Voltage regulator prefixes and capacitor source questions

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I've been a kit builder for decades (Dynakit, Hafler, Heath, and Audio Concepts) and hae managed to successfully design and build a few sets of speakers from scratch using Vance Dickason's LSCB. However, I don't have much actual electrical theory knowledge. I can read basic schematics.

I've had an Audio Alchemy Dac-in-the-Box lying around for years with a burnt out power supply. I couldn't find a replacement PS for it except for Alchemy's big buck "upgrades," and those were out of the question. Yesterday, with the help of a great guy at a local electronic supply place, we designed a regulated power supply for it, and I successfully built it from scratch. Felt great to resurrect a dead piece of gear; the DITB sounds better than ever.

Which brings me to my questions. I came across the attached article by Kalman Rubinson suggesting some cheap easy improvements for the DITB and power supply that I was thinking of trying, but he's suggesting using "AN78XX" and "AN79XX" regulators, but all I can find on Digikey or Mouser are LM prefixes. I used an LM7912 and an MC7812 to regulate my PS. What do the prefixes mean? Are they important?

Second, the article mentions using specific Nichicon Muse and Panasonic PP or V series caps. I can't seem to find those on Mouser or Digikey either, but I confess I'm overwhelmed with all of the data. Their word search boxes invariably return "No results," and if I drill down using their regular menus, I get so many options that I'm not sure which one is the specific cap mentioned.

I KNOW these are probably pretty stupid questions, but I could use some help. Thanks in advance to anyone who can spare a few minutes for me..

Best wishes,

Barry
 
Jack and Jon,

Thank you both for your replies. With your help, I've managed to find some AN (Panasonic!) regulators on the digikey site, but they don't stock them. PLUS, all of them are listed as "AN7XLXX" rather than "AN7XXX," which I gather means they're only acceptable for up to 100ma output, correct? Does it matter when used as Kalman suggested in the DITB mod? BTW.. I realize I forgot to attach it to my first post; I've attached it this time.

I did find the page w/ all of the Nichicon caps; thanks for the link. I WAS able to locate the right ones on that page, I believe. Thanks, also, for the suggestion of the Elna Silmics. If I'm looking at the correct equivalent Elna caps, they cost a LOT more than the Nichicon FG or KZ series caps. Here's the link to the page I'm looking at..

Aluminum Capacitors | Capacitors | DigiKey

Elna is $2.52 for a 330uf 50v while a Nichicon KZ is $0.73 or FG is $0.53? I thought the Elnas were a high quality yet budget alternative to the "Muse" caps. Here's the Nichicon page, just in case I'm looking at the wrong thing.

Aluminum Capacitors | Capacitors | DigiKey

Not that it really matters; I only need a few of them and the money isn't significant. I just want to make sure I get the right stuff and don't ruin my DAC. :)

Thanks again, very much, for your help.

Barry
 

Attachments

  • DITBMOD(1).pdf
    129.1 KB · Views: 28
Last edited:
The 78xx and 79xx regulators were invented by Fairchild. Why not use those? There prefix BTW was microA, greek mu being a character not even available to computers except through some weird alt sequence. Besides LM which is now TI, the other first line IC manufacturer is ST which use to be RCA but is now French something.
Electrolytic capacitors, I use the long service life grade, not pricey "audio grade" which have sucker written all over the specs. I haven't bought any bad sounding ones using that as the figure of merit. Panasonic, Nichicon, Rubicon all make great e-caps, also some short life ones for people looking to save $.20 If those are out of stock I've used United Chemicon CDE and mulitcomp (farnell house brand) but don't expect them to last as long. the last three have a looser specified end of life test (read datasheet) than the first three.
In non-electrolytic caps, polyprophylene is probably worth a 40% premium versus mylar or polyester film . In restoring old circuits, paper caps really do sound different than polyester. Those might be worth an "audio" premium". Mylar made my PAS2 excessively trebly.
In resistors, metal film really does hiss less than carbon comp especially 100kohm up. I redone my dynakits with them. But radio circuits might require low inductance carbon comp, they are still available.
 
Last edited:
Hey Hoosier -- I think you're conservative in putting polypropylene ahead of polyester (mylar).

A higher voltage rating is preferable to lower voltage, but you don't have to go overboard.

Not to beat the horse too much, but the Linear Low Dropout LT1963A/LT3015 regs sounded much better (to the ears of the NJ Audio Society panel) than the LM317/337. They are much more expensive, however.
 
LM317 is an adjustable Vreg, the OP is looking at 7812 and 7912.

You can pick an LM7812 or MC7812 or most any other brand, and they will all be more or less the same. Don't go nuts looking for the exact brand.

7812 is rated to at least 1 amp. 78M12 is rated to at least 500ma. The little LM78L12 is rated to 100ma. You chose the one with the current rating you need.
 
.......
Not to beat the horse too much, but the Linear Low Dropout LT1963A/LT3015 regs sounded much better (to the ears of the NJ Audio Society panel) than the LM317/337. They are much more expensive, however.

May I ask a couple of questions about your regulator test in Linear Audio Vol. 4 ?

1) You described the test being carried out using HP/Tektronics power supply as power source at 25V. However, LT1963A data sheet mentions maximum input voltage at 20V. So did you actually lower the input voltage for the LT1963A, or did you use some form of pre-reg upstream ?

2) In your listening test, in which the LT1963A/3015 excelled over the LM317/337, were you also using the HP/Tektronics PS as voltage source, or were you using transformers and rectifiers (as in a normal audio equipment) instead ?

3) Could you share data about the noise spectrum and output impedance of the power source upstream of the regulators in the listening test ?

Many thanks in advance.


Best regards,
Patrick
 
Second, the article mentions using specific Nichicon Muse and Panasonic PP or V series caps.

If there is "input cap", replacing with a bipolar (BP) one is good, unless you have used non-electrolytic for that.

So the upgrade is about using a better opamp? What is the cost of the upgrade including the opamp? Yes, not many opamps can work with +/-5V but LM4565 can but is "expensive", depends on your upgrade budget.

BTW, if supply voltage is determined by 7805/7905, upgrade can usually be done by only replacing the chip with 7806/7906 or the 8V version, depends on the circuit.

The 78xx and 79xx regulators were invented by Fairchild. Why not use those?

AN is considered by many as better sounding. I myself don't use any of them, except for the small current one (TO-92).

but the Linear Low Dropout LT1963A/LT3015 regs sounded much better (to the ears of the NJ Audio Society panel) than the LM317/337.

I don't doubt that. I'm just curious how they determine "sounding better". LM317/337 is very noisy, hence it is worse if noise is the criteria...

I have always considered LM317/337 a "not complete" design. They are very noisy, but do something "right" to the sound (I guess Zout?)...

78/79XX is not acceptable to me (to power the analog section) but LM317/337 is okay when used 2 or 3 in series. For analog, I have used 2 in series where the first one is it's discrete version.
 
Last edited:
Plus the LT317A I found in stock were all surface mount. You or your employer has to pay the royalties to Microsoft to use surface mount. Biannually to stay up to date. I can't get linux layout software to work properly and the part libraries are poor on the schematic maker. Plus a $45 soldering iron is not going to make it for surface mount. This is diyaudio, I'm glad everybody else has wads of cash to burn. I'm using used computers where microsoft destroys any "free" operating system with unauthorized updates three weeks fater an op system hits the internet - even one with a sticker op system number. No source CD, no windows. Linux is free, but *****y for making PC boards.
I don't see regulator noise as a huge problem, just use a bigger bypass cap and maybe a series inductor after that - used out of a dead PCAT power supply. I'm using zeners as regulators for op amps, and my hiss is lower than the gas pilot light in the room. The inductor is handy for keeping radio noise out of your preamp/mixer, too.
 
May I ask a couple of questions about your regulator test in Linear Audio Vol. 4 ?

1) You described the test being carried out using HP/Tektronics power supply as power source at 25V. However, LT1963A data sheet mentions maximum input voltage at 20V. So did you actually lower the input voltage for the LT1963A, or did you use some form of pre-reg upstream ?

2) In your listening test, in which the LT1963A/3015 excelled over the LM317/337, were you also using the HP/Tektronics PS as voltage source, or were you using transformers and rectifiers (as in a normal audio equipment) instead ?

3) Could you share data about the noise spectrum and output impedance of the power source upstream of the regulators in the listening test ?

Many thanks in advance.


Best regards,
Patrick

For the PSRR and Zout tests I used the Tek PS5010. These supplied Walt's "rail driver" as described in TAA in 1995. http://waltjung.org/PDFs/Regs_for_High_Perf_Audio_1.pdf This is 25V in and 18V out.

For the noise test I used a pair of 12V SLA's in series.

The LT1963A did not blow up in this case and I did not ask if if it were happy.

The listening test used a conventional torroid ps for all the regulators.

The results were not remarkably different from Walt's findings, using the LM317 as a "sanity test".

I didn't run any noise or impedance spectra on the sources. From my personal experience the PS5010 is much noisier than the HP
 
so "sounding better" may well be marginal stability in action

I have been critically very curious about this typical issue. The question (rhetorical) is: how exactly "marginal stability" (this can be replaced with similar variable) can improve anything??

My finding is that it is just some kind of side effect. I mean, when we increase "stability" (can be replaced with other variables), we introduce something else that degrades sound. It is not the stability that takes away the goods...

(The good thing is, I can make use of this "knowledge", at least to my own satisfaction, because I seem the only one who can hear it :D)
 
Well, yes...if you over-damp the error amplifier ...
Low drop regulators act as adjustable current sources and their feedback adjusts current to give the right voltage, giving a low dc output impedance. At high frequencies the output impedance is low due to the capacitor. Trouble happens in the middle, wrong capacitor value can cause a "LC resonance" of high impedance or even a negative resistance
 
May I ask a couple of questions about your regulator test in Linear Audio Vol. 4 ?

1) You described the test being carried out using HP/Tektronics power supply as power source at 25V. However, LT1963A data sheet mentions maximum input voltage at 20V. So did you actually lower the input voltage for the LT1963A, or did you use some form of pre-reg upstream ?

2) In your listening test, in which the LT1963A/3015 excelled over the LM317/337, were you also using the HP/Tektronics PS as voltage source, or were you using transformers and rectifiers (as in a normal audio equipment) instead ?

3) Could you share data about the noise spectrum and output impedance of the power source upstream of the regulators in the listening test ?

Many thanks in advance.


Best regards,
Patrick

Here are the devices I used -- the preamplifier for the listening test was Erno's "All FET" which has a PSRR of only 40dB, very low distortion and noise. The regulators were all housed in their own aluminum chassis (Bud) and banana jacks and plugs inter-connecting.

The power supply driving the regulator for the listening test used a conventional Amveco torroid, diode bridge etc. For the measurements, the TEKPS5010 was used.

The rail driver as described by WJ is also pictured. I now have a couple of transformers which make PSRR/LR rejection easier -- from a few Hz to a few hundred kHz.
 

Attachments

  • PowerSupply.jpg
    PowerSupply.jpg
    75.3 KB · Views: 154
  • RailDriver.jpg
    RailDriver.jpg
    67.8 KB · Views: 158
  • Tek_HP_Supplies.jpg
    Tek_HP_Supplies.jpg
    113.2 KB · Views: 155
Low drop regulators act as adjustable current sources and their feedback adjusts current to give the right voltage, giving a low dc output impedance. At high frequencies the output impedance is low due to the capacitor. Trouble happens in the middle, wrong capacitor value can cause a "LC resonance" of high impedance or even a negative resistance

If you over-damp the regulator's error amplifier you get poor transient response.

I have a setup for measuring gain and phase out to 40MHz, but rarely is it necessary to go beyond 2MHz with the regulators and power supplies I have encountered. You can see where folks get in trouble with tantalum and ceramics caps stuck onto the delivery end of a power supply.

It would be helpful to have this setup for HV supplies for tube preamps.
 
It is difficult to argue about a subjective test result. So I accept it as is.

But for the F5 Headamp circuit, which is very similar to character (including PSRR) to the Borbely JFET lineamp you used, we have the following observations.
It is perhaps worth noting that the total bias is much higher at 160mA per channel.
Rail voltages are +/-15V.

1) The circuit is totally unstable when fed by a Salas shunt.
2) 7815_7915 & 317_337 are fine.
3) 317_337 baed Nazar reg is also fine.
4) 317_337 followed by BJT cap multipliers, as well as ZD referenced cap multipliers, are subjectively preferred.

I also simulated in Spice the circuit distortion with 3 different supplies :
a) an ideal voltage source in Spice
b) a Spice voltage source with 0.2R internal reistance, followed by a 330uF electrolytic cap
c) an ideal voltage source in Spice, followed by the above mentioned cap multiplier.

At close to maximum class A output, the 3 PS have essentially no effect on output distortion to within 1%.

Yes, I know Spice is not real, and does not say anything about sound.
On the other hand, a noisy supply always has as disadvantage for headphones as it disturbs the listening.
So it is perhaps not easy to have a direct comparison as you did for LA Vol.4.


Just for your interest,
Patrick
 
Last edited:
Patrick -- I have the Salas shunt regs around the lab somewhere and the parts for the F5HA so I will put in on the Bode100 and see from whence the issue might arise.

One of the difficulties with Salas is that there really is no "reference design" -- it always seems to be a work in progress.

What music tracks did you use to compare?
 
It's a whole bunch of people listening to different music with different regulators for over a long period of time.
Some of them reported their preference at the F5 Headamp thread.
More detailed descriptions are only available at our private forum.

Here some of the comments :
http://www.diyaudio.com/forums/pass-labs/271926-f5-headamp.html#post4288998
http://www.diyaudio.com/forums/pass-labs/271926-f5-headamp-3.html#post4337967
http://www.diyaudio.com/forums/pass-labs/271926-f5-headamp-5.html#post4374209
http://www.diyaudio.com/forums/pass-labs/271926-f5-headamp-6.html#post4389697
http://www.diyaudio.com/forums/pass-labs/271926-f5-headamp-5.html#post4375137

"When my 2nd F5-HA was powered by a bench switch mode power supply, all the magic was gone and it wasn't enjoyable to listen for long period at all.
So I put a CRC power supply in between with a corner frequency of 15Hz to clean the power a bit, most of the magic is back :)
With the LM317/337+CM PS it sounds the best."

We have made boards for a LT1963A_3015 CM, with a 7818 upstream of the 1963A.
We also intend to test it with my TO220 version of the Didden Regs (that you also measured a couple of years ago).
Might take a while though.

And of course most interested in what your own findings would be.


Patrick
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.