best sounding 78XX/79XX drop in Regulator Replacements?

[fmak]

There is no debate there, but is there a difference between 20 and 24 bit sonics? There are not many DACs that can achieve 120dB S/N in any topology, even harder to maintain this through the entire signal chain.

------------------------------------------------------------------------
Yes, simply use digital processors such as the dCS972/974 or Meridian 580 and you can instantly tell differences in 16, 20, and 24 bit. Also dither levels and type.

 

[tiroth]

fmak, what you dither your OS/ASRC filter results down to is totally beside the point. You may need the 24 bits to avoid truncation errors. This does not mean that your DAC is producing 24 significant bits! This has been explained numerous times.

I thought we were discussing noise levels of actual analog circuitry.

 

[jewilson]

Just to get a 18bit resolution is an accomplishment. If you have any noise on the supply you will never get lest significant bits. As Fred said, PPRR on a DAC is not so good.

So it can be said that the supply of a DAC is just as important as the converter and more important than the OPAMP's.

 

[carlosfm]

Konnichiwa,
Maybe you know an excellent choice of 8-Pin DIL Case Op-Amp's that improve on the OPA2604 as well (I know the LM6172 but it may be too twitchy for the job), lowish cost preferred?
Sayonara

Kuei, I know you don't like it, but give another try to the OPA2132 or, cheaper (and probably the same, but I haven't tried it) the OPA2134.
It's a little more picky about bypassing, it really likes capacitance.
Between 47uf~100uf bypassed with a small 0.1uf ceramic (I put these inside the socket ) will do.
This way, it sounds a great deal better that the OPA2604.
It more precise in the midband an treble, and a more dynamic sound (it has a gorgeous bass, better that the 2604).
It's my favourite double op-amp, side to side with the LM6172.
But it only works well in these conditions.
Otherwise it sounds bad.

 

[ThorstenL]

Re: The numbers racket

Konnichiwa,

First of all I don't know what "get DAC and Op-Amp noiseless and retain the Op-Amp's PSRR as original" means.

Op Amps and DACs have finite PSRR which usually become lower as frequency increases and is lower than you might think.

I am quite aware of that. My original earlier post listed those nombers for the given Op-Amp, namely around 80db @ 20KHz. The Regulator has 500uV output noise. Reject this by at least 80db and in the Audio range the noise contribution by the Powersupply is 50nV and hence -155db below 2V RMS.

Noise modulation on digital supplies results in jitter.

Absolutely. However fmak rejected the regs based on certain numbers in the datasheet which when applied where Peter applied them and where I intend to apply them the numbers lack relevance.

The desirability for low noise on the supply rails seems pretty straight forward.

There is nothing wrong, per se with low noise. However, a few 100uV RMS noise from the regulator is usually swamped by dozends of mV supply line noise from (poor) decoupling. I can imagine points where the noise may become the limiting factor, however they are well past anything I am interested in.

I am trying to figure out where the 50 uV number come from....
broadband noise for a typical three terminal regulator?

You mean the 0.05uV? See above.

I really don't know where the -155dB number comes from...........

20 * LOG(0.05/2000000) = -152db, I did the original calculation quickly and on paper and rounded....

0.05uV noise compared to 2000000uV Signal at full scale.

-160dB in reference to 2 volts is 10 x 10^-9 or 10 nV .

Nope, it is not. It is 20nV, actually.

Doubling this for -154 dB is 20nV.

Nope, it ain't, it's 40nV or 0.05uV.

This is way below the noise floor of any audio equipment know to me.

My point exactly. Supplying a suitable Op-Amp in the I/V stage of a DAC fo the regulator branded as noisy and poor performance (despite having better ripple rejection than 78xx/79xx and despite much lower output impedance and better transient behaviour than these as well) would contribute only a very small amount of noise to the output noise of the Op-Amp, so little that it will not materially impact on a 24-Bit resultion, if the other noise sources did not exist.

Hence fmak's logic "noisy regulator - looses last bit resolution, this is why it sounds different" is shown to be ad absurdum and I conclude that in the right place and correctly applied the discussed regulators are good upgrades for production gear, IN PRODUCTION.

I find it amusing that you push all this numbers at us without any context of what they are in reference to,

The context was in an earlier post.

Sayonara

 

[Fred Dieckmann]

"You mean the 0.05uV? See above."

Actually I meant 50 nV, sorry about unit slip. I did my noise from 1 V, I plead guilty with the provision that I have been setting up my sound card for noise measurements based around 1 V p-p for 0 dB. I still believe that picking regulator base just on specs might not be total story. For the amount of RF in the raw supply, I would RC filter the input to the regulator. For a D to A converter a lot of the supply currents in the RF range and the contributions to low source impedance are completely from the output capacitor. The regulator output cap's performance at RF may be more important than in the audio range (Oscon anyone?)

I can get into a big discussion of 7800 regulators verses adjustable but it has been done to death. If you were going to stay with that type (78XX) regulator for cost, simplicity, what ever......... The NJMs have some of the better numbers for noise and PSRR and sound better than the several other vendors 78XX in my experience. Avoid any TO 92 type and stay with the TO 220. I would look at a 0.5 ohm resistance in series with the output cap to damp the resonance between the inductive output impedance of the regulator and the fairly high Q filter cap.

I still believe some context of where the numbers come from in the post where they are given is a real good idea. Most people don't want to wade through a bunch of post to see what you are talking about. Many people even provide a link to previous post that are relevant to the subject......

 

[ThorstenL]

Konnichiwa Diekman San,

I still believe that picking regulator base just on specs might not be total story.

It isn't, but as it so happnes, the answer from Peter was just what I was looking for. I am working currently on a "UK Edition" of certain gear, the deal is we send a bag of compenents to the factory and they fit them for us instead what they usually use. In order to allow for the use of certain Op-Amp's I need 12V instead of 15V regs for the Op-Amp analog stage. Now if we need to change anyway, why not change to something that works better?

Looks to me the LM's are just the ticket....

For the amount of RF in the raw supply, I would RC filter the input to the regulator. For a D to A converter a lot of the supply currents in the RF range and the contributions to low source impedance are completely from the output capacitor. The regulator output cap's performance at RF may be more important than in the audio range (Oscon anyone?)

All completely agreed. The point is I'm not designing, I'm POOGE'ing, with a fundamental constraint that I cannot change the PCB layout and that it must loads of bang, preferably no buck....



That is typhically not my playground. I'd love to redesign the whole units (I'm actually doing that too - but it amounts o a near complete re-build with Os-Con's all over the place, much improved supplies, C37 derived mechanical mods and on and on, you name it) but I'm severely constrained in what I can do an what not.

I can get into a big discussion of 7800 regulators verses adjustable but it has been done to death.

Exactly. IF I did design the whole box from scratch you can assume I'd be using semi-discrete regs with a Zenner/Emitter follower using TL431 and nice, large emitter geometry low noise transistors. That gets me at best 40db better noise levels as 78/79XX and a flat output impedance into the MHz range, plus super low noise shunts (much quieter than the normally just fine TL431) where it really matters (DAC's, Clock etc). But that's not an option here, it has to fit where currently a JRC 7815/7915 sits, with no further changes (other than capaictor type change).

Sayonara

 

[Bricolo]

Exactly. IF I did design the whole box from scratch you can assume I'd be using semi-discrete regs with a Zenner/Emitter follower using TL431 and nice, large emitter geometry low noise transistors. That gets me at best 40db better noise levels as 78/79XX and a flat output impedance into the MHz range, plus super low noise shunts (much quieter than the normally just fine TL431) where it really matters (DAC's, Clock etc). But that's not an option here, it has to fit where currently a JRC 7815/7915 sits, with no further changes (other than capaictor type change).

Designing a small PCB, with your zenner/EF reg, as 3 terminal reg (pin to pin compatible with a 78xx) can be done. Making this as small as a keyboard key seems possible.


Please, tell us more abou those low noise shunts!

 

[ThorstenL]

Konnichiwa,

Designing a small PCB, with your zenner/EF reg, as 3 terminal reg (pin to pin compatible with a 78xx) can be done. Making this as small as a keyboard key seems possible.

Abso-****ing-lutely possible. But budgetary pricing for LM2940 + LM2990 comes in at around $ 2 per set. I can't have the PCB's made for that, less have them stuffed and soldered.

Please, tell us more abou those low noise shunts!

Look at this:

http://www.wenzel.com/documents/finesse.html

This basic idea can extend to give excellent low noise levels and very low PSU node impedance designed into equipment, but it really is "design in" not "after market".

Sayonara

 

[Bricolo]

Konnichiwa,



Abso-****ing-lutely possible. But budgetary pricing for LM2940 + LM2990 comes in at around $ 2 per set. I can't have the PCB's made for that, less have them stuffed and soldered.



Look at this:

http://www.wenzel.com/documents/finesse.html

This basic idea can extend to give excellent low noise levels and very low PSU node impedance designed into equipment, but it really is "design in" not "after market".

Sayonara

It's not a reg, but it costed me less than $2

But it may take you too much time to do this, if you have dozons of regs to change

 

[fmak]

fmak, what you dither your OS/ASRC filter results down to is totally beside the point. You may need the 24 bits to avoid truncation errors. This does not mean that your DAC is producing 24 significant bits! This has been explained numerous times.

I thought we were discussing noise levels of actual analog circuitry.

---------------------------------------------------------------------------
No, Peter was talking about using the LM29xx for the TDA1543 and Tent clock. The former is perhaps 15 bit and the latter needs a lo-noise supply.

Someone else is talking about PSRR of opamps and dacs wrt supply. In extremis why use a regulator at all? Just design the whole thing to be fed from a clean supply and rely on PSRR and low cap impedeance, or use a series pass config.

There is a difference between attaining 20 bit resolution, achievable only in the best of dacs, and being fed with high bit depth streams which do make a difference to dac behaviour. Just look at Stereophile test data.

 

[jean-paul]

Abso-****ing-lutely possible. But budgetary pricing for LM2940 + LM2990 comes in at around $ 2 per set. I can't have the PCB's made for that, less have them stuffed and soldered.

You don't want to use LM2940, you want to use LM2937. Better and less noisy. 500 mA max. though.

So you are commercially working on a UK edition of a certain device and you are asking for info on a DIY forum !?!? I don't have a problem with that but I can imagine the company that gave you this order, well you know...

Please note that this all can start a hype/fashion of using noisy regulators where most of us changed them for low noise/low imp. types in the past years I experimented with a lot of regulators and came back to 78xx/79xx of a certain brand with good sonical results FWIW. Something I would not have believed when I started to test different types. Although a premature conclusion it may sound, till now I haven't had the best experiences with low drop regulators in audio. Still a discrete built simple TL431 + transistor regulator wins from low drop regulators. Again sonically, the best measuring regulators often offer dull sound. The results of using different caps after whatever regulator is important as, for example, good regulators with an OSCON can sound worse than with a normal electrolytic. I agree with Peter that a small electrolytic cap at a short distance from the regulator gives best results. I use standard series 037/038 from BCcomponents without exception at that location.

Reminds me of the LT regulator hype ( not wanting to start about inverting vs. noninverting )....

 

[fmak]

Please note that this all can start a hype/fashion of using noisy regulators where most of us changed them for low noise/low imp. types in the past years

Reminds me of the LT regulator hype.... [/B][/QUOTE]
----------------------------------------------------------------------------------
I agree, Jean-Paul.

Perhaps the noisy regulators will help dither digital signals!

I am saddened tho have learnt that cheap opamps and regulators are used in very expensive pooged commercial preamps. This highlighths the tradeoffs in component selection where not-so-good components can 'hide or disguise' the sonics in a particular set up

 

[ThorstenL]

Konnichiwa,

It's not a reg, but it costed me less than $2

I have done things just like that, in my case a TO220P BJT turned "upsidedown" and a little stripboard with SMD Cap's and resistors to make the Zenner (TL431) and RC filter to the base of the BJT.

However, this in not really an option to make and send to factory to fit.

But it may take you too much time to do this, if you have dozons of regs to change

Well, for the first batch of gear I would have to build a few 100 of the ******s, thanX but no thanX...



Sayonara

 

[ThorstenL]

Konnichiwa,

No, Peter was talking about using the LM29xx for the TDA1543 and Tent clock.

Not in this thread. Let's review the FACTS:

1) Peter wrote:

"Recently, in ML38 preamp, I replaced 78XX/79XX regulators with low dropout type from National, LM2937ET and LM2990T, and those are much better sounding than the other ones. This was also a recommended upgrade from 38 to 380 version."

Note that Peter write about a Mark Levinson preamplifier, not a DAC with TDA1543 or a Tent Clock.

2) You wrote:

"Who recommends the upgrade? The LM 2973ET is for automobile use and have poor noise and mediocre other characteristics."

3) I wrote:

"Hmmm. I have on previous occasions noticed that regulators that are "poor" on the Datasheet can sound rather good. Often "good" regulators have such a low output impedance that they swamp any bypass capacitor sonics and determine the sound, usually in a negative sense."

4) You wrote:

"Or may be one is loosing 1 or more bits thru, noise. Is the noise affecting the sound in some way?"

My contention was that when feeding Op-Amp stages or modern DAC's the "high noise level" of the discussed regulators (Nat Semi LM2937 & 2990) is not an issue in the sense in which you present it.

Would I use these regulators in a DIY DAC using a Tent Clock and a TDA1543? Not likely. But as replacement of generic 78XX/79XX regulators in commercial gear they look the piece, which is what I originally asked for.

Sayonara

 

[ThorstenL]

Konnichiwa,

You don't want to use LM2940, you want to use LM2937.

Maybe. I am comparing the two and feel like using then 2940, which is actually the complement to the 2990. The 2937 has no direct complement and is lower current. I'll have to try both.

Please note that pinout is different from 78xx/79xx regulators so they're not exact replacables.

Please NOTE that the pinout according to the Nat Semi datasheets is EXACTLY THE SAME as 78xx/79xx, so they ARE exactly replacable.

Still a discrete built simple TL431 + transistor regulator wins from low drop regulators.

I don't disagree. But I need something the factory can hand the people stuffing the PCB's with the instruction "use these instead of "brand X".

Again sonically, the best measuring regulators often offer dull sound. The results of using different caps after whatever regulator is important as, for example, good regulators with an OSCON can sound worse than with a normal electrolytic.

Absolutely.

Sayonara

 

[Bricolo]

My contention was that when feeding Op-Amp stages or modern DAC's the "high noise level" of the discussed regulators (Nat Semi LM2937 & 2990) is not an issue in the sense in which you present it.

Would I use these regulators in a DIY DAC using a Tent Clock and a TDA1543? Not likely. But as replacement of generic 78XX/79XX regulators in commercial gear they look the piece, which is what I originally asked for.

I don't clearly understand your point

Are you telling that a DAC has to have low noise power supplies (and opamps) in order to keep its full resolution (obvious), but that other stuff don't necessarely need low noise?

If your preamp has noisy regs, you'll also loose precision in the sound.
An audio rig with a, say 24 bit player, will need a player with 145dB SNR, but the preamp and amp will also need 145dB SNR in order to maintain the 24 bit resolution.

You can loose bits "after" the player...

 

[Peter Daniel]

---------------------------------------------------------------------------
No, Peter was talking about using the LM29xx for the TDA1543 and Tent clock. The former is perhaps 15 bit and the latter needs a lo-noise supply.

I never used LM29xx reg with TDA1543 or clock supply. To me AN800x sound the best in those applications. For comparison I tried 78xx of certain brand (that had the printing scraped off) and the AN800x was much better, IMO. I didn't compare it directly to TL431 + transistor, but from my previous experience with those regulators, it doesn't seem like AN800x sounds any worse than those, and it's so easy to implement and takes minimum space.

 

[carlosfm]

LM340Tx?

 

[Peter Daniel]

I don't know what it was as the print was gone

But I presume it was from National as the casing matched (there are small details on plastic case that are propriatory to each manufacturer)