| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Does the change from the 4,700uF to the 2,200uF capacitor make a big difference?
|
Not really. A 2,200uF Capacitor will result in approximatly twice the ripple, it's sort of six db of one, halve a dozend db of the other. I would think capacitor quality more important than value, so some changes in value to suit what is at hand are just fine.
| quote: | Originally posted by Fin
Also, I have seen some examples where something like: 2,200uF - 10R - 2,200uF - Reg. have been used. What difference does this type of arrangement make?
|
The larger first capaictor has two effects. First, it will make the ripple lower (but we have regulators to take care of that anyway) and it will reduce the conduction angle (the time out of 360degrees = one cycle of the mains voltage) significantly, resulting in narrower, higher amplitude current peaks being drawn on each of the mains voltage's halve waves.
I consider this mostly a "bad thing"..... ;-)
| quote: | Originally posted by Fin
Also, Thorsten spoke of using 100uF at the regulator pins - Dave has used 220uF. Is this significant?
|
Same as above, six db of one, halve a dozend db of the other. Use what is available, good quality and so on.
| quote: | Originally posted by Fin
For the power supplies to the servos, Thorstem suggests capacitors with about 10,000uF before and after the regulators:-
It looks like I may have to compromise on something here, due to availability. So I'm wondering what is most important - capacitor size or quality? |
Okay, this important to understand. The Servos and motors are operating at fairly slow speeds (mechanical components with a given mass) and the servos attempt to enforce a certain behaviour to those masses with their inertia. This means that the servo circuits and motor drivers usually draw current in an irregeluar manner, with quite high peaks on a low average current, a lot in fact like a power amplifier driving a speaker with music. So we are looking here more at a powersupply for a woofer amplifier than for a digital/class A analog circuit.
Does that illuminate the issues at all?
| quote: | Originally posted by Fin
A transformer with 12Vac secondaries produces about 17Vdc. I assume that a 25V capacitor is required here as 16V is probably a bit risky - even allowing for voltage drop across the diodes.
|
Yup, completely agreed. Use 25...35V.
| quote: | Originally posted by Fin
In quality capacitors like Panasonic FC or Rubycon YXF, the largest value in 25V is 4,700uF. In Nichicon Muse, there is 6,800uF. Cheaper brands like Jamicon, make a 10,000uF.
|
Same prnciple as before, use what's available. I used Samwaha "Tag" connection 10,000uF/35V "Audio Grade" tyoes, which where available and have good ripple rating and low ESR/ESL (for a capacitor of that size).
Sayonara |
|
|
| Dave S |
Fin,
You really need to try some of this stuff for yourself - and believe what you're hearing. Don't get hung up on detailed component choices, get the fundamentals right (grounding, layout, well proven circuits) and sort the "audiovile" stuff later.
FWIW, I'm now at stage 1d (from post #186) and the only one I have doubts about is 1b although I think this relates more to implementation than component or topology choice. At some stage I will reverse 1b or maybe try another implementation (to improve the grounding).
I found that a better servo PSU gave an increase in scale and control, but the improved 5V supply for the main CD processor gave a more fundmental improvement to the music. In light of this I may tweak or improve the 5V PSU at some later stage, whereas I'm happy that there is not a lot of mileage in tweaking the 12V PSU.
I also think my comment about the CD723 being as good as my Arcam was a bit premature (I'm sure I mentioned in an earlier post that I'm frequently wrong).
If you look at Thorsten's recommendations, it's all about sorting the fundamentals, there's no snake oil or requirement for Peruvian virgins etc - that's why I decided to go with his mods (except perhaps the C37, which I have absolutely no idea about).
Dave |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Dave S
I also think my comment about the CD723 being as good as my Arcam was a bit premature (I'm sure I mentioned in an earlier post that I'm frequently wrong). |
I would propose that once you are really through (right Op-Amp as I/V) it will be as good, if not better....
| quote: | Originally posted by Dave S
If you look at Thorsten's recommendations, it's all about sorting the fundamentals, there's no snake oil or requirement for Peruvian virgins etc - that's why I decided to go with his mods (except perhaps the C37, which I have absolutely no idea about). |
Actually, I am quite in favour of any number of highly esotheric principles in Audio Design, but only on top of and after teh fndamentals have been gotten right, not instead of getting them right.
Sayonara |
|
|
| Fin |
Before I think of more questions to ask - I must say again that this is an excellent thread (for me anyway) and thank all of you who have provided such excellent answers and information. Hopefully others are getting as much out of this as I have.
Hi Thorsten
| quote: | Originally posted by Kuei Yang Wang
Not really. A 2,200uF Capacitor will result in approximatly twice the ripple, it's sort of six db of one, halve a dozend db of the other. I would think capacitor quality more important than value, so some changes in value to suit what is at hand are just fine.
|
As I'm starting with a blank sheet and empty box of components, I might as well get the optimum values (if that is the right term).
So, I'll go with 470uF, 3,300uF or 4,700uF and some 100uF Nichicons Muse for the analogue stage. Then, just for variety, use Panasonic FC in the same values, for the DAC and decoder.
| quote: | Originally posted by Kuei Yang Wang
The larger first capaictor has two effects. First, it will make the ripple lower (but we have regulators to take care of that anyway) and it will reduce the conduction angle (the time out of 360degrees = one cycle of the mains voltage) significantly, resulting in narrower, higher amplitude current peaks being drawn on each of the mains voltage's halve waves.
I consider this mostly a "bad thing"..... ;-)
|
I actually understand this!
| quote: | Originally posted by Kuei Yang Wang
Okay, this important to understand. The Servos and motors are operating at fairly slow speeds (mechanical components with a given mass) and the servos attempt to enforce a certain behaviour to those masses with their inertia. This means that the servo circuits and motor drivers usually draw current in an irregeluar manner, with quite high peaks on a low average current, a lot in fact like a power amplifier driving a speaker with music. So we are looking here more at a powersupply for a woofer amplifier than for a digital/class A analog circuit.
|
So - the larger value here is quite important in order to provide a reserve for the irregular demand.
| quote: | Originally posted by Kuei Yang Wang
Does that illuminate the issues at all?
|
Yes - quite a lot - thanks.
| quote: | Originally posted by Kuei Yang Wang
Same prnciple as before, use what's available. I used Samwaha "Tag" connection 10,000uF/35V "Audio Grade" tyoes, which where available and have good ripple rating and low ESR/ESL (for a capacitor of that size).
|
The Jamicons have relatively low impedance and high ripple current figures - can't find ESR/ESL figures for them. I will look for the Samwahas as well.
Hi Dave
| quote: | Originally posted by Dave S
You really need to try some of this stuff for yourself - and believe what you're hearing. Don't get hung up on detailed component choices, get the fundamentals right (grounding, layout, well proven circuits) and sort the "audiovile" stuff later.
|
A good friend in The Netherlands has been telling me this for months, and I do understand that the layout and implementation is more important. We have been working on ideas for improving the grounding, decoupling and power supplies.
The only capacitors available locally are Jamicons (or Hitano - if I order from Sydney). So - for me to get types that others have tried and had success with - I will be ordering from RS Components and some suppliers in the US, etc. In some cases, the audio grade stuff is comming in cheaper than the normal stuff. For example, the 4,700 Nichicon Muse FX is AUD$4.20 from the US, while the Panasonic FC 3,300uF is AUD$6.00 from RS Australia. The delivery from the US is no more expensive than delivery from Sydney. I'm not specifically chasing audio grade parts - but if they are still appropriate, and cheaper - then, why not.
| quote: | Originally posted by Dave S
FWIW, I'm now at stage 1d (from post #186) and the only one I have doubts about is 1b although I think this relates more to implementation than component or topology choice. At some stage I will reverse 1b or maybe try another implementation (to improve the grounding). |
Seems like you are going well. What would (or could) you do differently? Are you talking about connecting the ground pins of ICs directly to the groundplane, ensuring that no other circuits share any of the ground or supply traces etc?
| quote: | Originally posted by Dave S
I found that a better servo PSU gave an increase in scale and control, but the improved 5V supply for the main CD processor gave a more fundmental improvement to the music. In light of this I may tweak or improve the 5V PSU at some later stage, whereas I'm happy that there is not a lot of mileage in tweaking the 12V PSU. |
It's good to hear that the improved servo supply brought about an improvement. I had read some comments by Guido Tent on this and tried to get more information on it by starting a thread:
Jitter induced by servos
Unfortunately, it didn't generate much interest. Now that Thorsten confirms this, and you have heard the results - there is no doubt. |
|
|
| Fin |
Hi Thorsten
How specific to the CD7xx are your power supply designs and recommendations?
| quote: | Originally posted by Kuei Yang Wang [/b]
Based on what you send me, the 624 is an earlier unit. I am not sure which DAC is used. My comments are very specific to the group of Machines including the Philips CD710/11/13/20/21/23 (probably also 12/22 and the Marantz CD36/46 et al. These have a TDA1545 DAC and similar chipsets and circuits.
[/B] |
| quote: | Originally posted by Kuei Yang Wang [/b]
Not really. A 2,200uF Capacitor will result in approximatly twice the ripple, it's sort of six db of one, halve a dozend db of the other. I would think capacitor quality more important than value, so some changes in value to suit what is at hand are just fine.
[/B] |
In light of the above coments, would it be reasonable to apply similar power supply arrangements to other CD players and to use similar capacitor values?
What type of factors would would cause a departure from the 3R3 - 470Uf - 10R - 4700uF approach? |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
How specific to the CD7xx are your power supply designs and recommendations? |
Modestly so, they apply everywhere where currents are not excessive (eg modern players).
| quote: | Originally posted by Fin
In light of the above coments, would it be reasonable to apply similar power supply arrangements to other CD players and to use similar capacitor values? |
Yes, but make sure that the supply can deliver the required current at the required voltage.
| quote: | Originally posted by Fin
What type of factors would would cause a departure from the 3R3 - 470Uf - 10R - 4700uF approach? |
Considerably higher supply currents for any of the sections/IC's in the player.
Sayonara |
|
|
| Fin |
| quote: | Originally posted by Kuei Yang Wang
Konnichiwa,
Modestly so, they apply everywhere where currents are not excessive (eg modern players).
Yes, but make sure that the supply can deliver the required current at the required voltage.
Considerably higher supply currents for any of the sections/IC's in the player.
Sayonara |
OK - the analogue output stage will consist of one dual opamp. If I use something like OPA2134 ( which a lot of people seem to like in this type of CDP), it will require about 8mA. This is all that will be on that supply. Should I change anything?
The TDA1545 in the CD723 uses about 6-8mA. The DAC in the 624 uses about 60mA (Idda=20ma, Iddd=40ma). The max is about 120mA. This is all that will be on this supply. Does anything need to be changed?
The SAA7310 decoder uses about 35-50mA. This is all that will be on this supply. Does anything need to be changed?
How much higher is "considerably higher"? Is there a range that these supplies are suitable for?
Sorry for asking these questions about a different circuit and a machine that you have not worked on - but maybe the 624 can pirate some ideas from the 723. |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Sorry for asking these questions about a different circuit and a machine that you have not worked on - but maybe the 624 can pirate some ideas from the 723. |
What I would suggest you do is to model each supply in Duncan Amplifiers PSU Designer.
The timeconstants in the circuit I propose (1,500uS/102Hz for the first RC and 47,000uS/3.4Hz) are intended to provide a reasonable level of filtering and a "slow down" of the charge pulses. I would keep these about the same but adjust the rest of the circuit.
For example, a circuit PSU supplying 10mA to a "Class A" circuit (no material change in current with signal) may very well use larger value Resistors and smaller value capacitors (eg. 10R/150uF/33R/1500uF) and with a fairly high impedance (low power) transformer. On the other hand a circuit drawing 100mA with a strong modulation of the current (say +/-80%) like a class AB Amplifier would require a lower resistive component and larger value capacitors (eg. 1R5/1000uF/4R7/10,000uF) and a higher power transformer.
Simply use PSUD to simulate the "worst case", meaning currents at the Datasheet (plus dynamic current) limits and the mains voltage at the lowest permissable and then make sure the output voltage after the Filter is enough to avoid the datasheet limit dropout voltage for the specified regulator. Take care to include a reasonable model of the mains transformer as well.
That way your PSU is designed for the specific application and suited for it. Usually that is the only way to get the best results. Also always check the step response of the PSU Circuit to a 50% change in current, more so when LC supplies are used!
Sayonara |
|
|
| Bricolo |
| quote: | Originally posted by Kuei Yang Wang
Konnichiwa,
Also always check the step response of the PSU Circuit to a 50% change in current, more so when LC supplies are used!
Sayonara |
can this be done in PSU designer? I never san this function (but missed it) |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Bricolo
can this be done in PSU designer? |
Yes.
| quote: | Originally posted by Bricolo
I never san this function (but missed it) |
Select a constant current load, select stepped load and enter the before step and after step value of current and the time you want the step to occur. Make sure the step is within the time window selected to be displayed and it may be usefull to introduce the step after the PSU's intial setteling is complete. This usually means very long display times (30 seconds) and zooming in strongly on the step.
Sayonara |
|
|
| Fin |
| quote: | Originally posted by Kuei Yang Wang
What I would suggest you do is to model each supply in Duncan Amplifiers PSU Designer.
The timeconstants in the circuit I propose (1,500uS/102Hz for the first RC and 47,000uS/3.4Hz) are intended to provide a reasonable level of filtering and a "slow down" of the charge pulses. I would keep these about the same but adjust the rest of the circuit.
|
Hi Thorsten
This seems like good software and it produces very nice graphs with colourful traces. It is also quite user friendly. However, I'm not sure what I'm supposed to be looking for.
Which trace should I be looking at, and what should it look like?
I assume it is the trace for the last component in the filter and that it should be as smooth as possible while maintaining the required voltage??
Also, where did you get the time constants from and what should I be doing with them?
Maybe someone could attach an image here of what it is supposed to look like - or even attach a psu file (with the extention changed to one that is valid). |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Which trace should I be looking at, |
All of them, intially at least, for didactic reasons.
| quote: | Originally posted by Fin
and what should it look like? |
I'd suggest you take a few different supplies, maybe a very basic one and vary topology and parts values and observe how the various currents and voltages change. Some items (such as the diode switching noise and LC tank circuits in the mainstransformer) are not modeled and you need to retain the need to damp these in the back of your hear when working with PSUD.
| quote: | Originally posted by Fin
I assume it is the trace for the last component in the filter and that it should be as smooth as possible while maintaining the required voltage?? |
In essence, but other traces (Transformer current, especially waveform) also have their uses.
| quote: | Originally posted by Fin
Also, where did you get the time constants from
|
Timeconstant is R*C and relates inversely to the -3db Frequency. 3180uS equals 50Hz, meaning an RC circuit with a riven R*C giving 0.003180 will have a -3db point of 50Hz. Using a 100Hz (120Hz for USoA) -3db point on the input filter keeps the higher harmonics down and "slows" the current flow from the rectifier, making overall the current draw less "spikey".
Using a much lower -3db Point (say 2.5Hz) for the second RC filter will reduce any 100Hz ripple substantially, as we have a 6db/Octave lowpass. So any (100Hz) ripple on the first cap will, with a 2.5Hz/0.064S Filter, be attenuated by around 32db. Following that with a good regulator (LM1085) will knock any ripple another 75db down.
Now having our example supply (3R3/470u/10R/4,700uF) with a 15V/30VA Transformer and schottky rectifiers supplying around 50mA we get 20.5V across the second capacitor, with 22mV peak-Peak (around 8mV rms) with a fairly close to sinewave waveform (few high order harmonics). Then using a single regulator (LM1085) correctly impemented we drop the ripply by a factor of around 10,000 to 0.8uV which is around two times the self noise of the regulator with 15V Output and 122db below 1V.
It is obvious that with only 22mV P-P ripple and 20.5V raw voltage at nominal our regulator always works with more than 3V across it, avoiding the worse performance "dropout" region. (20.5V - 10% = 18.6V so 3.6V across the Regulator.
So, with a simple RCRC Filter plus regulator we have made a supply with -122dbv noise level.
Had we omited the first RC and second R we would get a higher DC voltage at 22V but also 5 times the ripple and with a much worse waveform (very steep slopes and spikey, not rounded peaks suggesting a lot of high order harmonics. Had we then used a generic 7815 Regulator we would have had at least 10db less ripple rejection on top of 14db more ripple, so our output noise would be around -98dbV (still not bad) but likely with a lot more than a little 100Hz ripple, most likely a lot of higher order components would have made through the regulator.
Also, the conduction angle for the direct 4,700uF Capacitor is quite low, current is conducted only for 1mS within the whole 10mS halve cycle of the 50Hz mains wave with nearly 0.65A peak and a fairly "peaky" peak. The RCRC circuit has a much wider conduction angle, conducting almost 3mS out of 10mS of the mains halve wave and with a current peak < 0.35A and very gentle slopes.
Try following the above in PSUD and I think you will get a feel for what things mean.
Sayonara |
|
|
| Fin |
Hi Thorsten
Thanks for the lessons - I'm now getting results that look meaningful. And it's starting to make sense!
| quote: | Originally posted by Kuei Yang Wang
I'd suggest you take a few different supplies, maybe a very basic one and vary topology and parts values and observe how the various currents and voltages change. |
OK - I've done this - and it is really interesting!
Obviously many combinations can give a similar result.
Is the optimum for the first RC achieved by using a small capacitor/large resistor combination, that still obtains 1500uS, a smooths waveform and the required voltage?
| quote: | Originally posted by Kuei Yang Wang
In essence, but other traces (Transformer current, especially waveform) also have their uses. |
Also interesting - now that I'm using the right variables
| quote: | Originally posted by Kuei Yang Wang
Timeconstant is R*C and relates inversely to the -3db Frequency.
3180uS equals 50Hz, meaning an RC circuit with a riven R*C giving 0.003180 will have a -3db point of 50Hz. Using a 100Hz (120Hz for USoA) -3db point on the input filter keeps the higher harmonics down and "slows" the current flow from the rectifier, making overall the current draw less "spikey". |
OK - I understand the concept - but you must have obtained the 3180uS elsewhere.
Does this mean it is 1590uS for a -3db point of 100Hz?
Are there tables (or similar) with these values?
| quote: | Originally posted by Kuei Yang Wang
Using a much lower -3db Point (say 2.5Hz) for the second RC filter will reduce any 100Hz ripple substantially, as we have a 6db/Octave lowpass. So any (100Hz) ripple on the first cap will, with a 2.5Hz/0.064S Filter, be attenuated by around 32db. Following that with a good regulator (LM1085) will knock any ripple another 75db down. |
OK - I can see where the 0.064S comes from, having started with 3180uS for 50Hz, however, I might have some other questions about this later - just have to think of them first.
| quote: | Originally posted by Kuei Yang Wang
Now having our example supply (3R3/470u/10R/4,700uF) with a 15V/30VA Transformer and schottky rectifiers supplying around 50mA we get 20.5V across the second capacitor, with 22mV peak-Peak (around 8mV rms) with a fairly close to sinewave waveform (few high order harmonics). Then using a single regulator (LM1085) correctly impemented we drop the ripply by a factor of around 10,000 to 0.8uV which is around two times the self noise of the regulator with 15V Output and 122db below 1V. |
OK - I've tried to model this and got values quite close to those you quote. Getting there!
This arrangement seems to work for a current draw up to 115mA - then the voltage at C2 starts to drop below about 18.5V. If more current is required, the obvious solution is to use a larger transformer or reduce the values of the resistors while increasing the values of the capacitors?
| quote: | Originally posted by Kuei Yang Wang
It is obvious that with only 22mV P-P ripple and 20.5V raw voltage at nominal our regulator always works with more than 3V across it, avoiding the worse performance "dropout" region. (20.5V - 10% = 18.6V so 3.6V across the Regulator. |
Is it better if you start with a slightly higher voltage or is that just a waste?
| quote: | Originally posted by Kuei Yang Wang
Had we omited the first RC and second R we would get a higher DC voltage at 22V but also 5 times the ripple and with a much worse waveform (very steep slopes and spikey, not rounded peaks suggesting a lot of high order harmonics. Had we then used a generic 7815 Regulator we would have had at least 10db less ripple rejection on top of 14db more ripple, so our output noise would be around -98dbV (still not bad) but likely with a lot more than a little 100Hz ripple, most likely a lot of higher order components would have made through the regulator. |
Well - that's what is in the CDP now - so the new PS should be a big improvement.
| quote: | Originally posted by Kuei Yang Wang
Also, the conduction angle for the direct 4,700uF Capacitor is quite low, current is conducted only for 1mS within the whole 10mS halve cycle of the 50Hz mains wave with nearly 0.65A peak and a fairly "peaky" peak. The RCRC circuit has a much wider conduction angle, conducting almost 3mS out of 10mS of the mains halve wave and with a current peak < 0.35A and very gentle slopes. |
This also makes sense now.
| quote: | Originally posted by Kuei Yang Wang
Try following the above in PSUD and I think you will get a feel for what things mean. |
I did - and I think I've got most of it.
I have attached a few of the psu files (disguised as zip files).
This one is my attempt at modelling Thorsten's description above, ie. 15V/1A TF, 3.3R-470uF-10R-4700uf-50mA. |
|
|
| Fin |
This is a similar arrangement but C2 is reduced to 2,200uF and R2 increased to 20ohm. Is this OK to do?
BTW - after you have downloaded the file, just delete the ".zip" extension. |
|
|
| Fin |
| This one is similar again but it uses a 22V / 568mA TF and a 30mA current draw (1 opamp and regulator). |
|
|
| Fin |
| This one has an extra RC. Not sure if that is a good idea. |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Obviously many combinations can give a similar result.
|
Yes, I tend to look at conduction angle, ripple waveform and step response when comparing different circuits. Obviously the output voltage needs to right and ripple low, but it is IMNSHO preferable to have a large conduction angle and a bit higher ripple with a "soft" waveform than to minimise ripple at the cost of a unpleasant (spikey - indicates high order harmonics) ripple waveform and narrow conduction angle.
| quote: | Originally posted by Fin
Is the optimum for the first RC achieved by using a small capacitor/large resistor combination, that still obtains 1500uS, a smooths waveform and the required voltage?
|
More or less. All things are tradeoffs. The larger the input resistor and the smaller the input capacitor the wider the conduction angle but equally regulation suffers, maing such a circuit suitable only for "class A" circuits. You can compensate the resistive losses by boosting the primary volateg, but you then need to account for the higher potentioal "no load" voltage in the safety ratings for components.
Thus, for any given pricepoint and level of component quality there will be a "sweet spot" where you get best performance within a given budget and component quality.
| quote: | Originally posted by Fin
OK - I understand the concept - but you must have obtained the 3180uS elsewhere. |
Yes, 3180uS corresponds to 50Hz, a very common EQ value in Tape Head (and LP) playback EQ's, I know the usual EQ Timeconstants very well and also how they relate to frequency.
A good shorthand is to remember that 1uS = 159,115Hz.
| quote: | Originally posted by Fin
Does this mean it is 1590uS for a -3db point of 100Hz?
|
| quote: | Originally posted by Fin
This arrangement seems to work for a current draw up to 115mA - then the voltage at C2 starts to drop below about 18.5V. If more current is required, the obvious solution is to use a larger transformer or reduce the values of the resistors while increasing the values of the capacitors? |
Yup. I'd say increase the Transformer current rating and reduce Filter Resistor Values with increased Filter Capacitor Values.
| quote: | Originally posted by Fin
Is it better if you start with a slightly higher voltage or is that just a waste? |
More Voltage means more reserve for filtering/regulation (increase Input Resistor preferably) but also requires higher safety ratings for components. This is a tradeoff.
Sayonara |
|
|
| Fin |
Hi Thorsten
| quote: | Originally posted by Kuei Yang Wang
Yes, I tend to look at conduction angle, ripple waveform and step response when comparing different circuits. Obviously the output voltage needs to right and ripple low, but it is IMNSHO preferable to have a large conduction angle and a bit higher ripple with a "soft" waveform than to minimise ripple at the cost of a unpleasant (spikey - indicates high order harmonics) ripple waveform and narrow conduction angle.
|
This makes a lot of sense - especially if you can reduce the ripple at a later stage - if necessary. Is there a simple means of determining the conduction angle or length of time that a circuit conducts, and a guide to what one should be aiming for?
Previously, you mentioned that 3ms was better that 1ms - how high can it go (obviously 10ms is max. - and impossible)?
| quote: | Originally posted by Kuei Yang Wang
More or less. All things are tradeoffs. The larger the input resistor and the smaller the input capacitor the wider the conduction angle but equally regulation suffers, maing such a circuit suitable only for "class A" circuits. You can compensate the resistive losses by boosting the primary volateg, but you then need to account for the higher potentioal "no load" voltage in the safety ratings for components.
Thus, for any given pricepoint and level of component quality there will be a "sweet spot" where you get best performance within a given budget and component quality.
|
So - with my 2*22V 25VA transformer - I could increase the value of the first resistor and reduce the value of the first capacitor. However, I would need to maintain the peak to peak voltage at capacitor 2, below about 22mv. Is there any point in maintaining this at a lower level? What would be the minimum capacitor value you would consider?
As you say above, I would also need to ensure that all components before the regulator are rated above the no-load voltage of the rectified voltage. Therefore, I should use 50V capacitors and 5W resistors.
The higher primary voltage also provides the option to include additional filtration. Is it worthwhile adding more RC filters? I noticed that an extra 20ohm/2,200uF RC brings the peak to peak down to less than 1mv.
| quote: | Originally posted by Kuei Yang Wang
Yes, 3180uS corresponds to 50Hz, a very common EQ value in Tape Head (and LP) playback EQ's, I know the usual EQ Timeconstants very well and also how they relate to frequency.
A good shorthand is to remember that 1uS = 159,115Hz.
|
Now it all fits together.
| quote: | Originally posted by Kuei Yang Wang
Yup. I'd say increase the Transformer current rating and reduce Filter Resistor Values with increased Filter Capacitor Values.
|
Does the increase in transformer current rating cause the conduction angle to widen - therefore allowing an increase in the capacitor size - or - is the adjustment in the RC values just to minimise the required increase in the size of the transformer - and therefore, the cost?
This is just hypothetical, as my situation is the reverse - higher voltage and more current than is required.
| quote: | Originally posted by Kuei Yang Wang
More Voltage means more reserve for filtering/regulation (increase Input Resistor preferably) but also requires higher safety ratings for components. This is a tradeoff. |
I know this partly answers some of what I have asked above - but - are you suggesting that, due to the higher voltage of my transformers, I should increase the input resistor? If so, is there a reasonable value that you would not exceed? I assume that the capacitor needs to be reduced in value at the same time - or are you talking about altering the nature of the filter?
If you still have a higher than required voltage, what would you do next? More filtering or more regulators or both? Should I now start to consider pre/post regulators, with each optimised differently?
With the transformers that I have, it looks like there is plenty of reserve for filtering/regulation. I hope that, from a sound quality point of view, this is a good thing. |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Is there a simple means of determining the conduction angle or length of time that a circuit conducts, |
Look at the current through the transformer, zoom in if neccesary, that should make it clear....
| quote: | Originally posted by Fin
and a guide to what one should be aiming for?
Previously, you mentioned that 3ms was better that 1ms - how high can it go (obviously 10ms is max. - and impossible)? |
Choke input and purely resistive load manage close to 180degrees/10mS condcution time (50Hz mains).
| quote: | Originally posted by Fin
So - with my 2*22V 25VA transformer - I could increase the value of the first resistor |
Yes.
| quote: | Originally posted by Fin
and reduce the value of the first capacitor.
|
Yes, but this will increase ripple.
| quote: | Originally posted by Fin
However, I would need to maintain the peak to peak voltage at capacitor 2, below about 22mv. |
Not neccesarily. A little more or less rarely makes a huge difference. I usually apply to any changes in secondary circuits the 6db rule. Unless the improvement is near 6db (or the disimprovement) I don't loose sleep, unless the original result was marginal.
| quote: | Originally posted by Fin
As you say above, I would also need to ensure that all components before the regulator are rated above the no-load voltage of the rectified voltage. Therefore, I should use 50V capacitors and 5W resistors. |
Maybe, you can do a worst case simulation with PSUD and establish the actual values for LIMIT and also look at the operational behaviour.
| quote: | Originally posted by Fin
The higher primary voltage also provides the option to include additional filtration. Is it worthwhile adding more RC filters? |
Usually. If you slit the R and C you get more ripple rejection.
Eg. if instead of 10R/4700uF on our earlier example you use 3R3/2,200uF/3R3/2,200uF/3R3/2200uF after 3R3/470uF on the input you get 20.45V instead of 20.5V, but with only 3.4mV Peak-Peak ripple instead of 22mV....
Even if you change the input RC to 3R3/100uF you still have only 5.2mV Peak-Peak ripple with 20.2V and the conduction angle widens a little more.
If we "design in" your 22V/0.5A Transformer and use 22R/100u/10R/2,200uF/10R/2,200uF/10R/2200uF we get 26V DC with 0.5mV peak-peak ripple. Peak currents (for 50mA DC out) are below 0.2A and the current draw is very "rounded", almost a nice sinewave. nearly 4mS conduction time equal 72degrees conduction angle.
Comparing this (for fun) with a single 4,700uF Capacitor has a conduction angle of 27 Degrees and nearly 0.5A peaks with 32V DC overlaid with 90mV Peak-Peak ripple.
I think this illustrates well the differences.
| quote: | Originally posted by Fin
Does the increase in transformer current rating cause the conduction angle to widen - therefore allowing an increase in the capacitor size |
Yes, usually a larger cap need a larger transformer, but noit because the conduction angle widens, but rather due to issues inherent to transformers.
| quote: | Originally posted by Fin
- or - is the adjustment in the RC values just to minimise the required increase in the size of the transformer - and therefore, the cost? |
To a degree. In the above example our RCRCRCRC filtered supply draws 88mA RMS from the transformer for 50mA out, the single 4700uF Cap causes 133mA RMS to flow and the first circuit limits current peaks to 195mA, the second draws 435mA peaks.
| quote: | Originally posted by Fin
I assume that the capacitor needs to be reduced in value at the same time - or are you talking about altering the nature of the filter? |
You can increase the timeconstant if you like. There are no hard and fast rules, except perhaps the 6db rule. PSUD allows you to try a lot of arrangements, work out two you think most likely build both and listen.... ;-)
| quote: | Originally posted by Fin
Should I now start to consider pre/post regulators, with each optimised differently? |
Depends how far you want to go. Adding a TL/LM431 Shunt after the LM1085/317 series regulator is naturally a good thing, as is adding several RC filter circuits. The less noise the better. Where do you want to draw the line? You can go so far that even with AC powered gear there is no observable ground leakage and PSU noise is down to and limited purely by the active regulation device feeding the signal circuit. It does take a lot of effort to get this far and may not give better sound anyway...
Sayonara |
|
|
| Fin |
Hi Thorsten
Thanks again for taking the time to explain all of this in such detail. It's getting to the point where I'm starting to trust "my" interpretation of what the software is showing.
| quote: | Originally posted by Kuei Yang Wang
Look at the current through the transformer, zoom in if neccesary, that should make it clear....
|
It does - and it's so obvious now! Why didn't I think of that?
Thought it would be much more complex.
| quote: | Originally posted by Kuei Yang Wang
Choke input and purely resistive load manage close to 180degrees/10mS condcution time (50Hz mains).
|
So this is why some people use chokes.
| quote: | Originally posted by Kuei Yang Wang
Maybe, you can do a worst case simulation with PSUD and establish the actual values for LIMIT and also look at the operational behaviour.
|
OK - but I will still play it safe. If I see anything above 30V - I will use 50V components instead od 35V components.
| quote: | Originally posted by Kuei Yang Wang
Usually. If you slit the R and C you get more ripple rejection.
Eg. if instead of 10R/4700uF on our earlier example you use 3R3/2,200uF/3R3/2,200uF/3R3/2200uF after 3R3/470uF on the input you get 20.45V instead of 20.5V, but with only 3.4mV Peak-Peak ripple instead of 22mV....
Even if you change the input RC to 3R3/100uF you still have only 5.2mV Peak-Peak ripple with 20.2V and the conduction angle widens a little more.
|
Yes - I can see this effect.
I was going to ask the following question:-
"Shouldn't the resistor values be increased to maintain the same R*C values - or - has it changed now that we are using multiple filters? To replace the 10R/4,700uF, shouldn't we use 22R/2,200uF (and not 3R3/2,200uF)?
but then read your comment that there is no hard and fast rule. Also noticed in the software, that using the 22R values increases the time for the voltage to level out, to about 7 seconds. Is that a problem?
| quote: | Originally posted by Kuei Yang Wang
If we "design in" your 22V/0.5A Transformer and use 22R/100u/10R/2,200uF/10R/2,200uF/10R/2200uF we get 26V DC with 0.5mV peak-peak ripple. Peak currents (for 50mA DC out) are below 0.2A and the current draw is very "rounded", almost a nice sinewave. nearly 4mS conduction time equal 72degrees conduction angle.
Comparing this (for fun) with a single 4,700uF Capacitor has a conduction angle of 27 Degrees and nearly 0.5A peaks with 32V DC overlaid with 90mV Peak-Peak ripple.
I think this illustrates well the differences.
|
It sure does!
If you increase the first resistor to 50R, this increases the conduction time to almost 5ms and the output voltage is still 24.9V after 7 seconds. Is this a reasonable thing to do or am I going overboard?
| quote: | Originally posted by Kuei Yang Wang
You can increase the timeconstant if you like. There are no hard and fast rules, except perhaps the 6db rule. PSUD allows you to try a lot of arrangements, work out two you think most likely build both and listen.... ;-)
|
The 6db rule:- Trying to achieve a -6db roll-off at a particular frequency - using sucessive filters?? I did say earlier that I would probably have more questions about this.
| quote: | Originally posted by Kuei Yang Wang
Depends how far you want to go. Adding a TL/LM431 Shunt after the LM1085/317 series regulator is naturally a good thing, as is adding several RC filter circuits. The less noise the better. Where do you want to draw the line? You can go so far that even with AC powered gear there is no observable ground leakage and PSU noise is down to and limited purely by the active regulation device feeding the signal circuit. It does take a lot of effort to get this far and may not give better sound anyway...
|
Well - I was thinkig about the parts and components that I already have. The 22V 25VA transformer provides higher voltage and more current than needed. I've already got some LM317/337, LM1085 and LM431. To go a bit further than RCRC-Reg and use RCRCRC-LM317/337-LM431, will only require two additional 2,200uF/50V caps (one for +15V and one for -15V). Might be worth it? |
|
|
| Fin |
I started to have a look at the supply for the servos in a little more detail. Assuming that they require +/-12V and about 1A, there are two transformers to choose from. One is 2*12V 30VA (12V@1.25A + 12V@1.25A) and the other is 2*15V 30VA (15V@1A + 15V@1A).
Each of these were used in modelling the supplies in PSUD - with interesting results. The higher voltage/lower current transformer can actually provide more current to the supplies than the transformer with the higher current rating.
The 12V/1.25A unit can supply 1A, but the voltage drops to 13.4V - too low to be regulated to 12V. To stay above the dropout voltage of most regulators, only 600mA can be provided at 15V.
The 15V/1A unit can supply 1A, and the voltage only drops to 16.2V. It can provide 1.2A at a voltage of 15.2V - enough to regulate to 12V.
So - in this situation, a 15V transformer can supply twice the current of the same sized (30VA) 12V transformer. |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Assuming that they require +/-12V and about 1A |
How do you arrive at +/-12V/1A? Have you checked the Motor Driver IC's used? What do they have as current limits (most are short circuit proof)?
| quote: | Originally posted by Fin
Each of these were used in modelling the supplies in PSUD - with interesting results. The higher voltage/lower current transformer can actually provide more current to the supplies than the transformer with the higher current rating.
The 12V/1.25A unit can supply 1A, but the voltage drops to 13.4V - too low to be regulated to 12V. To stay above the dropout voltage of most regulators, only 600mA can be provided at 15V. |
Yup. However in most cases high current demands from the drive will be transient in nature, so the lower voltage, higher current transformer with large value reservoir capacitors may still be the better choice.... ;-) Try calculating realistically the regulation of the transformer and make sure you use Schottky Diode (models for simulation).
Sayonara |
|
|
| Fin |
Hi Thorsten
There is obviously more to it than I thought.........
| quote: | Originally posted by Kuei Yang Wang
How do you arrive at +/-12V/1A? Have you checked the Motor Driver IC's used? What do they have as current limits (most are short circuit proof)? |
It was just an approximation (or guess) to get an idea of what might be needed and to compare the transformers. Will look more closely at it now.
| quote: | Originally posted by Kuei Yang Wang
Yup. However in most cases high current demands from the drive will be transient in nature, so the lower voltage, higher current transformer with large value reservoir capacitors may still be the better choice.... ;-) Try calculating realistically the regulation of the transformer and make sure you use Schottky Diode (models for simulation). |
I've used the regulation figure supplied by the manufacturer but I need to alter things for Schottky Diodes. Results comimg soon.
BTW - do you have any comments on Post #221 |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
There is obviously more to it than I thought.........
|
Always is.
| quote: | Originally posted by Fin
It was just an approximation (or guess) to get an idea of what might be needed and to compare the transformers. Will look more closely at it now. |
First order approximations are fine for feasibility studies, for DESIGN you need real numbers.
| quote: | Originally posted by Fin
BTW - do you have any comments on Post #221
|
You mean this paragraph?
| quote: | Originally posted by Fin
Well - I was thinkig about the parts and components that I already have. The 22V 25VA transformer provides higher voltage and more current than needed. I've already got some LM317/337, LM1085 and LM431. To go a bit further than RCRC-Reg and use RCRCRC-LM317/337-LM431, will only require two additional 2,200uF/50V caps (one for +15V and one for -15V). Might be worth it? |
I agree with it, except stick to LM317 Regs, as you need to / should keep the rails seperate untill you get to the common kelvin return (often inaccuratly called Star Ground) it matters nor what "sex" your regulator is. Usually the N-Channel/NPN types are better!
So go for RC\RC\RC\LM317\LM334(CCS)\LM/TL431 by all means. It pretty much matches "what I would do", if space, complexity and cost are not much of an issue.
Place the TL431 local next to the supplied pins and the CCS (LM334 for up to 10mA, otherwise FET's, IXYS adjustable Current Regs etc...) per 431 on the main add in supply board. Add some ferrite beads and RF chokes to the CCS and if possible cascode the CCS. Another RC circuit prior to the CCS may also be beneficial. Also, make sure to split each and every filter resistor into 2pcs, one in the positive line and one in the negative line.
As you can see the amount of effort and complexity in PSU's is nearly unlimited. If you go RC\RC\RC\REG\RC\CCS\SHUNT you can be pretty certain that your PSU noise and impedance is limited by the shunt only, all other noise should be minimal. Adding a further transistor to the shunt (high speed PNP type - see TI datasheet fig 21) can increase loop gain and drop the output impedance.
Also make sure to bypass the Adj/Cathode Resistor with a suitable Capacitor (single Os-Con is fine as this remains capacitive beyond the bandwidth of the TL431) to keep the shunt impedance low. Of course you might use a shunt regulator with a Op-Amp and suitable shunt transistor and so on....
Sayonara |
|
|
| Fin |
Hi Thorsten
| quote: | Originally posted by Fin
......... but I need to alter things for Schottky Diodes. Results comimg soon. |
Finally figured out that the last two diodes in the dropdown list in PSUD are Schottky types. Using these in the simulation makes quite a significant and clearly visible difference to the shape of the waveform and the peak current values - thanks.
| quote: | Originally posted by Kuei Yang Wang
First order approximations are fine for feasibility studies, for DESIGN you need real numbers. |
Working on it now. This has turned out to be a bit more than just copying someone else's handywork, but it is also a great learning exercise.
While we are still on the supplies for the servo's, and if any one is looking for the larger value capacitors, I noticed that Percyaudio.com has Nichicon KG Gold Tune in 10,000uF/35V snap in type for US$4.95. I hope these are suitable as the price is acceptable (for me anyway). Nichicon KG
| quote: | Originally posted by Kuei Yang Wang
I agree with it, except stick to LM317 Regs, as you need to / should keep the rails seperate untill you get to the common kelvin return (often inaccuratly called Star Ground) it matters nor what "sex" your regulator is. Usually the N-Channel/NPN types are better! |
Understood - I think! This means they will both be 15V supplies until they connect to the "ground" on the main PCB - where one supply will be connected in reverse?
| quote: | Originally posted by Kuei Yang Wang
So go for RC\RC\RC\LM317\LM334(CCS)\LM/TL431 by all means. It pretty much matches "what I would do", if space, complexity and cost are not much of an issue.
|
Space:- Should be OK, especially if I can place a small board with the regulators directly above the circuit to be supplied.
Complexity:- Still figuring that out. Fortunately, the "horrible" DAC, with it's six supply pins, is very close to the edge of the PCB. There is an easy point where one supply can be connected which could feed all six 5V pins. Alternatively, each pin could be given its own regulator, connected to the supply side of the resistor feeding each pin. This would obviously be more complex - but I suppose it could be possible with a small board positioned directly above the DAC.
Is it worthwhile giving each pin its own 431?
If so, should the local decoulping stay in place?
Cost:- Might have solved this one. The main problem here is the cost (and availability) of quality capacitors. However, I just discovered Handmade Electronics in the US have Nichicon Muse KZ, FG, FX and ES at reasonable prices. Much cheaper that I can get Panasonic FC for - but I hope they are suitable. If anyone is interested, the datasheets are here:- KZ, FG, FX, ES.
| quote: | Originally posted by Kuei Yang Wang
Place the TL431 local next to the supplied pins and the CCS (LM334 for up to 10mA, otherwise FET's, IXYS adjustable Current Regs etc...) per 431 on the main add in supply board. Add some ferrite beads and RF chokes to the CCS and if possible cascode the CCS. Another RC circuit prior to the CCS may also be beneficial. Also, make sure to split each and every filter resistor into 2pcs, one in the positive line and one in the negative line.
|
I need to do some research on the ferrite beads and RF chokes etc. Might need to come back to this later.
Split each resistor:- For example - instead of 10R - 2200uF, you suggest - 5R - 2200uF with another 5R in the line to the other lead of the capacitor?
| quote: | Originally posted by Kuei Yang Wang
As you can see the amount of effort and complexity in PSU's is nearly unlimited. If you go RC\RC\RC\REG\RC\CCS\SHUNT you can be pretty certain that your PSU noise and impedance is limited by the shunt only, all other noise should be minimal. Adding a further transistor to the shunt (high speed PNP type - see TI datasheet fig 21) can increase loop gain and drop the output impedance.
|
I would like to try this. The power supply seems to be the area where most corners are cut - and now I can see why.
| quote: | Originally posted by Kuei Yang Wang
Also make sure to bypass the Adj/Cathode Resistor with a suitable Capacitor (single Os-Con is fine as this remains capacitive beyond the bandwidth of the TL431) to keep the shunt impedance low. Of course you might use a shunt regulator with a Op-Amp and suitable shunt transistor and so on....
|
This last step would get me really confused at this stage. I'll have a serious look at the 431 setup and see if I can get it all to fit neatly.
Hopefully everyone else who reads this discussion appreciates your efforts here as much as I do. |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Working on it now. This has turned out to be a bit more than just copying someone else's handywork, |
Copying only works where you try to get EXACTLY where whaever is copied is. That would mean for a starter identical starting points.
| quote: | Originally posted by Fin
While we are still on the supplies for the servo's, and if any one is looking for the larger value capacitors, I noticed that Percyaudio.com has Nichicon KG Gold Tune in 10,000uF/35V snap in type for US$4.95. I hope these are suitable as the price is acceptable (for me anyway). Nichicon KG
|
I would expect them to work well, after all,as I said, consider the Servos "Audio Power Amplifiers".... ;-)
| quote: | Originally posted by Fin
Understood - I think! This means they will both be 15V supplies until they connect to the "ground" on the main PCB - where one supply will be connected in reverse?
|
Yup.
| quote: | Originally posted by Fin
Is it worthwhile giving each pin its own 431? |
You know what I think of that DAC... ;-)
| quote: | Originally posted by Fin
If so, should the local decoulping stay in place?
|
Decoupling capacitors - yes.
| quote: | Originally posted by Fin
Split each resistor:- For example - instead of 10R - 2200uF, you suggest - 5R - 2200uF with another 5R in the line to the other lead of the capacitor?
|
Yes. This of course works only if the supplies remain seperate untill they are joined at the main kelvin return(s). The advantage is that you introduce a significant impedance also into the ground line, which would normally carry all sorts of leaked noise from the mains. Once you draw out all current loops (intentional and parasitic) in a complete interconnected system it should become clearer. Try this article masquerading as interconnect review:
http://www.sstage.com/articles/pete01.htm
Sayonara |
|
|
| Fin |
Hi Thorsten
Only two questions today.
| quote: | Originally posted by Kuei Yang Wang
You know what I think of that DAC... ;-)
|
Yes - Sounds best with no supplies going to it at all!
| quote: | Originally posted by Kuei Yang Wang
Decoupling capacitors - yes.
|
And the resistors, inductors...?
Do you know of any software like PSUD, that models decoupling or of any easy way to determine the appropriate size of the decoupling capacitors? |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
And the resistors, inductors...?
|
If every pin has it's own shunt regulator, fed via a CCS and suitable inductors and ferrite beads - remove them.
| quote: | Originally posted by Fin
Do you know of any software like PSUD, that models decoupling or of any easy way to determine the appropriate size of the decoupling capacitors? |
Not really, it ain't quite working that way. The TL431 needs a bypass cap of a given minimum size to prevent it oscillating and to avoid a rise of the supply line impedance at high frequencies. You can work that out from the datasheet. The additional bypasing requires complex parasics to be accounted for and to be modelled, this can become complex. Best stick to the system of decoupling often called "tripplets" or whatever new stuff people came up with recently.
Sayonara |
|
|
| Fin |
| quote: | Originally posted by Kuei Yang Wang
If every pin has it's own shunt regulator, fed via a CCS and suitable inductors and ferrite beads - remove them.
|
Any suggestions for the suitable inductors and ferrite beads?
Should they be used on digital and analogue suppy pins (I suppose that again depends on whether the analogue pins are truely analogue - like your comment about the Os-Cons for decoupling the analogue pins)? The existing arrangement only uses inductors on the digital supply pins - not on the "analogue" supply pins.
| quote: | Originally posted by Kuei Yang Wang
Best stick to the system of decoupling often called "tripplets" or whatever new stuff people came up with recently.
|
Thanks |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Fin
Any suggestions for the suitable inductors and ferrite beads?
|
High resonance frequency for the inductor and a lot of stated impedance for the ferrite bead at low frequencies.
| quote: | Originally posted by Fin
Should they be used on digital and analogue suppy pins
|
Digital - I'd condier them mandatory. Ananlogue you can usually manage without, except they are components of the "penny" type, so why skimp? At audio frequencies their effect is swamped by the CCS , but higher up they will block noise, no matter what the source.
Sayonara |
|
|
| Bricolo |
| any recommendation for the CCS? |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Bricolo
any recommendation for the CCS? |
Depends on many factors, like current, headroom voltage and so on. It may range fron the J5XX series FET CCS's over LM334 and LM1085 as CCS, cascoded versions of these and discrete circuits like C4S and so on.
Sayonara |
|
|
| Bricolo |
| for a dac drawing 60mA (ex: the 1543), the fet and 334 can't be used :( |
|
|
| Werner |
Don't know if this has been answered since, I haven't read the whole thread yet.
Bricolo got confused on where to connected the ground leg(s) of a shunt regulator: near to the bridge, or near to the load.
| quote: | Originally posted by Bricolo
do you mean, the shunt+cap's ground should be placed at the chip's ground? so no direct connection to the bridge?
I don't understand the last one :( |
The crux of shunt regulation is that is makes the (variable) load current run in a loop. This loop is formed by the load circuit, the shunt element, and any traces connecting these at the top (supply line) and bottom (signal ground potential).
The physical area of this loop must be as small as possible.
Hence.
If the shunt element is 'perfect' then the lines that connect its top to the supply voltage and its bottom to the supply null only carry DC current.
(As an aside: the ground node of the lower output voltage setting resistor of an LM317-type of regulator also must be connected with its own private trace to the load's local signal ground, as the bottom of said resistor constitutes the reference for the reg. See ALW's various postings on this.) |
|
|
| Dave S |
but it still lacks the "analogueness" of the Arcam!
Here are my notes:
-------------------------------------
CD723 Upgrade
Modification Sequence (intended to address the biggest problems first, so that later mods can be properly judged. Listening to be conducted at each stage to assess the sonic impact):
1. Upgrade PSUs
a) +/- 15V for I/V stage
b) +5.4V for DAC (gives 4.9V on DAC after 100R/220uF filter)
c) +12V for Servos
d) +5V for VAM1201 signal processing
e) +5V for clock
2. Modify circuits
a) Fit Tent Clock
b) Adjust for 2mA FSC
c) Change o/p filter from CF to KYW circuit
d) Try AD812
e) Output null circuit to remove o/p caps (or maybe use BG caps and leave the offset (single ended class A?)
f) Removal of muting transistors (and maybe replace by relay)
3. Upgrade components
a) BG NX Hi-Q for Vref
b) Silmic decoupling caps for I/V
4. Mechanical Mods
Implementation
1a and b) 20VA 0-15, 0-15 Avel Toroid with electrostatic screen. 11DQ090 rectifiers generating 2 identical rails with 3R3 – 470uF (FC) – 10R – 2,200uF (YXF) – connector to local reg board which contains 3 X LM317. Each LM 317 set for 46mA quiescent with 220uF FC to decouple each ref pin. Still using decoupling on audio board, 220uF 25V FC for I/V, 470uF 25V FC for DAC, 39uF 35V FC for DAC Vref.
Listening
1a) More detail and noticeably cleaner sounding (compared to 7815/7915 Audionics PSU). Still somewhat thin and sibilant but good separation between instruments.
1b) Much of the thinness has gone, the sound has much more body and greatly improved tonality and texture of voices and instruments. Increased detail and tunefulness on bass, treble seems less distinct however percussion sounds tighter and more realistic. Slightly mechanical quality to the music and maybe a bit dull.
Note: Concerned about sharing of OV line for +5 and +15 V lines which means that the DAC supply current (5mA) returns through approx 3cms of shared CAT5 wire. This will impact regulation of the +15V supply.
1c) Seemed to give more scale and control to the music. Not a massive difference but certainly not worse.
1d) Restored some of the top end and some liveliness. Quite a significant and fundamental improvement in sound. Might be worth investigating an even better PSU. Still concerned over 1b and whether or not it’s a real step forward.
Tried reversing 1b, which surprisingly made no or minimal change to the SQ (surprising because when it was first implemented it made a very noticeable difference). Noticed at this point that the tracking ability on some CDRs was poor. To improve the tracking ability I reinstalled the two 3R3 resistors in series with the supply lines to each servo IC. This made no difference to the tracking ability or the sound.
Changed to 2c using 120R and 13nF as output filter. This gave much clearer SQ, in retrospect the CF version is grainy and coloured.
2d. Left feedback resistor at 3K3 since with the existing bias resistor values this seems to give 2mA FSD.
1e + 2a Fitted Tentclock with the recommended PSU. Needed to reduce the bias resistors in the regulator by 10X to obtain the correct o/p voltage and decent load reg. SQ improved further, better timing and rhythms and generally cleaned up and more coherent sound.
2d. Fitting the AD812 seemed OK until connected into the system. Faint whisling can be heard during play. Instability seems to be occurring. Need to investigate further. Tried NE5532 again – nowhere near as good as the RC4227. NE5532 is closed in and thin sounding.
------------------------------------------- |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Dave S
but it still lacks the "analogueness" of the Arcam!
|
I note you have one of the old TDA1541 based mashines. In a different thread I pointed out that the TDA1541 has a "singular sound", I mean it. NOTHING ELSE sounds like it. I don't care if you have a cheap all plastic Philips CDP with TDA1541 or the latest Marantz CD-7 or a Marantz DA-12. They all share exactly this "analogueness" (they do differe very much on the rest though).
So, don't expect (ever) to match the old Arcam in this department.
| quote: | Originally posted by Dave S
Here are my notes:
|
Excellent work (except the 812.... Make sure the PSU Bypass is soldered directly to the chip and is low inductance.
So, one question.
Is your current 723 still mediocre, or is pretty decent but not a match for the Arcam?
| quote: | Originally posted by Dave S
Tried NE5532 again – nowhere near as good as the RC4227. NE5532 is closed in and thin sounding. |
The RC4227. Rarely saw it discussed, a quite nice diual Op-Amp. For me oversahdowed by real high speed stuff, but it deserves more mention/note for those people who seem to be "unlucky". To wit, I practically NEVER humming amplifiers, oscillating amplifiers, oscillating Op-Amp's, you name it. I just seem to get all such things most people call "dicey" to work first time, reliably, go figger.
Sayonara
PS, RC4227 is a Dual OP-27 IIRC.... |
|
|
| Dave S |
Firstly, I would like to thank KYW for the serious amount of time he's spent
helping Fin and me. I have seen a lot of threads where KYW comes under
attack for voicing his opinion, well as far as I'm concerned all the things
I have tried on his suggestion have been beneficial (and believe me I tried
many suggestions here and elsewhere on the lovely WWW and gone many steps
backwards!).
You will notice at this point that I have not tried passive trimming of the
output offset to remove the output caps. I'm not sure at this point that I
agree with KYW on this particular point - the dc offset (about 1.5V) as it
stands drifts around by up to 30mV, I don't want drift this on my volume
pot. Also, in the Arcam I tried trimming the offset (from a Ben Duncan
acticle) using LT1021 and although the sound improved slightly from the
emitter follower Arcam used, just removing the whole lot and running a
higher i.e. 3.5V dc offset sounded much better (does this put the opamp in
single ended mode?). Any bias currents added into the inputs of an opamp
must be super clean, as far as I can see.
KYW - To answer your questions:
1) The CD723 is definitely not mediocre now!
2) Is it better than the Arcam? During "conscious" listening i.e. in an
A-B comparison it's probably better. However the real test is "subconscious"
listening (like I'm doing now, whilst typing) and at the current mod status
the CD723 has just got to the stage where I don't have to swop back to the
Arcam after half an hour of so. I occasionally do "unconscious" listening as
well - not useful for judging hifi. As an aside, with injudicious mods I can
easily make the Arcam sound more overtly detailed (more like the CD723) but
it also loses the analogueness and I generally return to almost original
spec (albeit with improved clock and OPA627s throughout).
3) The RC4227 used to be sold by Maplin (in the days when they had some
interesting parts and some of them were even in stock). I think they were
made by Raytheon in their Scottish fab, which is long since gone. They were
probably made on about 300 micron technology with a haggis substrate. Jokes
aside, they are the best sounding duals I have in the bitbox.
4) I will have another go with the AD812, it's the first time I have had
a hooting op-amp and it only does it when connected to the system. I'm just
too lazy to drag my ancient Textronix into the living room to investigate.
5) Early on in this thread you mentioned something about bad sound
relating to Pana FC with Wima polyprop in parallel as output caps. I can
confirm that the player sounds better without the Wimas. I need some caps to
replace the FCs that will fit on the pcb, Sonicaps etc are too big, can you
suggest the best small option - one of the BGs?
Another aside: My friend's Cambridge Audio CD2 is markedly superior to the
Arcam. In another thread someone stated that most >£200 modern CD players
are better that the CD2. If so, please, please, please someone tell me which
one to buy 'cos when I took said CD2 to the local hifi shop it demolished a
Naim CD5i, comfortably bettered a £1600 Unison Research machine. Only an
Accuphase player at £3000 was more detailed but it was also rather
unengaging. This seems to support KYW's view that the TDA1541 does something
special.
Cheers,
Dave |
|
|
| Nuuk |
Dave, reading your comments in the last post, struck a chord with me as I have just been trying out different opamps in my new Gainclone buffers.
It seems that more detail almost equals less musicality in my experience and somebody else confirmed this on the Gainclone forum.
What I found, (flack jacket on) was that the NE5534 was actually preferable to the OPA627 when it came down to musicality. And I know another person who has finished with the 5534's in his modified CD723 after trying many candidates for that job!
And I wonder what opamps that CD2 uses? :att'n: |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Dave S
Firstly, I would like to thank KYW for the serious amount of time he's spent helping Fin and me. |
No sweat. I appreciate the way in which you do try to work out things for yourself and try to actually learn something. In such cases I am always to contribute whatever little knowledge and experience I posess....
| quote: | Originally posted by Dave S
I have seen a lot of threads where KYW comes under
attack for voicing his opinion, |
There are people who simply cannot deal with a well founded and voiced opinion that runs counter to theirs.
| quote: | Originally posted by Dave S
You will notice at this point that I have not tried passive trimming of the output offset to remove the output caps. I'm not sure at this point that I agree with KYW on this particular point - the dc offset (about 1.5V) as it stands drifts around by up to 30mV, I don't want drift this on my volume pot. |
Appreciated. When I made my mod's I figured the +/-10mV drift was "harmless". Never had any audible problems linked to DC on the volume control.
| quote: | Originally posted by Dave S
just removing the whole lot and running a higher i.e. 3.5V dc offset sounded much better (does this put the opamp in single ended mode?). |
No, it does not put it into SE Mode, but it would seem to move you away from the "zero crossing" point. With some class B Op-Amp output stages this will improve the sound, on modern high speed Op-Amp's the output stage needs a lot of quiescent current to get the bandwidth, so it is not much of an issue.
| quote: | Originally posted by Dave S
Any bias currents added into the inputs of an opamp
must be super clean, as far as I can see. |
Absolutely.
| quote: | Originally posted by Dave S
KYW - To answer your questions:
1) The CD723 is definitely not mediocre now! |
My original point was that there is a VERY NICE CD Player trapped in that CD-723 case only waiting to get out. It is not the last word in anything, but given the low cost and fairly easy mods I think it's worthwhile.....
| quote: | Originally posted by Dave S
3) The RC4227 used to be sold by Maplin (in the days when they had some interesting parts and some of them were even in stock). |
Yes, I got them there too. Maplin is absolutely the pit's nowadays, shame, ain't it?
| quote: | Originally posted by Dave S
I think they were made by Raytheon in their Scottish fab, which is long since gone. They were probably made on about 300 micron technology with a haggis substrate. |
And probably made by old scottisch ex-virgins... Well, it does explain a few things....
| quote: | Originally posted by Dave S
Jokes aside, they are the best sounding duals I have in the bitbox. |
Yeah, it was seriously good at the time.....
| quote: | Originally posted by Dave S
Early on in this thread you mentioned something about bad sound relating to Pana FC with Wima polyprop in parallel as output caps. I can confirm that the player sounds better without the Wimas. |
It will sound even better once the FC's are gone too (DC servo or Ofset Trim).
| quote: | Originally posted by Dave S
Sonicaps etc are too big, can you suggest the best small option - one of the BGs? |
You only need to use 2.2uF (if your load is 10k or higher) as coupling Cap, in that you can get ICW (aka "clarity cap") or Ansar Polypropylenes from Cricklewood electronics. I am (as you well know) not a fan of using BG's. If you are actually blocking sufficient DC even WIMA MKS4 tend to sound better (after breakin) to my ears. But you may feel different, so try them. For "open window" type sound try a pair of the Audyn Tinfoil Capacitors (Audiocom has them in the UK).
| quote: | Originally posted by Dave S
Another aside: My friend's Cambridge Audio CD2 is markedly superior to the Arcam. In another thread someone stated that most >£200 modern CD players are better that the CD2. If so, please, please, please someone tell me which one to buy |
Me too. Enquiring minds want to know.....
However, remember the concepts of moral affordability and of pecuniary affordability, and allow for some (a lot?) of wishfull thinking.
| quote: | Originally posted by Dave S
This seems to support KYW's view that the TDA1541 does something special. |
I call them as I hear them. Can't do any more. But tell you what, maybe we should start badmouthing the TDA1541 so we can buy up some more cheaply for our own use? How about a 16 or 32 pcs TDA1541 parallel DAC?
But I'm afraid this particular Schroedingers kitten has well and truely escaped it's Box (as you would of course if you where locked into abox with some nerve gas), maybe we should get Dirk Gently to investigate?
Sayonara |
|
|
| TimA |
| quote: | | I would like to thank KYW for the serious amount of time he's spent |
I'll echo that if I may. It has been very educational.
Tim. |
|
|
| Dave S |
I pretty much agree with Nuuk on this point. A good example is the difference between a CDR copy and the original CD, on first listen I actually thought the copy was better than the original, it seemed to be tighter and clearer with more detail.
After further listening I came to the conclusion that the copy loses a lot of the fullness, scale and ambience info - is fact it's basically lower resolution. The copy sounds more "digital". Interestingly, a CDR played through the Arcam sounds rather like the CD723 playing the original!!!
As another example: I used to be the dissapointed owner of a BMW 3 series. The CD changer in this car was virtually unused because it possessed a remarkable ability to present plenty of "detail" whilst being completely tuneless.
At the opposite extreme I confess to having owned a Linn system, this was very tuneful whilst presenting minimal detail. New or difficult music was easy to follow, however complex music lacked the interesting bits that keep you listening time after time. I nicknamed it the "nursery rhyme effect".
Nuuk - the CD2 does in fact use NE5534s or a 5532. However I don't agree this is a better part than the OPA627. In some systems I can believe NE553X could balance out imbalances elsewhere. I found a similar situation with Siemens stacked film polyester caps sounding better than Sonicaps in a coupling application, the more limited frequency extremes of the Siemens gave a better overall system sound until I improved some other areas (actually I stopped using Proac R2.5 clones with their overblown bass).
Back to the CD723 - the muting transistors will not be going back in the pcb!
Dave |
|
|
| Nuuk |
| quote: | | However I don't agree this is a better part than the OPA627. In some systems I can believe NE553X could balance out imbalances elsewhere. |
Well, I think this is what successful hi-fi is all about. It's a bit like making something to eat and we need the right balance of ingedients.
Some may say that Christmas cake tastes better than plain bread but you wouldn't make a sandwich with cheese and Christmas cake (I hope)! ;)
That doesn't make bread better than Christmas cake or vice versa! |
|
|
| Bricolo |
Just something I was thinking about:
Thorsten explained the conduction angle for the diodes, and said that (in this case ;)) the bigger the better. Because it causes less harmonics on the supply line.
Maybe that's an explanation for the gainclone's beloved sound, despite it's 100uF per rail. |
|
|
| Kuei Yang Wang |
Konnichiwa,
| quote: | Originally posted by Bricolo
Maybe that's an explanation for the gainclone's beloved sound, despite it's 100uF per rail. |
Maybe. I have thought the same thing.
Sayonara |
|
|
| Fin |
Yes - a big thanks to KYW for the huge effort he has put into this thread. I have certainly learnt a lot here - especially about the power supplies.
Thanks to Dave too, for the excellent reports on progress so far, and everyone else who contributed.
I have heard comments that suggest that this will be one of the great reference threads for CDP mods. Even our debates (or arguements) on where to start (CDP, mechanism, DAC, etc) should be useful to some. |
|
|
| Lourens |
Fin wrote:| quote: | | I have heard comments that suggest that this will be one of the great reference threads for CDP mods. Even our debates (or arguements) on where to start (CDP, mechanism, DAC, etc) should be useful to some. [/B] |
IMHO this was an excellent thread indeed. Though i am a recent and not a very active member, limiting my reading mostly to this ‘digital forum’, i enjoyed it very much. Not only because I learned a lot.
I was surprised by the elegant communication. Even the inevitable intrusions which only degrade or disrupt a thread so often, were very rare, and subsequently dealt with definitely or just ignored. :att'n: A very good example, i think, of what another (old, but revived) thread which i just noticed, is all about.
Thanks, Fin, Dave, Bricolo and last but not least KYW, alias Thorsten, a patient guru... Chapeau.
best regards,
Lourens |
|
|
| Fin |
Hi Dave, Thorsten.................
| quote: | Originally posted by Dave S
f) Removal of muting transistors (and maybe replace by relay)
|
What sort of relay is good to use here?
This is from Dave's update in Post #194 | | | |
