You might find this thread interesting then. I certainly did
It looks from that Slimdevices forum that there are major EMC issues with including a TacT preamp (or perhaps only with a Squeezebox) in a system. So perhaps the lower noise of the linear power supply is a secondary issue. I'd experiment with putting a mains isolation transformer in the input to the TacT (low capacitance, i.e. not a toroidal) preceded by a line filter. Also try this in whatever source you're feeding to the TacT. Also try using cables with very good quality screening (braid and foil together) between components. Since I guess your issue is a system issue, simply lowering the noise on your regulator probably won't have much impact on the sound.
Try an LM7805
Try an LM7805 with a current increasing transistor, as shown on page 23 of the LM7805 datasheet (the pdf on this linked page).
LM7805 Datasheet pdf - 3-Terminal 1A Positive Voltage Regulator - Fairchild Semiconductor
Try an LM7805 with a current increasing transistor, as shown on page 23 of the LM7805 datasheet (the pdf on this linked page).
LM7805 Datasheet pdf - 3-Terminal 1A Positive Voltage Regulator - Fairchild Semiconductor
Search for the Salas shunt regulator. Several people have been running it at high current too.
I'm using the Transporter (TP). The SB3 has no comparison to the TP, and AES output of the TP beats the crap out of SB3. Noise is not an issue with TP, as soon as I've reduced the jitter and noise of my TacT preamp, the system sounds better. I've done several iterations of jitter and noise reduction on my TacT preamp, and the system keeps on sounding better, so TP is not the limitation here.
FWIW, I do have a pretty advanced system. I've a balanced isolation transformer with Faraday shield (custom made by Toroid). This feeds my PS Audio PPP AC regenerator. It then distributes power to all my hi-fi gear with critical components not sharing the same bank of AC outlets as PPP provides AC isolation.
The noise floor of my system is already very low. I've also done hard core mods to my TacT 2.2xp preamp, the crystal oscillators driving the ASRC chips have been replaced with low jitter, low phase noise ones. I've also added many Sanyo OsCon caps on each IC chip I can find on the digital input and output boards, to further reduce noise.
Previously, I made an external LPS using a LinearOne LPS. It sounds a lot better than the SMPS it replaces, but my latest creation is a high efficient internal LPS, IMHO, sounds a hell lot better than the external LPS.
I'm using LT1085 to provide the +5V rail to my TacT preamp. The noise output on the unloaded +5V rail is less than 200uV peak-peak when measured on my digital scope. When it is loaded, the noise is just slightly above 2mV.
Today, I changed the LT1085's output cap from a Panasonic "Pureism" 470uF to a Panasonic FM 2200uF low ESR cap. The noise on the +5V rail went down to 1.2mV when system is loaded and running. I guess the 2200uF cap improves the transient response of the regulator. With this mod, I get even more reverbs and ambient from my music. I don't know if bigger output cap will further improve things, I'm doubtful on this, as I did the same on the +/-12V rails and see no futher noise reduction with a bigger cap. I'm really into the "golden ear" stage, so I'm looking at improving the last little mile.
I'm just wondering if there is anything better than what I'm doing now.
I did measure the +5V rail across the ASRC chip on the digital input board, and there is a 7.5mV ripple with 750Hz riding on it. So it looks like I need bigger and better caps to replace the decoupling caps that are already on the main board.
Perhaps I should use a super regulator on the digital input/output board and reduce this 7.5mV ripple to less than 1mV...
Last edited:
LM7805 is noisy, and using it with an external power transistor will make the output noisier!
If loaded noise is more than the unloaded, then you might need to do something on the load end of the DC supply lines, rather than at the supply itself.
The DC rails' conductors have inductance, and resistance. So the instantaneous demands for current at the load ends of those conductors cannot necessarily be met without causing voltage changes, unless you provide local sources for the current demands.
Do you at least have some reservoir capacitance at the point of load, if the DC supply lines are more than a couple of inches in length? [Note that you might want to use lower-grade capacitors, there, than you might at first think, i.e. with some ESR, for damping, to keep things quietest (in general, but "your mileage may vary").] Even if the DC supply paths are very short, you would still need at least a small ceramic capacitor for each active device's power pins.
You'd probably want an electrolytic (size depends on max instantaneous current that might ever be needed, but too big is probably better than too small; 10uF is good first guess) and a small ceramic in parallel across the DC lines, at each point of load. The ceramic should not be the highest quality, i.e. not an NPO or C0G. Otherwise you might get ringing, depending on the characteristics of the supply lines (so why gamble?).
I don't understand how it can really be affecting the sound, but you are there and have measured it, so I'll go with that. If you think that the source of the noise is the supply itself, then you could also add a low-pass filter in the DC positive rail, such as a small R (maybe 20-30 Ohms) in series with a high-current "RF Choke" type inductor (say 100uH +/-50 uH), followed by a large C to ground (say 2200 uF). That should take anything that's there down to the 10's of uV. But, of course, then you would really NEED to make sure that you had capacitors at the point(s) of load, to supply all of the demand for any fast changes in current. Think of them like small power supplies. Also note that they really might need to be right at the power pins of each active device. And especially make sure that the small ceramics are no more than a millimeter or so away from the pins.
If you already have some ten(s) of uF of electrolytic and a .01 to 0.22 uF (0.01 to 0.1 uF is typical) ceramic across each device's power pins, then the previous comments about needing local reservoir capacitance still apply, but at a board and/or subsystem level, in which case the electrolytics are typically larger, say 100uF or more as a first guess.
The bottom line is that it sounds like you have insufficient instantaneous current-delivery capability at one or more points of load where there are dynamic demands for current. So either increase (or optimize) the size of the capacitors you already have at the active devices' power pins, or add them, and/or add board-level or subsystem-level reservoir capacitances.
Other possibilities exist, of course. One example: If you have dynamic ground-return currents (e.g. from power pin ground returns) sharing a length of conductor with the ground reference for an input, then the inductance of the ground return conductor will cause a noisy voltage to appear back at the non-ground ends of each of the ground return conductors. That "bouncing ground" voltage will arithmetically sum with an input voltage for which it is the ground reference, which is "not a good thing".
While not directly related to your current inquiry, this article is good, and might be helpful:
http://www.diyaudio.com/forums/diya...udio-component-grounding-interconnection.html
Cheers,
Tom Gootee
The DC rails' conductors have inductance, and resistance. So the instantaneous demands for current at the load ends of those conductors cannot necessarily be met without causing voltage changes, unless you provide local sources for the current demands.
Do you at least have some reservoir capacitance at the point of load, if the DC supply lines are more than a couple of inches in length? [Note that you might want to use lower-grade capacitors, there, than you might at first think, i.e. with some ESR, for damping, to keep things quietest (in general, but "your mileage may vary").] Even if the DC supply paths are very short, you would still need at least a small ceramic capacitor for each active device's power pins.
You'd probably want an electrolytic (size depends on max instantaneous current that might ever be needed, but too big is probably better than too small; 10uF is good first guess) and a small ceramic in parallel across the DC lines, at each point of load. The ceramic should not be the highest quality, i.e. not an NPO or C0G. Otherwise you might get ringing, depending on the characteristics of the supply lines (so why gamble?).
I don't understand how it can really be affecting the sound, but you are there and have measured it, so I'll go with that. If you think that the source of the noise is the supply itself, then you could also add a low-pass filter in the DC positive rail, such as a small R (maybe 20-30 Ohms) in series with a high-current "RF Choke" type inductor (say 100uH +/-50 uH), followed by a large C to ground (say 2200 uF). That should take anything that's there down to the 10's of uV. But, of course, then you would really NEED to make sure that you had capacitors at the point(s) of load, to supply all of the demand for any fast changes in current. Think of them like small power supplies. Also note that they really might need to be right at the power pins of each active device. And especially make sure that the small ceramics are no more than a millimeter or so away from the pins.
If you already have some ten(s) of uF of electrolytic and a .01 to 0.22 uF (0.01 to 0.1 uF is typical) ceramic across each device's power pins, then the previous comments about needing local reservoir capacitance still apply, but at a board and/or subsystem level, in which case the electrolytics are typically larger, say 100uF or more as a first guess.
The bottom line is that it sounds like you have insufficient instantaneous current-delivery capability at one or more points of load where there are dynamic demands for current. So either increase (or optimize) the size of the capacitors you already have at the active devices' power pins, or add them, and/or add board-level or subsystem-level reservoir capacitances.
Other possibilities exist, of course. One example: If you have dynamic ground-return currents (e.g. from power pin ground returns) sharing a length of conductor with the ground reference for an input, then the inductance of the ground return conductor will cause a noisy voltage to appear back at the non-ground ends of each of the ground return conductors. That "bouncing ground" voltage will arithmetically sum with an input voltage for which it is the ground reference, which is "not a good thing".
While not directly related to your current inquiry, this article is good, and might be helpful:
http://www.diyaudio.com/forums/diya...udio-component-grounding-interconnection.html
Cheers,
Tom Gootee
Last edited:
Hi Tom,
Thanks for your thoughtful reply.
The exisiting digital circuit is a commercial product, and each IC chip has already got a 0.1uF and a 10uF SMD bypass caps on the backside of the circuit board right behind each IC chip.
The main board has a total of ten low grade 470uF electrolytic caps for distributed capacitance.
On the digital input board, I measured the noise ripple across the 10uF bypass cap on the ASIC and it measured 40mV with a 750Hz frequency. Since then, I have added one 270uF Sanyo Oscon SP cap on each chip on the input board (just solder them in parallel with the existing bypass caps), this brings the noise down to 7.5mV. My goal is to bring this down to <1mV and make that 750Hz frequency go away. For now, I haven't done the probing to find out the source of that 750Hz ripple/noise.
One of the things I'm planning to do is to replace those main board 470uF caps with 1000uF. Originally, I thought I'd change them to Panasonic FC or Sanyo Oscon SP. Now after reading your post, I'm not sure if low ESR cap is the way to go. I do have some 1000uF 25V Panasonic Pureism caps and Nichicon MUSE KZ caps. Do you think these will be better choice than the Panasonic FC or Sanyo Oscon? Or will this be a trial and error process?
Another thing I have in my mind is dedicated IC regulator for each individual digital input/outboard board (these boards plug into the main board via a bus connector). Maybe a LM317 to provide a really stable and low noise +5V supply. I do have a spare Belleson Super Regulator (11uV of noise) I can use too instead of LM317. I also have a LT1763 IC regulator I can use (20uV of noise).
My goal is to keep the ripple and noise down to hopefully 200uV or less in various parts of the circuit.
With my new Linear PS, I do hear a significant improvement in sonic quality, despite the device being all digital (save except for the ADC board, in which it requires +/-12V). I used soft recovery shottky rectifiers and soft recovery HEXFREDs in my new LPS, and separate tranformers for +5V and +/-12V rails. I do find that the AD1896 ASRC chip and possibily the AES transceivers are sensitive to noise. To get the ultimate sound, I find it necessary to keep power supply noise down to a bare minimum. BTW, I'm using LT1085 as the main regulator for the +5V rail.
Thanks for your thoughtful reply.
The exisiting digital circuit is a commercial product, and each IC chip has already got a 0.1uF and a 10uF SMD bypass caps on the backside of the circuit board right behind each IC chip.
The main board has a total of ten low grade 470uF electrolytic caps for distributed capacitance.
On the digital input board, I measured the noise ripple across the 10uF bypass cap on the ASIC and it measured 40mV with a 750Hz frequency. Since then, I have added one 270uF Sanyo Oscon SP cap on each chip on the input board (just solder them in parallel with the existing bypass caps), this brings the noise down to 7.5mV. My goal is to bring this down to <1mV and make that 750Hz frequency go away. For now, I haven't done the probing to find out the source of that 750Hz ripple/noise.
One of the things I'm planning to do is to replace those main board 470uF caps with 1000uF. Originally, I thought I'd change them to Panasonic FC or Sanyo Oscon SP. Now after reading your post, I'm not sure if low ESR cap is the way to go. I do have some 1000uF 25V Panasonic Pureism caps and Nichicon MUSE KZ caps. Do you think these will be better choice than the Panasonic FC or Sanyo Oscon? Or will this be a trial and error process?
Another thing I have in my mind is dedicated IC regulator for each individual digital input/outboard board (these boards plug into the main board via a bus connector). Maybe a LM317 to provide a really stable and low noise +5V supply. I do have a spare Belleson Super Regulator (11uV of noise) I can use too instead of LM317. I also have a LT1763 IC regulator I can use (20uV of noise).
My goal is to keep the ripple and noise down to hopefully 200uV or less in various parts of the circuit.
With my new Linear PS, I do hear a significant improvement in sonic quality, despite the device being all digital (save except for the ADC board, in which it requires +/-12V). I used soft recovery shottky rectifiers and soft recovery HEXFREDs in my new LPS, and separate tranformers for +5V and +/-12V rails. I do find that the AD1896 ASRC chip and possibily the AES transceivers are sensitive to noise. To get the ultimate sound, I find it necessary to keep power supply noise down to a bare minimum. BTW, I'm using LT1085 as the main regulator for the +5V rail.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.