Hello,
I have several LM399 running at 10V and 7V. What's the best way to make it run at 5V please?
10V with precision INA105 1/2 gain?
Thanks
Alex
The LM399 give out some 7 V. So it would need some precision divider with a factor of about 0.71. There is no need to go up to 10 V first. However this is not a very common divider ratio. There may be approximate values from some resistor arrays.
The best way is to mutliply by 3/4 by an LT1043. And use 5.2V instead 5V as your reference..
Depending on the application 2/3 or 7 V could also be a good option.
The LM399 give out some 7 V. So it would need some precision divider with a factor of about 0.71. There is no need to go up to 10 V first. However this is not a very common divider ratio. There may be approximate values from some resistor arrays.
Indeed, 0.71 is a very peculiar factor.
I suppose two resistors 200k and 499k can do a voltage divider with necessary ratio ~ 5V
LM399 * 3/4 = 5.2V
did you ever test the cirquit in real life?
Up to now I have only simulations in LTSPICE.
Nope, sim only.
Btw, in the simulation there are ripples of about 40uVpp (it is to expect), thus additional filtering would be nice to have. With 1uF the ripple is 50uVpp.
PS: the "Branadic's" ripple is lower, around 17uVpp..
Attached picture shows my way of getting nominal 5V.
I mentioned it in a separate post previously.
https://www.eevblog.com/forum/metrology/lm399-based-2-5v-reference/msg2319243/#msg2319243
The tempco is roughly 1ppm/K. Long term stability not tested.
1ppm tempco is a very good result in my view. I would really like to see the long-term stability.
You are using two dual opams - LT1112. Why not using quad opamp LT1114 and terminating the forth one?
Thanks
Alex
EDIT: You have a 50k pot. Finding a pot with a good temco is hard and they are much more expensive and only give 15ppm at best. What's R3 value please?
The tempco itself of the 50k 10turn pot is not important imho, the matching tempco between the 2 portions of the pot does..
What could be a matching tempco of such A and B parts of the pot?
The tempco itself of the 50k 10turn pot is not important imho, the matching tempco between the 2 portions of the pot does..
What could be a matching tempco of such A and B parts of the pot?
The datasheet does not say. They just say about tempco for the pot.
This is prudent to assume that it would be not less than pot's tempco.
http://www.vishaypg.com/docs/63056/12801285.pdf
It would be great if experts here would be kind enough and measure the matching tempco of, for example, a Vishay 100ppm 50k 10t pot trimmer in say 25k/25k and 5k/45k wiper positions, for example..
The pot's setting only has a reduced effect due to the resistors (10 K and 1 M). So the pots TK would also have a reduced effect, like some 1% of the pot's TK. So a 100 ppm/K (divider ratio) pot could contribute some 1ppm/K to the output. The ratio TK is often not that bad with pots.
Btw, in the simulation there are ripples of about 40uVpp (it is to expect), thus additional filtering would be nice to have. With 1uF the ripple is 50uVpp.
PS: the "Branadic's" ripple is lower, around 17uVpp..
Hello,
I think the higher ripple has something to do with your (unbuffered) LM399 model.
I would try a 1uF capacitor in parallel to the LM399.
with best regards
Andreas
It comes from switching inside the LTC1043, see below. The red trace is the signal at the LTC1043 clock capacitor C1 (Cosc pin).
An ideal 6.9V source used instead the LM399.
Hello,
now the large negative spikes from previous simulation have gone.
-> we probably need some buffer or capacitor for a real world cirquit.
The rest looks like charge injection.
I always have the suspect that the charge injection cancellation cirquit
works in the "wrong" direction for the 2:1 divider cirquit from the data sheet.
At least I get different gains when exchanging input and output ports for the 2:1 divider.
(measured something like -9 ppm with datasheet pinning against -2ppm with reversed pinning)
So it would be interesting what happens if we exchange S1 against S2 and S3 against S4 pin.
By the way: if you use 10 nF for both OSC-Pins you get about the double switching frequency (1 kHz)
against the typical LTC1043 400-500 Hz frequency. That is why I use 20 nF in my simulation.
with best regards
Andreas
Looks symmetrical in sim. Mirrored Sx pins. 22nF clock cap.
PS: added for reference divide by 2 (3.45V out). Ripple is 6.67uVpp.
PPS: the charge injection for LTC1043 at 15Vcc and Vin=7V is 3.5pCb (DS).
With a single switch involved it should create a voltage pulse on an 1uF capacitor
deltaV = 3.5e-12/1e-6 = 3.5uV
With above schematics we have several switches involved.
Attached picture shows my way of getting nominal 5V.
I mentioned it in a separate post previously.
https://www.eevblog.com/forum/metrology/lm399-based-2-5v-reference/msg2319243/#msg2319243
The tempco is roughly 1ppm/K. Long term stability not tested.
Interesting.
You have managed to get 2.5V reference from LM399 - I saw it in another post.
Would you please post the full schematic - LM399 (~7V) - 5V - 2.5V?
There isn't much secret. You can get 5V from the schematic I posted previously, and use another LT5400 to divide 5V to 2.5V and buffer it with a precision opamp such as OPA2277, AD8676, etc. For best stability, remove the trimming pot.