I just bought out the last 4 LM329B's at Digikey, and they have 36 LM329A's left. Octopart shows no more available from the usual reliable sources.
The LM329 is at the heart of the GFA-565; They are essential parts for my BFA-565 circuit board kits, and I'm not sure what to replace it with! I have a couple ideas I would love some feedback on.
In a nutshell, the LM329 in the GFA-565 is doing double-duty. It's set up as a simple shunt regulator, with two 6.9V references in series for 13.8V output.
1. It supplies +/-13.8V power to the servo op-amp, consuming about 0.5ma per rail.
2. It supplies +/- 13.8V base reference voltage to the cascodes.
Schematic:
What are the important parameters here?
I suppose thermal coefficient is not really so important in this application. The servo can keep on top of the drift.
I have two ideas, both based on publications by Walt Jung himself, who incidentally, designed the GFA-565 circuit, so if one of these two solutions works out, it would be true to form.
1. Walt wrote an app note in 2019 describing the use of two 1N5234B zener diodes in series, with one forward and one reverse, for a combined Vf of 6.8V and a noise density even better than the LM329, around 20nV/Hz. One concern I can imagine, is the dynamic impedance of the zener is about 7ohms, so am I correct that the output impedance would be 7 ohms in series with impedance of the second, 0.6V forward-biased diode? Is it also 7 ohms in the forward-biased diode?
2. The "PM329" Zener-Based 6.9V Shunt Voltage Reference
This is a really interesting option to me, and it seems it would easily exceed requirements. There is a 13V version of this diode, the 1N5243B, which would make this a 13.6V reference, and so it would not need two references in series as OEM.
The first option would be simple and dirt cheap, two diodes in series, and if it's truly as good as the LM329 in this application, then great! But I'll do the PM329 discrete solution if that's best... I'm not sure I know what all the most important parameters to consider here, opinions requested...
The LM329 is at the heart of the GFA-565; They are essential parts for my BFA-565 circuit board kits, and I'm not sure what to replace it with! I have a couple ideas I would love some feedback on.
In a nutshell, the LM329 in the GFA-565 is doing double-duty. It's set up as a simple shunt regulator, with two 6.9V references in series for 13.8V output.
1. It supplies +/-13.8V power to the servo op-amp, consuming about 0.5ma per rail.
2. It supplies +/- 13.8V base reference voltage to the cascodes.
Schematic:
What are the important parameters here?
- Low noise obviously. The LM329 is hard to beat in that respect.
- Dynamic impedance: Does this matter to the bases of the cascodes? The LM329 is around 1 ohm dynamic impedance.
I suppose thermal coefficient is not really so important in this application. The servo can keep on top of the drift.
I have two ideas, both based on publications by Walt Jung himself, who incidentally, designed the GFA-565 circuit, so if one of these two solutions works out, it would be true to form.
1. Walt wrote an app note in 2019 describing the use of two 1N5234B zener diodes in series, with one forward and one reverse, for a combined Vf of 6.8V and a noise density even better than the LM329, around 20nV/Hz. One concern I can imagine, is the dynamic impedance of the zener is about 7ohms, so am I correct that the output impedance would be 7 ohms in series with impedance of the second, 0.6V forward-biased diode? Is it also 7 ohms in the forward-biased diode?
2. The "PM329" Zener-Based 6.9V Shunt Voltage Reference
This is a really interesting option to me, and it seems it would easily exceed requirements. There is a 13V version of this diode, the 1N5243B, which would make this a 13.6V reference, and so it would not need two references in series as OEM.
The first option would be simple and dirt cheap, two diodes in series, and if it's truly as good as the LM329 in this application, then great! But I'll do the PM329 discrete solution if that's best... I'm not sure I know what all the most important parameters to consider here, opinions requested...
A TL431 is a very nice shunt regulator, and they can be very quiet, especially with some passive post-filtering. Their dynamic output impedance is low, around 0.22Ω, and relatively flat too, making it easy to get high, wideband PSRR, especially fi you can drop some voltage across a series resistor. They're made by more than a few companies too, so supply issues won't be too daunting.
https://www.mouser.com/c/semiconductors/power-management-ics/voltage-references/?q=LM329
Only 2000 in stock, more on order but a year away.
Only 2000 in stock, more on order but a year away.
Oh! Sheesh, the way I searched for them on Octopart, they didn't show in the results till the second page.
Well, it's less urgent then, but I still want to come up with a plan B.
According to https://www.analog.com/en/products/lm329.html#product-overview some packaging options are out of production, but others are not and the LM329 is even suitable for new designs.
Gerhard, in the GLED thread ( https://www.diyaudio.com/community/threads/mr-jungs-ultra-low-noise-vref-the-gled431.285346 )discussed using two zeners in series:
"An NXP BZX84C2V7 or 3V3 is 6 dB above 1nV/sqrt(Hz) or 2nV/sqrtHz, just pick it
from the tape. And the 1/f corner is at 10 Hz."
I had purchased a pack of 100 LM329DZ from DK, now have only a few left. I think they were $0.25 apiece! The data from the ADI data sheet on their version shows the noise to be the same.
"An NXP BZX84C2V7 or 3V3 is 6 dB above 1nV/sqrt(Hz) or 2nV/sqrtHz, just pick it
from the tape. And the 1/f corner is at 10 Hz."
I had purchased a pack of 100 LM329DZ from DK, now have only a few left. I think they were $0.25 apiece! The data from the ADI data sheet on their version shows the noise to be the same.