (as lazy reader of these electron splitting threads, which are really interesting. :) ) of the LTZ1000, why everybody is doing analog controller of the oven. Put it digital and run it through MPC or even digital PID....?  :o
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf