Just one measurement so far -- a Fairchild LM317 WITHOUT its adjust pin bypassed, using CCIR filter measures 295uV -- with the adjust pin bypassed with a 100u/25V Panasonic this goes to 29.3uV. These numbers seem roughly consistent with the charts in AA 2/95. I gave the LM317 the benefit of the same 100u/25V Panasonic on the output since the Super Reg benefits from same.
The AD825 Super Reg (again CCIR filter) with no pre-regulator is coming in at 3.3uV, and the battery pack itself 1.5uV, this is about the residual of the Tektronix AA5001. The super-reg is bootstrapped and I put C5 across the zener as shown in the 2000 article.
The AD825 Super Reg (again CCIR filter) with no pre-regulator is coming in at 3.3uV, and the battery pack itself 1.5uV, this is about the residual of the Tektronix AA5001. The super-reg is bootstrapped and I put C5 across the zener as shown in the 2000 article.
I have some AD797's around -- it appears that the authors ditched the part as it was prone to some non-linear behavior in this application. When I made some tests with my HP3581, the AD797 was definitely quietest.
fwiw, a small tantalum cap on the adjust pin (6.8uF) had about the same effect of knocking down the noise as a 100uF electrolytic for the LM317 -- but tantalum's have this unfortunate problem of going short when they die.
fwiw, a small tantalum cap on the adjust pin (6.8uF) had about the same effect of knocking down the noise as a 100uF electrolytic for the LM317 -- but tantalum's have this unfortunate problem of going short when they die.
PA sent me a couple of his boards to check out. I also got another set of Old Colony Boards -- I have yet to run the impedance tests on PA's, but here are some noise results I ran last night:
PA Regulator -- Gold IC socket for error amp -- LM329 Reference:
P-A SR AD825-- on Aries AD797 LT1028
No filter 15.14Uv 5.67uV 6.95uV
A-Weighted 2.38 1.03 1.00
CCIR-486 6.40 2.04 2.02
CCIR-2K(Avg) 3.03 0.941 0.921
22-22K 3.52 2.37 1.70
Jung-Didden SR
No filter 28.60 8.75 19.96
A-Weighted 4.52 1.63 1.06
CCIR-486 12.50 4.20 2.91
CCIR-2k(Avg) 5.91 2.07 1.43
22-22k 5.19 1.77 1.29
The voltage source was a battery, and everything is contained in an Oreo cookie tin. The socket on PA's boards raised the noise just a hair, so you would probably best just soldering in the IC of your choice. I was pleasantly surprised by the performance of the LT1028, but note that the wideband noise is higher -- there is no evidence of oscillation. LT1028's and AD797's are selling at around the same price these days.
I only had WIMA 100nF 100V -- which are a bit of a tight fit -- on the board. 63V polyesters would probably work better.
PA Regulator -- Gold IC socket for error amp -- LM329 Reference:
P-A SR AD825-- on Aries AD797 LT1028
No filter 15.14Uv 5.67uV 6.95uV
A-Weighted 2.38 1.03 1.00
CCIR-486 6.40 2.04 2.02
CCIR-2K(Avg) 3.03 0.941 0.921
22-22K 3.52 2.37 1.70
Jung-Didden SR
No filter 28.60 8.75 19.96
A-Weighted 4.52 1.63 1.06
CCIR-486 12.50 4.20 2.91
CCIR-2k(Avg) 5.91 2.07 1.43
22-22k 5.19 1.77 1.29
The voltage source was a battery, and everything is contained in an Oreo cookie tin. The socket on PA's boards raised the noise just a hair, so you would probably best just soldering in the IC of your choice. I was pleasantly surprised by the performance of the LT1028, but note that the wideband noise is higher -- there is no evidence of oscillation. LT1028's and AD797's are selling at around the same price these days.
I only had WIMA 100nF 100V -- which are a bit of a tight fit -- on the board. 63V polyesters would probably work better.
You come with very good news and I really appreciate your work. Thanks. 
It feels good also that my boards not only looks good (according to myself ) but it's interesting to see that my 4-layer board is a tiny bit better than the Jung/Didden design but if you want to be really sure the same active parts should have been used.
It feels good also that my boards not only looks good (according to myself ) but it's interesting to see that my 4-layer board is a tiny bit better than the Jung/Didden design but if you want to be really sure the same active parts should have been used.
Last edited:
OK, cool.
See the attached below. It's from the Sabre ESS data sheets for 3.3V use on the analog section of their DAC chip. Pretty simple. 3.3V in, 3.3 out.
I use basically the same in my DAC (ESS 9006) sans the offset and with different caps. Seems to work fine, very quiet. I have not used the AD797 but have tried some others. (some are not stable)
But what I don't understand is what the heck is it regulating? The only reference is the voltage at C136. Certainly the opamp can supply enough current, but it should be "dirt in, dirt out" - right?
Anyhow, it does work and seems very quiet. But I'm at a loss as to what it's really doing, other than supplying some current.
See the attached below. It's from the Sabre ESS data sheets for 3.3V use on the analog section of their DAC chip. Pretty simple. 3.3V in, 3.3 out.
I use basically the same in my DAC (ESS 9006) sans the offset and with different caps. Seems to work fine, very quiet. I have not used the AD797 but have tried some others. (some are not stable)
But what I don't understand is what the heck is it regulating? The only reference is the voltage at C136. Certainly the opamp can supply enough current, but it should be "dirt in, dirt out" - right?
Anyhow, it does work and seems very quiet. But I'm at a loss as to what it's really doing, other than supplying some current.
Attachments
The answer is simple: nothing. Garbage in - garbage out. And it's actually a little less quiet than the reference voltage DVCC, slightly filtered at the input.
OTOH, it allows you to filter the DC input, and then buffer that filter with the opamp. That's what it does. I would prefer to use a quiet reference as input, like a precision 2.5V ref with the opamp having some gain, but it's the designers choice I guess.
jd
The AD797 may get very upset with that capacitive loading. And those 1.2V or 2.5V references are not at all quiet. And adding gain will certainly add output noise as well.
Actually you can filter DVCC more heavy than shown, and then use for instance an LME49600 as buffer. If syn08 agrees to that
jd
I used Panasonic ceramic caps in my prototype - 4.7uF (1206 package)
http://industrial.panasonic.com/www-data/pdf/ABJ0000/ABJ0000CE4.pdf
ESS recommends the following opamps in descending order of preference:
The eval board uses the NE5534, but it oscillated badly in my version. Wanted to use the LME49720, but it was out of stock. So used an Analog Devices opamp (can't remember what - where are my notes?) Worked well, plenty stable.
http://industrial.panasonic.com/www-data/pdf/ABJ0000/ABJ0000CE4.pdf
ESS recommends the following opamps in descending order of preference:
- AD797
- LME49710, LME49720, LME49740, LM4562
- NE5534, NE5532
- Rohm 4560
The eval board uses the NE5534, but it oscillated badly in my version. Wanted to use the LME49720, but it was out of stock. So used an Analog Devices opamp (can't remember what - where are my notes?) Worked well, plenty stable.
Last edited:
I tested the regulator for PA -- so far for noise only -- the other tests to run are impedance and transient response.
A long while back, TI responded to National's recommendation on bypassing the adjust pin on the LM317 -- while it does reduce the noise it also slows down the regulator. I would imagine that if you had a circuit was subject to dramatic change in load, as for instance in some "logic" this could be an issue.
I'll note also that WJ recommended using shielded cable for the "sense" inputs of his regulator. Jan did the painstaking analysis of the regulators and the results were described in Audio Amateur in 1995.
If someone would convince Ed Dell to issue another CD-ROM with the 1990-1999 issues of AA, Speaker Builder etc. they would have my vote in the next presidential election.
A long while back, TI responded to National's recommendation on bypassing the adjust pin on the LM317 -- while it does reduce the noise it also slows down the regulator. I would imagine that if you had a circuit was subject to dramatic change in load, as for instance in some "logic" this could be an issue.
I'll note also that WJ recommended using shielded cable for the "sense" inputs of his regulator. Jan did the painstaking analysis of the regulators and the results were described in Audio Amateur in 1995.
If someone would convince Ed Dell to issue another CD-ROM with the 1990-1999 issues of AA, Speaker Builder etc. they would have my vote in the next presidential election.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.