Double regulation
Double reg's are wholly beneficial in my experience, if one has a high enough supply voltage to support the drop. As a bare minimum it doubles line / ripple rejection.
LT1086's help here owing to lower dropout.
If you're using both, use LM317 as the pre-reg, 1086 as the final reg, as the final reg sets noise levels.
Andy.
Double reg's are wholly beneficial in my experience, if one has a high enough supply voltage to support the drop. As a bare minimum it doubles line / ripple rejection.
LT1086's help here owing to lower dropout.
If you're using both, use LM317 as the pre-reg, 1086 as the final reg, as the final reg sets noise levels.
Andy.
Andy,
Thanks for the info! I think you're the regulator guru around here, so I hope you don't mind if I quiz you
I was planning to use a ground plane, so do you mean that C1 & C2 should return to the point there the ground wire attaches to the ground plane, or would it be better to forget about the ground plane and use star grounding and local ground planes?
Also, do you think it's worth putting a ferrite bead after the reg, or is it more important to just keep the regulated supply traces as short as possible??
Thanks,
Mark.
Thanks for the info! I think you're the regulator guru around here, so I hope you don't mind if I quiz you
I was planning to use a ground plane, so do you mean that C1 & C2 should return to the point there the ground wire attaches to the ground plane, or would it be better to forget about the ground plane and use star grounding and local ground planes?
Also, do you think it's worth putting a ferrite bead after the reg, or is it more important to just keep the regulated supply traces as short as possible??
Thanks,
Mark.
Guru's a bit strong
...most of them are fake too
Ground planes are an interesting question - the clock supply wasn't quite as quiet as I'd have liked, and I suspect this was due to the ground plane being used for all 0v returns. At these levels everything has finite impedance, even a big ground plane. Current flows in this can give rise to noise, but are beneficial to many devices, particularly digital.
What I'd do in a new design is use the ground plane as the return for all decoupling, but have a seperate Kelvin sense ground, configured as above - consider the ground plane as a star point, but keep 'GNDA' seperate and close to the device it's connected to.
I'm not kidding w.r.t. sensitivity of regulator noise - just 1mm (!) away from the Kelvin sense point on a track with current flowing in it is measurably noisier than at the GNDA point. On a mains-based PSU it manifests itself as harmonics of the charging pulses flowing in the smoothing capacitors (50 / 100Hz in the UK, depending on 1/2 or full wave rectification).
The next two posts will show what I mean.
Andy.
...most of them are fake too
Ground planes are an interesting question - the clock supply wasn't quite as quiet as I'd have liked, and I suspect this was due to the ground plane being used for all 0v returns. At these levels everything has finite impedance, even a big ground plane. Current flows in this can give rise to noise, but are beneficial to many devices, particularly digital.
What I'd do in a new design is use the ground plane as the return for all decoupling, but have a seperate Kelvin sense ground, configured as above - consider the ground plane as a star point, but keep 'GNDA' seperate and close to the device it's connected to.
I'm not kidding w.r.t. sensitivity of regulator noise - just 1mm (!) away from the Kelvin sense point on a track with current flowing in it is measurably noisier than at the GNDA point. On a mains-based PSU it manifests itself as harmonics of the charging pulses flowing in the smoothing capacitors (50 / 100Hz in the UK, depending on 1/2 or full wave rectification).
The next two posts will show what I mean.
Andy.
Improved Kelvin sensing
Here's an improved wiring scheme ,on exactly the same PSU - not quite perfect, but dramatically better.
The Kelvin sense point is now returned to a clean 'audio' 0V star point at the o/p socket of the PSU (where the connected equipment 'sees' the PSU).
It still has a return back to the main 0V star point at the smoothing cap, but the noise change is obvious.
The regulator is now regulating at the point of interest, i.e. where we are connecting to it - the connected device doesn't care what goes on inside the box!
Here's an improved wiring scheme ,on exactly the same PSU - not quite perfect, but dramatically better.
The Kelvin sense point is now returned to a clean 'audio' 0V star point at the o/p socket of the PSU (where the connected equipment 'sees' the PSU).
It still has a return back to the main 0V star point at the smoothing cap, but the noise change is obvious.
The regulator is now regulating at the point of interest, i.e. where we are connecting to it - the connected device doesn't care what goes on inside the box!
Attachments
Shunt Regulator
What do you guys think of this shunt design I drew up?
As Spartcus mentioned I think it should provide rejection of noise after the OP-AMP has given up. Also the raw supply sees a constant current all the time which maybe an advantage.
It will clearly be fairly ineffectient but judging by some of the Amp designs around here I'd say most of you can live with that.
What do you guys think of this shunt design I drew up?
As Spartcus mentioned I think it should provide rejection of noise after the OP-AMP has given up. Also the raw supply sees a constant current all the time which maybe an advantage.
It will clearly be fairly ineffectient but judging by some of the Amp designs around here I'd say most of you can live with that.
power supply for the KWAK-CLOCK
Hi All,
Just finished a power supply for the KWAK-CLOCK , PI filter composed of 1.5 mH and 2200µF and BC550C and TL431 as zener diode. This gave me the opportunity to compare it it with a simple LT1086-5 implemented as in the datasheet.
I must admit that the difference is very subtle sonically. The LT1086 has sligtly more aura in the soundfield and the TL431/BC550 has slightly more peace in the soundfield. Difficult for me to find the right words in English.
When I take the scope as a rescue I see that the TL431 still has a risidue of ripple and and some trace of the 16.9344 MHz clock.
With the LT1086-5 I see a straight trace on the scope i.e. no ripple or clock residue.
I will try the <b>VERY</B> low noise MC 33761 and report later.
It would be interesting to find the relationship between powersupply noise and phase noise of the oscillator. The scientific papers I found thus far did not gave me a clue , only demonstrated the superior IQ of the writers of the paper.
Hi All,
Just finished a power supply for the KWAK-CLOCK , PI filter composed of 1.5 mH and 2200µF and BC550C and TL431 as zener diode. This gave me the opportunity to compare it it with a simple LT1086-5 implemented as in the datasheet.
I must admit that the difference is very subtle sonically. The LT1086 has sligtly more aura in the soundfield and the TL431/BC550 has slightly more peace in the soundfield. Difficult for me to find the right words in English.
When I take the scope as a rescue I see that the TL431 still has a risidue of ripple and and some trace of the 16.9344 MHz clock.
With the LT1086-5 I see a straight trace on the scope i.e. no ripple or clock residue.
I will try the <b>VERY</B> low noise MC 33761 and report later.
It would be interesting to find the relationship between powersupply noise and phase noise of the oscillator. The scientific papers I found thus far did not gave me a clue , only demonstrated the superior IQ of the writers of the paper.
Hi Elso
Clock is working a treat - thanks for that.
Try an LT1086 adjustable, set to 5V. The ability to decouple the adjustment terminal lowers noise and improves line rejection considerably over a fixed regulator.
Large (but sensible) values of divider resistors and a capacitor that's low in impedance relative to the resistor it's across lowers noise at LF too.
You may even find (I haven't checked the data for a 1086-5) that an LM317 is quieter too, configured as suggested.
I did find once some data relatign PSU noise to jitter in CMOS gates, ECL etc. If I find it I'll pass it on.
If in your player the Kwak-Clock passes through another CMOS gate try feeding this from the same low-noise supply. The difference in the Marantz is a great as fitting the clock!
Andy.
Clock is working a treat - thanks for that.
Try an LT1086 adjustable, set to 5V. The ability to decouple the adjustment terminal lowers noise and improves line rejection considerably over a fixed regulator.
Large (but sensible) values of divider resistors and a capacitor that's low in impedance relative to the resistor it's across lowers noise at LF too.
You may even find (I haven't checked the data for a 1086-5) that an LM317 is quieter too, configured as suggested.
I did find once some data relatign PSU noise to jitter in CMOS gates, ECL etc. If I find it I'll pass it on.
If in your player the Kwak-Clock passes through another CMOS gate try feeding this from the same low-noise supply. The difference in the Marantz is a great as fitting the clock!
Andy.
Andy,
re- your query about bitclock re-clocking.
The 74AC00 that was recommended is a NAND gate, and is probably being used as an inverting buffer with its two inputs tied together. Simpler alternatives would be the 74AC04 and 74AC07, straightforward inverting and non-inverting buffers respectively.
The ideal would be to generate the bitclock directly from the masterclock, but this obviously isn't possible if they are at different frequencies. Buffering can at least clean up edges, which helps. Use a small value (~50 ohms) resistor in series with the buffers output to damp any overshoot.
re- your query about bitclock re-clocking.
The 74AC00 that was recommended is a NAND gate, and is probably being used as an inverting buffer with its two inputs tied together. Simpler alternatives would be the 74AC04 and 74AC07, straightforward inverting and non-inverting buffers respectively.
The ideal would be to generate the bitclock directly from the masterclock, but this obviously isn't possible if they are at different frequencies. Buffering can at least clean up edges, which helps. Use a small value (~50 ohms) resistor in series with the buffers output to damp any overshoot.
One suggestion. How about using voltage reference chips as clock supply. These can handle at least couple of ma and they have very high stability and very low noise output + they usually have noise figures given in datasheets...
***
I plan to try it some time, but at the moment I'm engaged with amp project and unfortunately my CD67 died last week. The servo IC seems to be bad. The IC loads the clock input so heavily that the amplitude drops to near nothing.... : (
To get a new SAA7372GP is hard here in Estonia. If any of you guys know where to get it (international shipping) and lower price, then please let me know. I would very much like to save my "good friend" The price one of the local suppliers told me was about 50 EUR and it would be some 3-4 weeks delivery time : (
Ergo
***
I plan to try it some time, but at the moment I'm engaged with amp project and unfortunately my CD67 died last week. The servo IC seems to be bad. The IC loads the clock input so heavily that the amplitude drops to near nothing.... : (
To get a new SAA7372GP is hard here in Estonia. If any of you guys know where to get it (international shipping) and lower price, then please let me know. I would very much like to save my "good friend" The price one of the local suppliers told me was about 50 EUR and it would be some 3-4 weeks delivery time : (
Ergo
Hi alfsch
yeah bit of a silly post really, its 8V in via a 7805 and R5 is a pot. It seems if I turn the pot way down, the voltage out gets close to 5V but everything goes into meltdown mode. Current was measured at 385 mA I turned voltage back to 3.6V and everything is fine, and voltage is sufficient for a tentlabs oscillator.
Thanks for your fast reply
yeah bit of a silly post really, its 8V in via a 7805 and R5 is a pot. It seems if I turn the pot way down, the voltage out gets close to 5V but everything goes into meltdown mode. Current was measured at 385 mA
I turned voltage back to 3.6V and everything is fine, and voltage is sufficient for a tentlabs oscillator.Thanks for your fast reply
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.