Glenn2 said:
With a poor reg you need more filtering, because it's bandwidth is poor, but the series 'R' will raise the output impedance. A wideband reg can do with a small L and cap (higher cut-off frequency). Mostly, a big cap is not allowed anyway, because some type of regs can become unstable with a too-low ESR at the output. So, when you have oscillations, sometimes it's better to use a crappy reg
Regards,
Ray
6h5c said:
Unfortunately, the thing with public forums is that neither you nor I know who we are talking to, really.
This BY FAR is not the first CDP I have tweaked, and especially not the first DAC or ADC I've implemented (or in this case re-implemented).
Please note that you have matter-of-factly stated that my approach will pollute the analog supply line, and did not offer any explanation as to how. While it is impossible to avoid this absolutely, this is not a church, so in a practical world, we have to work with relatives. My approach is based on experience and sound theroy, implemented in practise, measured and listened to ad nauseaum. It was tailor-made for this CDP, given it's particulars, and the constraints of how far I was alowed to go in making changes.
Yes, I have used the 'classical' approach you do many times. I am sorry but if you don't trust me if you don't trust measurement and theory as a sound start (but only a start) to a particular approach, then I cannot offer you better arguments save to implement it for yourself, measure it, listen to it and decide.
Now, you may not like the sound of it, as this is, after all, a subjective measure. But, if that turns out to be the case, then do not expect me to just take your word that the culprit is 'polluting the analog power supply', unless you put up some measurements, a theory that fits or a way to avoid it with an equivalent approach, so that it's presumeably improved sound can be judged on some sort of equal ground. Because, without it, I am sorry, this is a knee-jerk reaction.
Also, please keep in mind that I have used LEDs as shunt elements for a reason, one that has to do with your later posts on series regulator instability. I assume your reference to 'output filtering' has to do with this. The reason why my approach can get to very high PSRR is that there are no real or actively simulated elements to lower it - for instance, a series integrated regulator in general tends to have a slightly inductive output, which can oscillate with an unsuitable chosen output filter cap, unless it is damped by a resistor. In my case, the shunts are connected to the local ground plane under the DAC (to establish a proper reference point for them), but fed through simple FET current sources. Both work well up to UHF if layout ind pin length are taken care of. Neither will oscillate unless a whole lot of extra hardware is added on the side. Neither will be affected by the parallel cap on the shunt element, save for the internal resonances of the cap itself (which all caps have), which is why there are multiple caps and in one case a snubber. However, you pay for this with less than ideal long term regulation - but in this case, it is not a signifficant problem. If I was making a measurement ADC, this approach would be completely invalid.
If your reference to output filtering has to do with the analog output filter, then you are making my point for me - more PWM hash will get out to the output through the 'usual way' than through the power lines.
Any regulator between the analog supply and the DAC will pollute the analog supply in some way, simply because these things are not perfect. Not only because they do not measure perfectly, but there's stray inductance and capacitance, unwanted coupling, radiated noise...
The solution is simple: avoid all this by using a separate supply and leave the analog supply for the opamps. Keep it as far away as possible...
I know a good shunt regulator will do a good job (i use them myself now and then), also with LED's as reference. Maybe i'm missing the point here, but what's the point exactly?
BTW, i'm curious how you measure the performance of these things in the UHF range, at -120dB.
Regards,
Ray
The solution is simple: avoid all this by using a separate supply and leave the analog supply for the opamps. Keep it as far away as possible...
I know a good shunt regulator will do a good job (i use them myself now and then), also with LED's as reference. Maybe i'm missing the point here, but what's the point exactly?
BTW, i'm curious how you measure the performance of these things in the UHF range, at -120dB.
Regards,
Ray
6h5c said:
ilimzn said:
Ray,
I think his point is cheap / value-for-money improvements, or lower part count for some other reason. I'm just guessing... all this is way above my head. It's already been shown quite nicely that simply giving each circuit its own psu and regulator gives massive gains in sound quality, but I think you guys are discussing a less brute-force approach to improving the player.
Simon
6h5c said:
Sure - but this is all there anyway, even when the supply and regulator is separate - in the end it ties to a common mains supply. The point is, you have to assess what the weakest link in the chain is, and once you get to sufficient attenuation (below the nosie floor of the DAC is already a challenge due to the PCB design...), the question is, how far into the territory of diminishing returns you want to go, especially if there are other constraints.
Yes, but even so the wiring is already a good 'antenna' especially if the regulator is low impedance. The trick of course is that the DAC itself must see the lowest possible impedance closest possible to the chip. This is actually whay I wanted to use shunts, I put them in place of one of the decoupling caps.
The only problem is, doing what you propose was outside the constraints set for the project (original board, transformer, basic PSU parts, lowest parts count.
In my case the LED _IS_ the shunt regulator. So, no amplification, feedback, stability issues. It just comes down to LED dynamic impedance vs FET current source dynamic impedance, as a 'divider' of sorts. The pint is, put the shunt as close as possible to the chip, but eprate it from it's power supply by a large dynamic impedance, so that all AC currents flow locally in a very short loop close to the chip, makin the ground plane segment under the DAC effectively as close as we can get to a point, without actually unsoldering the chip and cutting the PCB.
Not UHF, VHF - up to 32MHz to be exact. HP3586 selective level meter with tracking generator. Very useful piece of equipment if a bit tedious to use for this purpose.
SimontY said:
Yes, exactly. The point was to use the least you can to get the most you can, given the original transformer, PSU, PCB. Yes, some tracks had to be cut to insert proper voltages, and some decoupling caps moved under the PCB in SMD form, but unless you look carefully, you would not notice anything overly different, save different size and make components in some places. For instance, the FET CCS were inserted into positions originally used by resistors that make up local RC filtering.
Well, at least I am discussing it
One of the things that was not changed, were the 7812/7912 regulators for the analog output stage. Although these are fairly atrocious, ehat makes them passable is the local RC filtering for the OPamps - it hides the nature of the 78/79xx behind the RC filter. However, the character of the power supply then becomes mostly defined by the quality of the local capacitor (the C part of the RC), so no free lunches. It just happened to be simpler for me to put quality caps in there than to rip apart the PCB and add a new PSU. So, once again, this is not an all-out mod. But, even so, the increase in sound quality is pretty amazing.
I'm very interested in this optimal grounding stuff though. I'm still a bit confused about the grounding for the PWM outputs!
Seems as though joining them to the digital ground under the dac (the one that the clock uses) is optimal instead of the the separate L+R spurs that go to the output jacks.
Also Illimz mentions three places (I think) that the digital and analogue grounds meet. Many of the local ground areas (like the one under the DAC) seem joined by jumpers to the ground plane on the top of the board. Surely this creates many more loops?
I'm wondering if all of those jumpers should be removed apart from the one nearest to the digital star ground point. Or maybe just the one near the output jack to help isolate the analogue ground.
Glenn
Seems as though joining them to the digital ground under the dac (the one that the clock uses) is optimal instead of the the separate L+R spurs that go to the output jacks.
Also Illimz mentions three places (I think) that the digital and analogue grounds meet. Many of the local ground areas (like the one under the DAC) seem joined by jumpers to the ground plane on the top of the board. Surely this creates many more loops?
I'm wondering if all of those jumpers should be removed apart from the one nearest to the digital star ground point. Or maybe just the one near the output jack to help isolate the analogue ground.
Glenn
Ok, probably a typo then. You use a high-impedance probe with that?
Well, i'm sorry if I don't seem very enthousiastic, but I was expecting some Grand Finale after all the long posts with theory.
Maybe it's because using shunts near each circuit as local regulators is not new to me (maybe some remember the TL431 discussion that was held elsewhere).
Ray
DOS DOS DOS
It seems I finally have enough courage to dismantle my CDP to place the recently built DOS.
Wish me luck guys.
1 - Open the top lid, remove the back,
2 - Disconnect all the wires from the 4 additional PSU´s and remove them
3 - Disconnect the "Evil" ribbon from the mech
4 - Lift the pcb and turn it over (carefull with the display ribbon)
5 - Remove the +-15v pcb that feeds the opamps
6 - Remove the opamp caps
7 - Remove the DAC digital srayreg to be able to access the DAC output resistors.
8 - Place the pcb back again and fix/glue the DOS
9 - Place the four wires from DAC to DOS input
10 - Turn the PCB over again and solder the DAC -> DOS input wires
11 - Place the DAC digital srayreg and cap and solder again.
Etc Etc
It seems I finally have enough courage to dismantle my CDP to place the recently built DOS.
Wish me luck guys.
1 - Open the top lid, remove the back,
2 - Disconnect all the wires from the 4 additional PSU´s and remove them
3 - Disconnect the "Evil" ribbon from the mech
4 - Lift the pcb and turn it over (carefull with the display ribbon)
5 - Remove the +-15v pcb that feeds the opamps
6 - Remove the opamp caps
7 - Remove the DAC digital srayreg to be able to access the DAC output resistors.
8 - Place the pcb back again and fix/glue the DOS
9 - Place the four wires from DAC to DOS input
10 - Turn the PCB over again and solder the DAC -> DOS input wires
11 - Place the DAC digital srayreg and cap and solder again.
Etc Etc
