Just to make an official warning : I'm new to modding of electronics...
the setting : I have an old philips CD950 in good condition. (Just to refresh your mind : CDM9 transport, SAA7350/TDA1547 DAC, NE5532 output filter)
As a learning project I would like to make some changes just to see how far I can go with it. Off course I'm after the 'best' sound, but also in the least expensive way (aren't we all ?)
Things I've done so far is changing the oscillator crystal with a TentXO --> definite improvement here (Thanks Guido !)
Next on my list is upgrading the electrolytic caps in supply lines and output to Black Gates.
Now for the questions : I'm also looking for a suitable replacement for the NE5532 dual opamps, which are used only for a 3rd order filter. The schematic used in my player is essentially the same as in the TDA1547 datasheet. (Yes I've got the service manual)
I've searched the net and these forums a lot and found this list of possible suitable replacements :
NE5532 (8 V/µs ; 10 MHz)
OPA2604 (25 V/µs ; 20 MHz)
OPA2134 (20 V/µs ; 8 MHz)
AD712 (20 V/µs ; 4 MHz)
AD826 (350 V/µs ; 50 MHz)
LM6172 (3000 V/µs ; 100 MHz)
As you can see I've done my homework and I've listed the slewrate and unity gain bandwith. Some ones are the dual versions of some single opamps praised by some...
The thing that strikes me is that all of the types mentioned are 'faster' or better responding than the original 5532, however the one loved the most is the AD826 (827 in single version). The LM is faster, so should it be even better ?
Or are there other things to look for ?
As I said : I'm new to this, so I'd like to learn this the right way, thus asking for some input of the pro's...
thanks
the setting : I have an old philips CD950 in good condition. (Just to refresh your mind : CDM9 transport, SAA7350/TDA1547 DAC, NE5532 output filter)
As a learning project I would like to make some changes just to see how far I can go with it. Off course I'm after the 'best' sound, but also in the least expensive way (aren't we all ?)
Things I've done so far is changing the oscillator crystal with a TentXO --> definite improvement here (Thanks Guido !)
Next on my list is upgrading the electrolytic caps in supply lines and output to Black Gates.
Now for the questions : I'm also looking for a suitable replacement for the NE5532 dual opamps, which are used only for a 3rd order filter. The schematic used in my player is essentially the same as in the TDA1547 datasheet. (Yes I've got the service manual)
I've searched the net and these forums a lot and found this list of possible suitable replacements :
NE5532 (8 V/µs ; 10 MHz)
OPA2604 (25 V/µs ; 20 MHz)
OPA2134 (20 V/µs ; 8 MHz)
AD712 (20 V/µs ; 4 MHz)
AD826 (350 V/µs ; 50 MHz)
LM6172 (3000 V/µs ; 100 MHz)
As you can see I've done my homework and I've listed the slewrate and unity gain bandwith. Some ones are the dual versions of some single opamps praised by some...
The thing that strikes me is that all of the types mentioned are 'faster' or better responding than the original 5532, however the one loved the most is the AD826 (827 in single version). The LM is faster, so should it be even better ?
Or are there other things to look for ?
As I said : I'm new to this, so I'd like to learn this the right way, thus asking for some input of the pro's...
thanks
Look at settling time. Do some work, and figure out how many bits that 0.1 and 0.01% represents. Tha will give you an idea which ones are capable of high resolution.
One of the reasons often cited as to why current feedback op-amps do not work well in this application is their settling time. They are generally fast, but don't always settle nicely.
Jocko
One of the reasons often cited as to why current feedback op-amps do not work well in this application is their settling time. They are generally fast, but don't always settle nicely.
Jocko
Thank you for your response...
When looking at those settling times you mentioned, indeed the first 4 choises had settling times around 1 microsecond.
The last two (AD826 and LM6172) were far better.... with LM6172 being the slower one at 65 ns.
I'm not sure though what exactly you mean by how many bits.
If I calculate the number of samples (@ 44.1 kHz sampling rate) that would be, I get around 2 or 3 samples respectively
Does that mean that it 'lasts' 2 or 3 samples before the signal is back to 'zero', thus delaying everything a bit or am I completely off ?
Anyway I broke down and ordered me some of each kind, so I'll be doing some tests soon
Are there even better choices ?
When looking at those settling times you mentioned, indeed the first 4 choises had settling times around 1 microsecond.
The last two (AD826 and LM6172) were far better.... with LM6172 being the slower one at 65 ns.
I'm not sure though what exactly you mean by how many bits.
If I calculate the number of samples (@ 44.1 kHz sampling rate) that would be, I get around 2 or 3 samples respectively
Does that mean that it 'lasts' 2 or 3 samples before the signal is back to 'zero', thus delaying everything a bit or am I completely off ?
Anyway I broke down and ordered me some of each kind, so I'll be doing some tests soon
Are there even better choices ?
0.01% is 1 part in 10000, or still 6.5 counts in a 2^16=65535 counts full scale system.
My hunch says AD817/AD826. Too many people reporting implementation problems with LM6171/6172s.
I would not necessarily go the Black Gate route. It is costly and possibly keeps you from learning by experimentation. Anyway, more than one talented designed (i.e. not me) will tell you that 95% of the sound is in the circuit, not in fancy parts.
Better look at the existing c ircuit, replace supply caps with new ones that are (capacitance and ESR value) suited for the job, and try to eliminate elcaps from the direct signal path, or replace them with good polypropylene foil caps. (If necessary size-wise BG non-polar or Elna Silmic in anti-series.)
If the 950 really follows the datasheet: I see not use for loading the output opamp with 1k5, apart from quickly discharging the 100uF output elcap.
Replace that resistor with 15k at least (and why not 100k?), so that the output cap can be an order of magnitude smaller. If you know the smallest external load you're going to drive you calculate the smallest allowed cap as the one giving -3dB below 2Hz.
If you want to use IC sockets, then contemplate going for these tiny pins that are buried into the PCB, i.e. no added L or stray C.
SIL socket
If the capacitors in the actual low-pass filters are ceramics, then you may want to replace these with something like Evox PFR foil caps.
http://www.evoxrifa.com/cap_catalog/pulsecap/pfr.pdf
My hunch says AD817/AD826. Too many people reporting implementation problems with LM6171/6172s.
I would not necessarily go the Black Gate route. It is costly and possibly keeps you from learning by experimentation. Anyway, more than one talented designed (i.e. not me) will tell you that 95% of the sound is in the circuit, not in fancy parts.
Better look at the existing c ircuit, replace supply caps with new ones that are (capacitance and ESR value) suited for the job, and try to eliminate elcaps from the direct signal path, or replace them with good polypropylene foil caps. (If necessary size-wise BG non-polar or Elna Silmic in anti-series.)
If the 950 really follows the datasheet: I see not use for loading the output opamp with 1k5, apart from quickly discharging the 100uF output elcap.
Replace that resistor with 15k at least (and why not 100k?), so that the output cap can be an order of magnitude smaller. If you know the smallest external load you're going to drive you calculate the smallest allowed cap as the one giving -3dB below 2Hz.
If you want to use IC sockets, then contemplate going for these tiny pins that are buried into the PCB, i.e. no added L or stray C.
SIL socket
If the capacitors in the actual low-pass filters are ceramics, then you may want to replace these with something like Evox PFR foil caps.
http://www.evoxrifa.com/cap_catalog/pulsecap/pfr.pdf
Jocko are you referring to the fact that if you want the full precision of the DAC (16bit as an example) you would need to know how long it takes to settle to (1/2^16)*100 %
~0,0015%
Assuming 88k Fs you have ~10us per sample and if it takes 1 us or more to settle to 0.01% you might not realise the full precision?
I am also interested in your statement on settling nicely, is it only the time or the way it settles (i.e overshoot or damped).
~0,0015%
Assuming 88k Fs you have ~10us per sample and if it takes 1 us or more to settle to 0.01% you might not realise the full precision?
I am also interested in your statement on settling nicely, is it only the time or the way it settles (i.e overshoot or damped).
Be careful when comparing settling times, though. The datasheets
usually only states one figure for settling time, but they are not
always comparable. Some datasheets give the settling time to
0.1% and others the settling time to 0.01% for instance. Once
audio company selling op amps for modding made a big affair
of a certain op amp having a fast settling time and compared it
to a lot of other op amps. What they failed to say was that for
some of the slower op amps, it was actually the settling to to
a better precision that stated in the datasheet. Incidentally,
most other figures they stated for the various op amps were also
misleading or even wrong, but that's another story perhaps.
usually only states one figure for settling time, but they are not
always comparable. Some datasheets give the settling time to
0.1% and others the settling time to 0.01% for instance. Once
audio company selling op amps for modding made a big affair
of a certain op amp having a fast settling time and compared it
to a lot of other op amps. What they failed to say was that for
some of the slower op amps, it was actually the settling to to
a better precision that stated in the datasheet. Incidentally,
most other figures they stated for the various op amps were also
misleading or even wrong, but that's another story perhaps.
Werner thanks for your comments
Concerning the caps, the pcb will probably not support repeated exchange of parts on the same place, so I kinda went for the good ones anyway. I have not heard of any capacitor sockets
The supply caps are bypassed with a 100 n SMD (ceramics ??) on the bottom of the board so I guess that's ok ?
The output cap is indeed a 100 micro non polar... There are 2 parallel 100k resistors on the output (1 for the line out, 1 for the headphone opamps) and 2 muting transistors, with another FET(also for muting I guess). So the load would be what, like 50k ?
If I leave the circuit as is (which is the initial plan --> go one step at a time) I'll have to use the large cap.
The caps in the feedback loops (both external opamps and for the dac) are those blue ones you can see in this picture http://eddie.dyec.com.tw/diy-demo/audio-diy/cd_dvd_modify/philipscd951/951photo/icsocket.jpg .
I guess (read : hope) those are not just ceramics
Anyway we're drifting from the topic here, maybe I'll better continue this in another thread....
Christer, hjelm
I indeed found some inconsistencies with the reporting (or not) of those settling times, but I guess it's just to give the end user an idea.
Indeed the thought that those active devices are responding too slow and thus introducing non linearities is not very promising....
I guess there's no way around that other then using the best opamps with respect to response time AND settling time ?
Concerning the caps, the pcb will probably not support repeated exchange of parts on the same place, so I kinda went for the good ones anyway. I have not heard of any capacitor sockets
The supply caps are bypassed with a 100 n SMD (ceramics ??) on the bottom of the board so I guess that's ok ?
The output cap is indeed a 100 micro non polar... There are 2 parallel 100k resistors on the output (1 for the line out, 1 for the headphone opamps) and 2 muting transistors, with another FET(also for muting I guess). So the load would be what, like 50k ?
If I leave the circuit as is (which is the initial plan --> go one step at a time) I'll have to use the large cap.
The caps in the feedback loops (both external opamps and for the dac) are those blue ones you can see in this picture http://eddie.dyec.com.tw/diy-demo/audio-diy/cd_dvd_modify/philipscd951/951photo/icsocket.jpg .
I guess (read : hope) those are not just ceramics
Anyway we're drifting from the topic here, maybe I'll better continue this in another thread....
Christer, hjelm
I indeed found some inconsistencies with the reporting (or not) of those settling times, but I guess it's just to give the end user an idea.
Indeed the thought that those active devices are responding too slow and thus introducing non linearities is not very promising....
I guess there's no way around that other then using the best opamps with respect to response time AND settling time ?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.