LM399 based 10 V reference

[jorgemef]

..and what is your voltage regulator there? The resistors values around it are quite large..

It is a HT7550. I keept the resistors high to keep low bias current. The circuit is slow due to the time to charge the capacitors. Takes around 1 minute to reach final voltage of 14.5V. I am feeding it with a 5S battery.

[Kleinstein]

With parts from a somewhat dubious source I would not really count on the TC. They only say TC down to 10 ppm/K. This means this series of resistor may offer down to 10 PPM/K grades, but the parts actually sold may still be a lesser grade.
Using resistors from the same batch can result in good TC matching, but it is not guaranteed.

The first point to improve would be to improve the ground routing: take out the shart GND wire near the LT1001 and give the heater current a separate return path to the regulator or power connector.

[dietert1]

If a temperature variation of 29 - 27.9 = 1.1 K gives 57 - 37 = 20 uv this is about 3 ppm/K of 7V after the amplifier, more than expected for a LM399. For the 10 V output the TC is more like 6 ppm/K. Maybe the meter used to determine these graphs exhibits some TC to start with.
The difference between 3 and 6 ppm/K is certainly caused by the voltage divider. What is the TC spec of those resistors?

Regards, Dieter

1/2W 207. As per seller specs is about +-10ppm/C. https://pt.aliexpress.com/item/1005006320806775.html

So you mean the circuit is performing as per component specs, and I need to look for strategy to reduce the circuit TC like adding copper resistance on the lower side of the feedback network (and other resistor on the upper side to maintain the 10v)?

In general 0.1 % resistors are specified with TC of 10 or 15 ppm/K, e.g. PTF56 resistors. So those 1 % resistors you got may be more like 50 or 100 ppm/K and even if they are from the same lot and you use kind of averaging, 3 ppm/K as a residual TC is better than expected. Yes, one can use copper wire to tune the TC. In order to do that one needs some kind of temperature chamber in order to vary the temperature of the circuit while the meter remains at near constant ambient temperature.

Regards, Dieter

[iMo]

..and what is your voltage regulator there? The resistors values around it are quite large..

It is a HT7550. I keept the resistors high to keep low bias current. The circuit is slow due to the time to charge the capacitors. Takes around 1 minute to reach final voltage of 14.5V. I am feeding it with a 5S battery.

It is a CMOS 100mA low drop regulator with low 5uA quiescent current (at no output current), but why you want to have those resistors so big? I would use something 50-100x smaller.. Bias current is not important here. With such large resistors it may induce some instability, noise etc..

[Andreas]

Hello,

the PCB looks not clean, I would wash the PCB before testing.
The "hook" may also introduce errors with many contacts of not equal metals.

1.5 / 2 degrees span to measure T.C. is also not enough.
You may have several degrees temperature differences across a board depending on where you put your temperature sensor.
(especially when you have a heated reference).

That is also the reason why I would not try to compensate several ppm/K with copper resistors. Some 0.1 ppm/K would be ok.
(you never know which temperature should be sensed by the copper).

Did you check wether the heater is working. (how is the current consumption of the PCB?)
The last time when I had a LM399 with large T.C. I recognized that the heater was not working (had a bad contact).

with best regards

Andreas

[iMo]

3ppm/C of the divider's ratio TC is would be a pretty good result, but as Andreas wrote the 1C span might be not enough, moreover, his temperature resolution is rather low. I saw that in my measurements too - I had large diffs in TC between smaller and larger temp spans.

[iMo]

The Vreg variant.. The two 10u input/output capacitors C1 and C2 around the Vreg should be wired to ground. In your wiring the two important decoupling capacitors C1 and C2 are "missing", because they are wired to ground via 300k..

[jorgemef]

The first point to improve would be to improve the ground routing: take out the shart GND wire near the LT1001 and give the heater current a separate return path to the regulator or power connector.

The heater GND and Power follow different path. They meet almoust at power main decoupling cap (470nF).
The ground to the zenner is almoust common to the ground from the LT1001 which is buffering it, and the VZ is very close from IN+ from same buffer opamp.
I increased now the size of the ground and power lines up to the heater lines derivation to try to reduce eventual effects by adding a layer of solder of 1mm.
I also noticed that the 10nF may not be enough to the amplifier stage bypass capacitor. I think I will increase that to 100nF. I see some examples on LT1001 datasheet with it and I have a second board where the noise was close to 300uV. After much scratching my head and many tests I decided to test with outer 22nf cap in the bypass position and the noise immediatly went down, so I guess will go with 100nF.

I think I will follow iMo suggestion for the reduction of the biasing resistors. Makes litle sense to spare 1mA of biasing current when the LM399 are consuming 17-20mA to the heater.

As for the controlled temperature room to test this stuff still need to think how I can do it. The air flow of the air conditionair is not the best way to test. The best I can do is turn it off during the night and goes to 18ºC and turn on during day and goes to 24ºC. But this also impacts the Keythly2000.

[jorgemef]

The Vreg variant.. The two 10u input/output capacitors C1 and C2 around the Vreg should be wired to ground. In your wiring the two important decoupling capacitors C1 and C2 are "missing", because they are wired to ground via 300k..

I took this from the datasheet. I found this strange but I guessed it may be to keep the noise low between the output, input and gnd.

[iMo]

..and you have to recalculate the R1 and R2 - see below, the same principle as with the LM317. I would wire the C1 and C2 to ground..
For Vout=15V   
R2=2*R1

[jorgemef]

Went back to #1251 to iMo simulations. :/ I guess will use only 22nF on the one misbehaving or even test a different SMD part of 10nF not to make the circuit more susceptible to the supply noise.

[iMo]

Went back to #1251 to iMo simulations. :/ I guess will use only 22nF on the one misbehaving or even test a different SMD part of 10nF not to make the circuit more susceptible to the supply noise.

The #1251 is a simulation only.. The opamp models are mostly behavioral (not sure about OP07 there), thus my simulation might not be perfect (or it could be imprecise). Anyhow, the larger the capacitor in the feedback the slower the opamp's response and therefore it may respond to the power rail noise/ripple slower (experts here may clarify).. Also mind your schematics is a little bit different.

[Kleinstein]

For the heater ground I somewhat mixed up the heater and zener GND. So the layout in this respect is OK, though not great with the OP amp supply shared with the ref. votlage.

For the capacitor in the LT1001 feedback (C7)  10 nF should be large enough. It may be an issue if the output emitter follower sees very little current. Currently this is only R16. At least for the PCB this is 1 K and not 5 K as in the plan. 5 K would be borderline low current for the reference.  A larger capacitor for C7 would be needed with really large capacitive load..


The position of the capacitor (C1 in the HT75xx datasheet) at the regulator input is defintely wrong - it looks like a mistake in the regulator data sheet.  As shown the input side capacitor would couple supply noise directly to the regulator feedback, forming a divider with the capacitor on the output side. So one would get a PSRR of only 6dB, once the capacitors take over. The capacitor at the output side may be OK.

edit:
The capacitor at the output side is also bad, limiting the speed how fast the regulator can work. 
The

[jorgemef]

I replaced the 10nF SMD with another one of same value on the second board and noise came down, so probably stress from soldering or bad part.
So now noise on the second board at 10V output is of 10uV over 30 samples/seconds like in the other board.

I have used 3K for the current refference. Need to update the schematic later.

Will update the C1 position and reduce the voltage divider resistances there to make the response faster

[iMo]

..Will update the C1 position and reduce the voltage divider resistances there to make the response faster

Your board will take perhaps 80-100mA when cold, after couple of seconds when already hot the current drops down to perhaps 30-40mA. If I were you I would go with R1=1k and R2=2k (in my Vreg schematics) or something like that.
C1..100uF/35V and in parallel 100n/50V ceramics.

.. and the power loss (==heat) of your voltage regulator: with 21V input (with your 5S) and 15V output the loss when zener is cold will be aprox 0.55W (!!), when zener is hot 0.17W (!)..
Your HT7550 will be hot (like 40-45C at T=23C ambient, add say 5-10C coming off the 399 and others). Do not use more than 16-17V input. It could be your 5S battery is too much (21V max), try with 4S instead (16.8V max).

.. and re #1262 point 5. - when the battery voltage goes down by 1V, for example, the heat dissipation at the 7550 goes down by aprox 30-40mW.
That creates a temperature drop around the 7550 by aprox 3-4degC (doublecheck that).. Multiply it by your TC..

[jorgemef]

Do not use more than 16-17V input. It could be your 5S battery is too much (21V max), try with 4S instead (16.8V max).

I considered using the 4S but then the lower range of the battery pack would be 12V. That is why I added the jumper to be able to set the config of the LDO to 12V.
But I see the 10V output changes when switching from 15V to 12V by 40 uV. Not sure if the LM399 behaves well with 12V or something else going on with the opamp headrooms.
The datasheet specifies 2V dropout for the HT7550 so that would be additional challange for the regulation with a 4S.

I will do some measurements. If I see problematic on keeping the regulator onboard will probably build a doughter board to keep the regulator temperature out and just shunt VIN/VOUT on the board layout for the HT7550.

[iMo]

I can see "100mV" dropout in my DS (at 1mA out), with 30-40mA it could be more..
Would be better to have the Vreg off the reference board and use a higher input voltage, sure..

[jorgemef]

I remembered Dave video that opamp offset changes with supply voltage. I have two in series. Now I realize why changing from 15V towards 12V the 10V output changes. I imagine that same happens in a smaller scale as the battery empties or you have changing load  but the LDO keeps this in check for some degree.

[David Hess]

I remembered Dave video that opamp offset changes with supply voltage. I have two in series. Now I realize why changing from 15V towards 12V the 10V output changes. I imagine that same happens in a smaller scale as the battery empties or you have changing load  but the LDO keeps this in check for some degree.

PSRR (power supply rejection ratio) is the change in offset voltage for a change in the magnitude of the supply voltage.  CMRR (common mode rejection ratio) is the change in offset voltage for a change in the voltage at the non-inverting input.  If the voltage at only one of the supply pins changes, then both apply.

For precision operational amplifiers, PSRR can usually be ignored, and CMRR can be ignored if the non-inverting input does not change much.  Changes in offset of better than 1 microvolt per volt of power supply or common mode change is typical for precision parts.

[Kleinstein]

With battery operation it would make sense to have a low drop out regulator. This way one could use more of the battery voltage. This is more improtant than the ground current of the voltage regulator.

The PSRR of the OP-amps is usually quite good. A second path is via the reference heater. The heater voltage has a slight effect on the actual temperature.
The voltage regulator should keep the voltage variations small enough to not have a problem from the supply side.

With a 12 V supply the output swing for the LT1001 may not be sufficient. The specs give some 1.5 to 2.5 V that are lost at the upper end and there is also the emitter follower that looses some extra 0.6 V. So one could just hit the upper output limit and to drive the output that hard it may need a little extra voltage at the input. From this I would consider some 13 V as the lower limit for the LT1001.

[Andreas]

Hello,

there are better output cirquits for low headroom.
See e.g. DATRON 4912 cirquit for 10V output from a 12V supply.

with best regards

Andreas

[iMo]

I can see "100mV" dropout in my DS (at 1mA out), with 30-40mA it could be more..

In Farnell's datasheet for the HT7550 I've found the definition for the "Dropout Voltage" 

Dropout voltage is defined as the input voltage minus the output voltage that produces a 2% change in the output voltage from the value at VIN= VOUT+2V with a fixed load.

For the 7550 it is 25mV typ., max 55mV (at 1mA out). Thus it seems it cannot be powered from 4S (for the 15V out), indeed..

Holtek's datasheet calls it the "Voltage Drop", for 7550 it is 100mV typ. (at 1mA out), with no further explanation.

PS: you may have a look at the TI's slup239a.pdf talking the dropout voltage confusion..

[jorgemef]

I had to change the feedback network on one of the sets to have headroom to compensate with coper wire and have adjustment floor to the 10v.
The thing went from +10uv/C to -50uV/C so the resistors are not that great.
Anyway with two coper wire resistors I made the thing come down to 3uV/C. I think I will call it done now.
I added the coper wire resistors in the other side of the board where the feedback resistor network are with thermal glue to have proper thermal coupling.

Also some picture of the heat distribuition on the board with Vin=20v.

Cheers,
Jorge

[ivo]

1/2W 207. As per seller specs is about +-10ppm/C. https://pt.aliexpress.com/item/1005006320806775.html

So you mean the circuit is performing as per component specs, and I need to look for strategy to reduce the circuit TC like adding copper resistance on the lower side of the feedback network (and other resistor on the upper side to maintain the 10v)?

You can't expect cheap good resistors from aliexpress! They are probably gathering the lowest-spec rated TC parts, not the best (yes the best is 10ppm rated, but which rated parts did they decide to sell? If they don't tell you...). But it is not impossible to get okay parts.

https://www.lcsc.com/product-detail/Chip-Resistor-Surface-Mount_Milliohm-HoAR0805-1-10W-10KR-0-1-TCR5_C2912582.html single 10k resistors rated at 5ppm.

Even better is a Vishay TOMC16031002BUF which is 8 resistors for ~$6 all thermally matched in a single part, divide and combine it how you like for a divider. 10k & 22k is a good start for 6.9V->10V noninverting opamp. Include one or two if you have an order from a western supplier you need to make.

[Kleinstein]

Poor quality resistors not only have a problem with the TC but also with long term drift and noise.  For a reference source the long term drift is the more serious issue. Even with high grade parts it is hard to get specs on the long term drift and already the possibilty for drift is an issue, as one would not easily know if they drift.
The TC part is only the point that is easy to measure and compare.  A resistor array would really be the more sensible choice.