Dear Jonathan,
Thanks for the elaborate responses about the OS-CONs.
It is very interesting to read the detailed description of your
design philosphy. If one replaces a few techincal terms here
and there it could just as well be a text-book description on
how to do a software project. I wish more programmers had
your understanding of system design. There wouldn't be as
much crappy software around then.
Yes, that is what is usually known as the 90/10 rule in software
engineering. Either you spend 10% of the time designing the
program and 90% debugging it, or you spend 90% of the time
designing it and 10% of the time debugging it. Almost always
the second approach results in the 100%s
amounting to a much shorter total development time.
Same thing as in software. Although the program may get more
complex in the sense of getting bigger, at least if measured in
source code, it will get much less complex in the sense that it
has a much clearer structure, making it easier to understand,
debug, maintain and modify. I would suppose this design
philospophy is also common for any type of sufficiently
large-scale engineering projects. You always gets to a limit
where it becomes necessary to control the system complexity,
but few probably apply it for smaller-scale projects like amplifiers.
(Here I mean an amplifier is small-scale compared to say the
control system of jet fighter, or a power plant.)
Yes, that is often experienced by programmers tought to do
programming "the right" way, especially if the bosses are not
competent programmers themselves as is, or at least used to
be, common. I have friends who started working in industry
and already the second day the boss asked them "Why haven't
you started coding yet?!!!" Of course my friends were already
deep down into understanding the problem specification and
the overall design of the program. The coding phase would
mostly be routine craftmanship for them once the overall design
was clear and the important technical decisions taken.
I would think that audio equipment are usually still of such
small size that the major challenge is in the details and it is
usually possible to have sufficient control over the whole
system without moving up to a more abstract layer. For
software, you very quickly get into a complexity where things
get into a mess if you don't do it.
There was an interesting observation made once (over 20 years
ago now, I think) at the university where I studied. There were
two similar masters programmes, roughly an EE oriented one
and a more CS oriented one. Both did the first course in digital
design in the second year. However, while the CS programme
had the first programming courses already during the first year,
the EE programme didn't have any programming courses until
the third year. Although no statistic study was done, it was
observed that the CS students performed much better in the
digital design course. This was assumed to be because
the programming courses had forced them into approaching
design on a more abstract level, leading to a better ability to
split up the system into functional blocks etc.
Thanks for the elaborate responses about the OS-CONs.
It is very interesting to read the detailed description of your
design philosphy. If one replaces a few techincal terms here
and there it could just as well be a text-book description on
how to do a software project. I wish more programmers had
your understanding of system design. There wouldn't be as
much crappy software around then.
jcarr said:
Yes, that is what is usually known as the 90/10 rule in software
engineering. Either you spend 10% of the time designing the
program and 90% debugging it, or you spend 90% of the time
designing it and 10% of the time debugging it. Almost always
the second approach results in the 100%s
amounting to a much shorter total development time.
Same thing as in software. Although the program may get more
complex in the sense of getting bigger, at least if measured in
source code, it will get much less complex in the sense that it
has a much clearer structure, making it easier to understand,
debug, maintain and modify. I would suppose this design
philospophy is also common for any type of sufficiently
large-scale engineering projects. You always gets to a limit
where it becomes necessary to control the system complexity,
but few probably apply it for smaller-scale projects like amplifiers.
(Here I mean an amplifier is small-scale compared to say the
control system of jet fighter, or a power plant.)
Yes, that is often experienced by programmers tought to do
programming "the right" way, especially if the bosses are not
competent programmers themselves as is, or at least used to
be, common. I have friends who started working in industry
and already the second day the boss asked them "Why haven't
you started coding yet?!!!" Of course my friends were already
deep down into understanding the problem specification and
the overall design of the program. The coding phase would
mostly be routine craftmanship for them once the overall design
was clear and the important technical decisions taken.
I would think that audio equipment are usually still of such
small size that the major challenge is in the details and it is
usually possible to have sufficient control over the whole
system without moving up to a more abstract layer. For
software, you very quickly get into a complexity where things
get into a mess if you don't do it.
There was an interesting observation made once (over 20 years
ago now, I think) at the university where I studied. There were
two similar masters programmes, roughly an EE oriented one
and a more CS oriented one. Both did the first course in digital
design in the second year. However, while the CS programme
had the first programming courses already during the first year,
the EE programme didn't have any programming courses until
the third year. Although no statistic study was done, it was
observed that the CS students performed much better in the
digital design course. This was assumed to be because
the programming courses had forced them into approaching
design on a more abstract level, leading to a better ability to
split up the system into functional blocks etc.
-----------------------------------------------------------------------------Peter Daniel said:
My own take from cap changes is that better caps can reveal problems in other areas such as longer warm up times, opamp sonic problems, and longer settling times. I'd give everything a month beforev a verdict.
The super E seems to reveal more at the top and bottom than a single BGN. The reason why I think this is liked more than FK is that it is a 'mellower' cap and less revealing.
Re: What was the application?
Coupling; BGNs are too expensive and too big. I prefer OSCON SG or SP for digital; FK for IMPORTANT decoupling positions (expensive), Elena Stagets for eveuyday and Rubycon ZAs, ZLs (magnetic leads) for low esr.
Most of the time decoupling analogue PS doesn't seem too critical, provided there is no groun, oscillation or regulator noise problem.
----------------------------------------------------------------------Oli said:
Coupling; BGNs are too expensive and too big. I prefer OSCON SG or SP for digital; FK for IMPORTANT decoupling positions (expensive), Elena Stagets for eveuyday and Rubycon ZAs, ZLs (magnetic leads) for low esr.
Most of the time decoupling analogue PS doesn't seem too critical, provided there is no groun, oscillation or regulator noise problem.
Hi All
Forgive my ignorance [yet again
], I have a couple of further questions.
Regarding decoupling caps on the 8412 etc.
The data sheet shows .1uf yet i often see some quite large[single] values used in its place.
?
Another question regarding ldo reg cap values.[I suppose
this is related in that the cap value effects both devices]
I have been investigating regs [especially those that are BIGGER than a grain of rice
] , for example Burr Brown Reg 103.
The data sheet states:
"None of the versions require an output capacitor for regulator
stability. The REG103 will accept any output capacitor
type less than 1µF. For capacitance values larger than 1µF,
the effective ESR should be greater than 0.1Ω. This minimum
ESR value includes parasitics such as printed circuit
board traces, solder joints, and sockets. A minimum 0.1µF
low ESR capacitor connected to the input supply voltage is
recommended."
Mmm ..how does one go about using such a device if the device it is supplying requires or benefits from a higher value[low esr type] cap?
Cheers
Setmenu
Forgive my ignorance [yet again
], I have a couple of further questions.Regarding decoupling caps on the 8412 etc.
The data sheet shows .1uf yet i often see some quite large[single] values used in its place.
? Another question regarding ldo reg cap values.[I suppose
this is related in that the cap value effects both devices]
I have been investigating regs [especially those that are BIGGER than a grain of rice
] , for example Burr Brown Reg 103.The data sheet states:
"None of the versions require an output capacitor for regulator
stability. The REG103 will accept any output capacitor
type less than 1µF. For capacitance values larger than 1µF,
the effective ESR should be greater than 0.1Ω. This minimum
ESR value includes parasitics such as printed circuit
board traces, solder joints, and sockets. A minimum 0.1µF
low ESR capacitor connected to the input supply voltage is
recommended."
Mmm ..how does one go about using such a device if the device it is supplying requires or benefits from a higher value[low esr type] cap?
Cheers
Setmenu
Hi jewilson
I can't even explain what I don't know correctly
I think I meant to say something like:
If I wished to use the mentioned regulator ,that I would be
unable to use it to supply my devices if I intend to use a bypass
cap that has a low esr with a value over 1uf, or instability could result.
Types such as OSCONs, Black Gates etc.?
Is the above the correct interpretation of below?
"None of the versions require an output capacitor for regulator
stability. The REG103 will accept any output capacitor
type less than 1µF. For capacitance values larger than 1µF,
the effective ESR should be greater than 0.1Ω. This minimum
ESR value includes parasitics such as printed circuit
board traces, solder joints, and sockets. A minimum 0.1µF
low ESR capacitor connected to the input supply voltage is
recommended."
The reason for all this is my wish to be able to run the thing from as small a battery pack as possible, hence the ldo reg investigations.
I suppose I could just up the cell count [probably will need to anyway] and shove some standard 3 pin in there...
UGH.. I just hate being so ignorant..I feel a real duck out of water here
Just dunno what I am talking about...sigh..
Perhaps I should stop vandalizing this very interesting thread with my posts eh?
Cheers
Setmenu
I can't even explain what I don't know correctly
I think I meant to say something like:
If I wished to use the mentioned regulator ,that I would be
unable to use it to supply my devices if I intend to use a bypass
cap that has a low esr with a value over 1uf, or instability could result.
Types such as OSCONs, Black Gates etc.?
Is the above the correct interpretation of below?
"None of the versions require an output capacitor for regulator
stability. The REG103 will accept any output capacitor
type less than 1µF. For capacitance values larger than 1µF,
the effective ESR should be greater than 0.1Ω. This minimum
ESR value includes parasitics such as printed circuit
board traces, solder joints, and sockets. A minimum 0.1µF
low ESR capacitor connected to the input supply voltage is
recommended."
The reason for all this is my wish to be able to run the thing from as small a battery pack as possible, hence the ldo reg investigations.
I suppose I could just up the cell count [probably will need to anyway] and shove some standard 3 pin in there...
UGH.. I just hate being so ignorant..I feel a real duck out of water here
Just dunno what I am talking about...sigh..
Perhaps I should stop vandalizing this very interesting thread with my posts eh?
Cheers
Setmenu
First, your choice of regulators should not be biased on a minimum acceptable cap for that regulator or ESR value. The one specification should be that the regulator will remain stable with a minimum 100uf or more. If the regulators cannot handle that then do not use it. For example, look at the specification of a LM317 and LM337 they are decent regulators, even a 78XX or 79 series regulators or ok when used with low ESR caps.
You should be looking for a regulator with good performance specification. Most regulators require a minimum size cap to function, however, that does not mean you should use the minimum size. In fact, you will not get good filtering using the minimum size.
If you wanting to building to be battery powered thing, you should be looking at a "low drop out regulator" they are the most efficient. Most low drop out regulators are not high of performance, however with good filtering they will get close. When using a LDO Regulators you really need goof out filter caps, because their noise performance and ripple is worst. Check out Linear Technology they makes a bunch LDO regulators.
You should be looking for a regulator with good performance specification. Most regulators require a minimum size cap to function, however, that does not mean you should use the minimum size. In fact, you will not get good filtering using the minimum size.
If you wanting to building to be battery powered thing, you should be looking at a "low drop out regulator" they are the most efficient. Most low drop out regulators are not high of performance, however with good filtering they will get close. When using a LDO Regulators you really need goof out filter caps, because their noise performance and ripple is worst. Check out Linear Technology they makes a bunch LDO regulators.
setmenu said:You are correct of course.
I was working my self into a right old frenzy at the weekend
I have since done some more reading and investigation.
I have ordered some ADP3301 regs to try.
these are low dropout , low noise and can tolerate a wide range of caps/esr.
Cheers
Setmenu
At best these are above 100nV/SqrrtHz, which is about the same as an LM317, not realy low....
cheers
I have little idea of what noise figs mean.
I just try to find products that use the same definitions and go for the lowest figure
The ADP 3301 quotes:
OUTPUT NOISE VNOISE f = 10 Hz–100 kHz
@ 5 V OUTPUT CNR = 0 100 uV rms
CNR = 10 nF, CL = 10 uF 30 uV rms
They say it is low noise, so I believe them??
Setmenu
I just try to find products that use the same definitions and go for the lowest figure
The ADP 3301 quotes:
OUTPUT NOISE VNOISE f = 10 Hz–100 kHz
@ 5 V OUTPUT CNR = 0 100 uV rms
CNR = 10 nF, CL = 10 uF 30 uV rms
They say it is low noise, so I believe them??
Setmenu
setmenu said:I have little idea of what noise figs mean.
I just try to find products that use the same definitions and go for the lowest figure
The ADP 3301 quotes:
OUTPUT NOISE VNOISE f = 10 Hz–100 kHz
@ 5 V OUTPUT CNR = 0 100 uV rms
CNR = 10 nF, CL = 10 uF 30 uV rms
They say it is low noise, so I believe them??
Setmenu
Also worth a look are LT1963 regs.
These have following attributes:
Low drop out
Very low noise (14uV rms 10Hz~100kHz)
Fast transient response
Stability with wide range of OP capacitors
Refer data sheet for more details.
Cheers,
Terry
Hi Terry
I had spotted the 1963 in my searches.
The reason I got into such a fret over the reg 103 is the fact I could just simply order it from RS
All the others seem to be thin on the ground in the UK
I get sooo sick of hunting for components
The nice folk at Analog Devices UK are sending some samples
of the 3301 [could not find it anywhere else as usual].
I have yet to speak the Linear Technoloy...As I cannot find
this one by other means either
As for noise, I cannot recollect reading of an off the shelf reg
that anyone here jumped up an down about, [I probably missed that one!] preferences generally being shown for discreet types.
Soo..what kind of noise figure would be considered good then?
Assuming I understand the answer of course .
Cheers
Setmenu
I had spotted the 1963 in my searches.
The reason I got into such a fret over the reg 103 is the fact I could just simply order it from RS
All the others seem to be thin on the ground in the UK
I get sooo sick of hunting for components
The nice folk at Analog Devices UK are sending some samples
of the 3301 [could not find it anywhere else as usual].
I have yet to speak the Linear Technoloy...As I cannot find
this one by other means either
As for noise, I cannot recollect reading of an off the shelf reg
that anyone here jumped up an down about, [I probably missed that one!] preferences generally being shown for discreet types.
Soo..what kind of noise figure would be considered good then?
Assuming I understand the answer of course
.Cheers
Setmenu
The fact is, many of these regulators will work fine if the correct filtering is used, even a PI filter helps and ferrite beads will not hurt either.
For the digital stuff except for the receiver, noise is not that big of deal unless it ends up on your analog power and ground. That another discussion.
For the digital stuff except for the receiver, noise is not that big of deal unless it ends up on your analog power and ground. That another discussion.
jewilson said:
Guido,
That all depend on his application. It not so important to have ultra low noise regs for a digital application.
Low noise is important in digital, as some aspects of digital are not so digital. Domains where low noise is a must are at least the clock, analogue PLL supplies and DAC supplies.
Low means low, 5nV/Sqrrt Hz at least up to 100kHz
cheers
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.