LM317: I need assistance identifying better replacements for LM317/337 series devices; needs below

Hello Grant,

For LM317, the only option I see is LM350 series. For LM337, the option I see is LM2991 in our portfolio using TO-220. Below is the comparison table, please let us know if you have any questions.

Thanks,

Prakash L

Hi Prakash, but I'm confused by a few things.  First, the PSRR my biggest concern, is roughly the same -- or am i misreading? 

next i see for the LM2991 i see a max Vout of -3V. or are you ignoring he magnitude and referring only to the sign. Somewhat odd but maybe that explain it.

Do i have better options if i accept that i cannot have TO-220/THT? 

Thanks,

Grant

Grant,

Yes, not a great improvement in PSRR. Also, the sign is the reason for the confusion like you mentioned. If you can live with a max Vin of 26V, then I suggest that you try LM241C and LM2941 devices. If that doesn't work, then your best bet is TPS7A47 series for positive and TPS7A33 series for negative with 20 pins. Hope this helps, we are working on devices that might fit this portfolio in the future. However, don't think they are planned in anything less than 8 pins.

Thanks,

Prakash L

thank you helpful, and your follow-up (below) is even more helpful (not entirely what i hoped to hear, but helpful!) :-)

Strangely i cannot "reply" to that one.

One last request since replacements are apparently complicated:  can you provide links to any documents (aside from the datasheet) on how to get better performance out of the 317/337 pairs?

TIA,

Grant

Grant,

Checking with our applications team, will get back to you as soon as I hear from them.

Thanks,

Prakash L

Thanks Prakash!   My biggest concern, since LC will effectively filter much hash, is power supply ripple and its overtones, so 120, 240, 480 Hz....  the nice thing about improving the 317/37 performance is that they are so widely available, and in essentially matched complements.

Grant

Grant,

I am assigning this to our apps team. Also, they would like to know if you could share your circuit schematics or any other details.

Thanks,

Prakash L

Thank you for asking!  I appreciate that someone is taking this seriously.

Since I am developing a family of modules and building blocks, being specific is difficult; there is no one final schematic. But i will try. Note i listed some of the most important characteristics in my initial request.

in all cases the circuits are driven by a common raw supply, shielded from the low-noise circuitry.  That has independent, PCB mounted transformers for digital (noisy) and analog (clean) circuits, with independent grounds. The grounds are linked by 220 ohms in series with 2mH to tie the grounds to similar potential, but isolate noise.  This is also common practice in all mixed chips (including your own DACs).

The raw supplies have the following topology, sized per the load at minimum 200% sizing:

1. transformer
2. bridge rectifier
3. RC with ~ 5-50 ohms and 1000 - 2200 uF

Followed by a common regulator
4. Capacitance at input to regulators - 4700-2200, , with 0.1 film bypass
5. LM317/337 with 10-100uF filtering the adj pin; heatsink as needed for <150F (not C), R1 220-500 ohms
6. 1000-2200uF immediately after with ~0.47 film bypass
7. if possible, series inductor < 1 ohm o control hash/noise

Followed by a set of building blocks in different combinations, each with its own local electrolytic filtering and film bypass on each rail

nearly everything is split-symmetrical rails (e.g.: minus and plus 15V)

Digital supplies are not an issue

Total draw never exceeds 350mA; regulators are 1.5A rated

+ rail alwasy draws more due to presence of buffers from +5V to ground and +15V to ground

Loads  primarily operate class-A , but some have variable draw

Gain of circuits varies greatly from V0/Vin <1 to Vo/Vin > 1500X (64-65 dB gain)

It is primarily the circuits with 65 dB on gain that are presenting the challenge, as i presume you appreciate.

Clearly my PSRR is not as good as i hoped, btu the low noise front ends are all differential and ought to have at least an extra 12-18dB PSRR

These are used in professional instrumentation and audio signal processing, or so they are intended. Right now they are hum generators. with < 46 dB overall SNR at 1500X gain

I am looking at quieter/less sensitive circuit typologies but i have also tested that a "perfect" power supply eliminates the problem. Maybe i mis-read the datasheet but the 317 claims something like 70 dB ripple rejection. I am seeing less than 45 dB, sometimes much less. Its the difference between awesome and flawed.

TIA

Hello,

         I am looking into this, I will get back to you on Tuesday (09/07, after the long weekend) at the latest. Thanks!

Regards,

Srikanth

Hi Grant,

               Regarding the PSRR, may I ask what your frequency range of interest is? Beyond 10KHz, it is difficult to get more than 50dB of PSRR from LM337. If the TPS7A47 and TPS7A33 pair does not work for you, could you consider TPS7A39? It has both, positive and negative LDO in a package. It satisfies your input voltage and output voltage range and has very good PSRR performance over frequency. Perhaps the load current is a drawback. LM317/337 devices are fairly old devices, as a result we, unfortunately, do not have documentation on using them as a pair. Thanks!

Regards,

Srikanth 

Hi and thank you.  I will look at the pair of devices suggested.  My noise issues are primarily low frequency (60 Hz and several overtones) - as i noted, i can handle higher frequencies with an LC segment more effectively.  TIA,  Grant

Hello,

         Hope the devices (TPS7A47-TPS7A33/TPS7A39) will work out for you. Thanks!

Regards,

Srikanth

One is not available in a THT part -- making evaluation and testing difficult especially with existing prototype boards... so while they may be useful in future production, i am still looking for a better solution, and the advice on how to best achieve my goals with the known working parts, the old, obsolete, 317/337 (which really need a like for like improvement!)

The dual part, TPS7A39, looks very interesting.  It does not have sufficient current for the entire +/- 15V draw, but could work well on a single board.  Is here any kind of adapter PCB that would allow me to test a couple units on existing, THT, hand-stuffed boards?  There's no way i can hand add that tiny, multi-pin part.  TIA, Grant

is that this:?

https://www.digikey.com/en/products/detail/texas-instruments/TPS7A39EVM-865/8106159

Which is pretty much what i'm looking for - although it seems pretty complicated and (ok, its relative) costly for what I need.  I just found this references in the documentation.  Are you saying you have similar for the other parts too?  Remember my problem: i need to test these before i use them.  I cant do it by hand. Idealy i'd like to do early production with larger THT, but compromises have to be made ( in the TI product line anyway, less so in others i'll note)

In full production SMT is no only OK, its preferred.  But i'm not there yet.

Grant

Do you offer them as samples?

Also, what is the efficacy of using two three-terminal regulators (78xx, 79xx, 317, 337...) in tandem.  Is the ripple reduction and PSRR effectively multiplicative (which, if the reference is incoming noise, it ought to be)?

i realize the inherent/internal noise contribution may not be improved much if at all.

TIA