Best CD clock upgrade?
- Thread starter hacker
- Start date
James Evans
Bedroom Bodger
I'm constantly amazed how many freebies I manage to get out of analog and TI, especially analog - I've been ordering samples consistently for months with no questions asked. Must cost them a fortune in couriers doing this to all the global bodgers!
It's worth keeping as a test point, but maybe I should re-locate the clock outputs along the RH edge (or maybe front edge?) of the board, to make connection easier - the primary DAC - clock wires should be short and could act as a leaf spring, all one would then need is suitable insulation under the PCB to prevent shorting. The secondary feed is then made with a coax or twisted pair with 0V terminated at the sending end only.
What do you think?
Andy.[/QUOTE]
Clock outputs along the RH edge to make clock wires as short as possible sounds very good.
Mounting: I actually like stand-offs (at least two for such a pcb)--as additional option to lead-wire leaf spring. Because they are electro-mechanical, XOs will perform differently under different damping conditions: I wouldn't be surprised if, in few months time, people would be posting on the sonic difference between sorbothane (not that good, from my experience) or EAR products (better) or aviation-standard dampening material under their flea power. Soldering braid to ground plane is a nice trick, if you have a ground plane in the vicinity of XO position, which might not be the case on some Japanese designs.
L1: final choice would depend on frequencies that appear in circuit. (I am curious what Andy's scope probe will turn up there in both LF and HF: i.e. will this supply also take care of sub 10Hz noise?). Panasonic P9818 (min. 110R at 100MHz) is a good allrounder, but there are some excellent products from Phillips/Murata, which is what Tent probably uses. I would select from a variety of products in line with the frequencies that appear upon measurement, but I think L1 will not be as critical as the local decoupling cap.
Could we substitute BC550 for BC547? (I like the low noise version better. ) BC 547 has wide spread of current gain, which has led to problems for some people (who started a forum at diyaudio to solve it) in setting the voltage of the divider in the Tent schematic for his XO supply (which also uses BC547) to get the 5V +/- 5%.
Can't wait for the boards to become available!
With AD sample program, it actually works: I do usually end up buying larger quantities of the sampled products after testing them. Isn't that what everybody does?
What do you think?
Andy.[/QUOTE]
Clock outputs along the RH edge to make clock wires as short as possible sounds very good.
Mounting: I actually like stand-offs (at least two for such a pcb)--as additional option to lead-wire leaf spring. Because they are electro-mechanical, XOs will perform differently under different damping conditions: I wouldn't be surprised if, in few months time, people would be posting on the sonic difference between sorbothane (not that good, from my experience) or EAR products (better) or aviation-standard dampening material under their flea power. Soldering braid to ground plane is a nice trick, if you have a ground plane in the vicinity of XO position, which might not be the case on some Japanese designs.
L1: final choice would depend on frequencies that appear in circuit. (I am curious what Andy's scope probe will turn up there in both LF and HF: i.e. will this supply also take care of sub 10Hz noise?). Panasonic P9818 (min. 110R at 100MHz) is a good allrounder, but there are some excellent products from Phillips/Murata, which is what Tent probably uses. I would select from a variety of products in line with the frequencies that appear upon measurement, but I think L1 will not be as critical as the local decoupling cap.
Could we substitute BC550 for BC547? (I like the low noise version better. ) BC 547 has wide spread of current gain, which has led to problems for some people (who started a forum at diyaudio to solve it) in setting the voltage of the divider in the Tent schematic for his XO supply (which also uses BC547) to get the 5V +/- 5%.
Can't wait for the boards to become available!
With AD sample program, it actually works: I do usually end up buying larger quantities of the sampled products after testing them. Isn't that what everybody does?
Andrew L Weekes
Reverse Engineer
Well the number of tubes of AD825s and AD8065's I've been through at a three figure sum each leaves me with little guilt over the odd free sample
Andy.
Andrew L Weekes
Reverse Engineer
It's just possible to squeeze another one in between one in between C2 and LED1, if I nudge C2 along a bit - I'll do that later.
To be honest, given the application (DIY) and the minimal transport it's likely to be subjected to, I think 1 stand off would be adequate here. The reality is it's probably impossible in most cases to use the standoff, given the requirement for the clock to be as close to the DAC as is humanly possible, for best results. For a commercial app. I wouldn't hesitate to have two fixings. It's also advantageous to get the clock board with as little seperation as possible to the main board, although an M3 nut as a standoff would work well here.My experience of the innards of most CD players is that the solder fixing is most likely to be acheivable.
Andy.
P.S. BC547C (which is what the schematic should say!) has the same hfe spread as a BC550, noise figure isn't critical at this point of the circuit, but it's pin compatible, so anyone can try it.
Andrew L Weekes
Reverse Engineer
My scope probe will look noisy as hell due to the bandwidth of the 'scope
Some calibrated measurements on a super-reg based version of this supply though gives: -
-120dBv @ 6Hz
-150dBv @ 100Hz
-160dBv @ 1kHz
then we hit the noise floor of my measurement system up to 100kHz.
martin clark
pinko bodger
Guess that'll do then
Andrew L Weekes
Reverse Engineer
In response to James's request for mounting holes, I've looked at it again and have managed the following: -
1. Place 2x M3 mounting holes at the LHS corners of the board
2. Place the primary clock outputs at the far right hand edge
3. Reduce board length by 8mm!
without compromising layout.
Pic's later, I think that will do it for now
Andy.
1. Place 2x M3 mounting holes at the LHS corners of the board
2. Place the primary clock outputs at the far right hand edge
3. Reduce board length by 8mm!
without compromising layout.
Pic's later, I think that will do it for now
Andy.
laverda
pfm Member
Andrew L Weekes said:In response to James's request for mounting holes, I've looked at it again and have managed the following: -
1. Place 2x M3 mounting holes at the LHS corners of the board
2. Place the primary clock outputs at the far right hand edge
3. Reduce board length by 8mm!
without compromising layout.
Pic's later, I think that will do it for now
Andy.
The wizard Weekes.......................
Andy you must belong to the 'Magic Circle' ......being able to perform such 'eye defeating' tricks as this...........let along the skill in manipulating the circuit to perform at its best.
Great stuff
Debbie McGee......is missing out here one thinks!
Andrew L Weekes
Reverse Engineer
Here it is, rev.2...
It's just a crude 3D render, the components are just generated from an extrusion of the outlines, so not truly representative, but it brings it to life a bit better.
Andy.
It's just a crude 3D render, the components are just generated from an extrusion of the outlines, so not truly representative, but it brings it to life a bit better.
Andy.
QuadisQuad
Member
Hello
I've used an LC Audio Clock XO3 I think with an LC Audio PSU on an old Micromega Duo to great effect, I can't compare to others, but can tell you it works very well. Also I removed the mains transformer from the Transports case and this also was a great improvement!.
Richard
I've used an LC Audio Clock XO3 I think with an LC Audio PSU on an old Micromega Duo to great effect, I can't compare to others, but can tell you it works very well. Also I removed the mains transformer from the Transports case and this also was a great improvement!.
Richard
Some questions:
1) Is the AD797 the optimum device for this application? It's not that fast, plus I've heard many times that its stability is marginal. How about something like the AD817?
2) If another device is used, will the 100 ohm resiters at the op-amp's input be neccessary?
3) The three polyester caps - is there a reason not to use say a small value Oscon instead? I wouldn't have thought absolute DC levels would be a problem here.
4) Finally, if a 5v reference was used instead of the LED, could the potential divider at the output be done away with? Just looking at options to simplfy the implementation.
Thanks for any thoughts,
Dan
1) Is the AD797 the optimum device for this application? It's not that fast, plus I've heard many times that its stability is marginal. How about something like the AD817?
2) If another device is used, will the 100 ohm resiters at the op-amp's input be neccessary?
3) The three polyester caps - is there a reason not to use say a small value Oscon instead? I wouldn't have thought absolute DC levels would be a problem here.
4) Finally, if a 5v reference was used instead of the LED, could the potential divider at the output be done away with? Just looking at options to simplfy the implementation.
Thanks for any thoughts,
Dan
martin clark
pinko bodger
Quick answers:
1) This application is designed around the AD797! It is *plenty* fast, and the magic is the the 47pF cap (re-read the early pages of this thread). It enables a *very* low output impedance at frequencies other opamps cannot come close to - effectively it decompensates the opamp and cancels output distortion. The other reason we're using the 797 is its exceptionally low noise contribution; here it adds less noise than a 500ohm resistor! Most other opamps will not come close on this front, esp. FET types given the low impedances this design uses (which also contributes to the result).
Stability is not a problem providing it is taken care of in the design - as it is here, which brings us to:
2) the 100ohm resistors. Strictly no, but they will do no harm and may still be benficial if other bipolar opamps are used. The value is low enough the effect on input noise is negligable, <<1v/SQRT(hz).
3) Film caps have no current leakage noise, unlike electrolytics - although I haven't tested oscons for this. Oscons do work OK if you wish though.
4)Using reference here doesn't really confer any advantage - it's no quieter, it'll be more expensive, and you'll get no visual indication that the thing is working! LEDs are forward-biased PN junctions and, as such, very, very quiet; 6-8uV over the audio bandwidth is typical. The only downside is that the output voltage will drift at -2mV/degC. It's not significant and though a voltage reference would be somewhat better in this respect the drift is not a parameter that is significant to performance here.
If you wanted to use a 5v reference you could, but definitely keep the RC filter components - bandgap references will be more stable WRT temperature than the LED, but noisier! You can do away with the output divider with a 5v reference, but you MUST keep the 100ohm input resistors in that case with the AD797. Adding a 1Kohm resistor to ground would be a good idea too, to draw a slight DC bias current. So in the end you save...one resistor.
Anyway - the idea for this board is to provide a low-power but very high performance regulator for the experimenter.. so experiment!
1) This application is designed around the AD797! It is *plenty* fast, and the magic is the the 47pF cap (re-read the early pages of this thread). It enables a *very* low output impedance at frequencies other opamps cannot come close to - effectively it decompensates the opamp and cancels output distortion. The other reason we're using the 797 is its exceptionally low noise contribution; here it adds less noise than a 500ohm resistor! Most other opamps will not come close on this front, esp. FET types given the low impedances this design uses (which also contributes to the result).
Stability is not a problem providing it is taken care of in the design - as it is here, which brings us to:
2) the 100ohm resistors. Strictly no, but they will do no harm and may still be benficial if other bipolar opamps are used. The value is low enough the effect on input noise is negligable, <<1v/SQRT(hz).
3) Film caps have no current leakage noise, unlike electrolytics - although I haven't tested oscons for this. Oscons do work OK if you wish though.
4)Using reference here doesn't really confer any advantage - it's no quieter, it'll be more expensive, and you'll get no visual indication that the thing is working! LEDs are forward-biased PN junctions and, as such, very, very quiet; 6-8uV over the audio bandwidth is typical. The only downside is that the output voltage will drift at -2mV/degC. It's not significant and though a voltage reference would be somewhat better in this respect the drift is not a parameter that is significant to performance here.
If you wanted to use a 5v reference you could, but definitely keep the RC filter components - bandgap references will be more stable WRT temperature than the LED, but noisier! You can do away with the output divider with a 5v reference, but you MUST keep the 100ohm input resistors in that case with the AD797. Adding a 1Kohm resistor to ground would be a good idea too, to draw a slight DC bias current. So in the end you save...one resistor.
Anyway - the idea for this board is to provide a low-power but very high performance regulator for the experimenter.. so experiment!
Andrew L Weekes
Reverse Engineer
No need for me to add anything to what Martin has said, he's summed it nicely (1v/rt Hz might be a tad high though ).
I would highlight the comment about the 'magic' of 47pf (50pf, ideally) cap. I think many here are going to be *very* surprised if they try listening to the clock without this cap, then add it once familiar.
It's effect is little short of astonishing, sonically, in all the places I've tried it.
This is well an truly a board for experimentation, and I'd urge people to do just that, but I've tried AD817, AD825, AD8065, AD8620, OPA627, some whizzo TI THSxxx something or other and nothing has come close to the AD797 here, but it's not without it's difficulties. I feel we've addressed most of those though, it should work well. If not we'll have fun and learn a lot fixing it
Anyway, I'm happy with the board, the modest size reduction helps lower the costs that little bit. I will run Gerbers soon, and pass them to anyone who wants to get quotes (I'll get quotes from my usual supplier, Minnitron), then we can compare numbers. Before that we'll need a good indication of numbers wanted - there's a direct realtionship to price here.
Meanwhile I have a panic on at present as my server is conspiring against me and I've had a HD meltdown, no data loss (good backups + RAID5) but a PITA to re-install the OS.
F***ing computers!
Andy.
).I would highlight the comment about the 'magic' of 47pf (50pf, ideally) cap. I think many here are going to be *very* surprised if they try listening to the clock without this cap, then add it once familiar.
It's effect is little short of astonishing, sonically, in all the places I've tried it.
This is well an truly a board for experimentation, and I'd urge people to do just that, but I've tried AD817, AD825, AD8065, AD8620, OPA627, some whizzo TI THSxxx something or other and nothing has come close to the AD797 here, but it's not without it's difficulties. I feel we've addressed most of those though, it should work well. If not we'll have fun and learn a lot fixing it
Anyway, I'm happy with the board, the modest size reduction helps lower the costs that little bit. I will run Gerbers soon, and pass them to anyone who wants to get quotes (I'll get quotes from my usual supplier, Minnitron), then we can compare numbers. Before that we'll need a good indication of numbers wanted - there's a direct realtionship to price here.
Meanwhile I have a panic on at present as my server is conspiring against me and I've had a HD meltdown, no data loss (good backups + RAID5) but a PITA to re-install the OS.
F***ing computers!
Andy.
martin clark
pinko bodger
Er, yes, thats only 30dB below UK mains. I meant 1nV!(1v/rt Hz might be a tad high though
If it's any help on the board MOQ I'd take a 100 to hold stock against any future PFM enquiries (will add a PFM sales link on the acoustica page at some point).
Andrew L Weekes
Reverse Engineer
Oscons can be quite leaky, not sure how they compare to Tant's which are quite good in that regard.
Andy.
Fretless Eric
Musketeer Bathos
martin clark said:
Nice one Martin, that should be sufficient enough for a decent price. I suspect though that they'll be gone in no time
I'm up for two straight off the bat, please.
[sorry about the transistor error Andy. Also I'm still smarting from my own HD crash so you have my sympathies. I lost a fair bit though]