Hi,
is there a special reason, why You made the Servo a linear amplifier with passive post-filtering?
The feedback loop of the Opamp is not bandwidth limited, asking for HF noise to enter the I/V input if C11 and C12 are not paralleled with HF-caps.
Besides the feedback resistor RBA5 is of very high value, also feeding noise.
Why not using a ´slow´ OPamp, i.e. integrator in first place?
Were there any issues at start-up? Did the input, i.e DAC-output always start and settle to 0V? I´d expect the voltage at the DAC-output to reach levels of voltage so that the protection diodes of the DAC spring into action.
jauu
Calvin
is there a special reason, why You made the Servo a linear amplifier with passive post-filtering?
The feedback loop of the Opamp is not bandwidth limited, asking for HF noise to enter the I/V input if C11 and C12 are not paralleled with HF-caps.
Besides the feedback resistor RBA5 is of very high value, also feeding noise.
Why not using a ´slow´ OPamp, i.e. integrator in first place?
Were there any issues at start-up? Did the input, i.e DAC-output always start and settle to 0V? I´d expect the voltage at the DAC-output to reach levels of voltage so that the protection diodes of the DAC spring into action.
jauu
Calvin
Not if you use a toroid; toroids are wideband as I'm sure you are aware.
Thus my mention of EI separate-bobbin xformers.
Like I said, all kinds of solutions are available, it's a matter of which compromise you chose. The shield would attenuate high frequency noise and pulses on the mains, but it also increases the capacitance to the mains. What I'm saying is, is that there are some issues with floating mains supplies you must carefully consider. In contrast to just a battery ;-)
jan
Floating supply
I have tried the "floating powersupply" presented in post number 677 by CeeVee on my CEN BJT version (#107) with no good luck!
Still to much capacitive coupling to mains eg. ground, resulting in considerabel distortion. I changed back to a battery supply.
So be carefull with another supply then the suggested battery by EUVL.
I have tried the "floating powersupply" presented in post number 677 by CeeVee on my CEN BJT version (#107) with no good luck!
Still to much capacitive coupling to mains eg. ground, resulting in considerabel distortion. I changed back to a battery supply.
So be carefull with another supply then the suggested battery by EUVL.
For those who need a stable current drain for the likes of PCM1794 :
http://www.edn.com/archives/1998/031398/06di.pdf
(Have not tried myself ....)
Patrick
http://www.edn.com/archives/1998/031398/06di.pdf
(Have not tried myself ....)
Patrick
ditto, I have ordered some chokes as a last ditch effort before just admitting defeat and buying more batteries =) 28x A123 (up to 4 for the bias) takes up a fair bit of space, maybe i'll buy another 4 9v but my charger, although multi-chemistry is optimized for lifepo4
Indeed.
EI Core transformers are far better than toroid ones and R-Core transformers are better that EI ones.
Which is why I suggested a second tandem R-Core transformer without a shield.
Of curse there are various solutions. As far as I can see, my suggestion of two R-Core transformers seems to give the best possible solution (save, of course, batteries with filtered output).
So its not the best possible solution then?
Best #2? How do you 'see' that?
I would be interested in the actual signal/noise ratio of your build power supply.
jan
I'll be able to relate to it should you specify what you consider as best #1.
A power supply doesn't have signal to noise ratio, it has noise figure expressed either in volts or as a ratio to the supply voltage.
YOU said that the floating mains was second best to a battery, not me.
Hey, great, you caught me on a word! From the context it should be clear that I was talking about the S/N ratio of the complete I/V unit supplied by your power supply.
What is disappointing is that the people that actually (have) build these things are constantly countered with speculation.
But, again, build it and prove that your solution is the best. I mean, you DO build things, right? Being at diyaudio
jan
The servo is the way it is because I started with the one in the previous circuit I posted then removed the integrator capacitor because it was a unipolar electrolytic and I first thought that was the reason the servo on the output was locking up (the electrolytic was the right way around for the correct operating point of the servo but was the wrong way round when the servo was locking up) then removed the input filter which was the real reason for the lockup.
After removing the bits that weren't working I ran a spice simulation and found that using the op-amp without the input filter and connecting the op-amp input directly to the SEN output would increase the distortion of the output signal significantly (from -120dB to -60dB) so I gave up trying to servo the output and used the circuit that remained on the input.
I do have a problem with high frequency noise and it may be that some is from the servo circuit. The feedback resistor is high because it gave the best results in the original simulation I did and that simulation didn't include noise. I considered adding the integrator capacitor back again but the servo is working as shown so I didn't bother.
The servo seems to reliably settle on the right operating point. I don't know what the transient looks like as it is getting there. Whilst I was trying to servo the output I saw it lock up with -0.6V on the DAC output and also saw it get to about +5V on the DAC output in a swing the other way. The DAC chip seems to be fairly robust and is still working. I did put a resistor in the current drain in an attempt to protect the DAC.
As I said in my first post, I am an analog novice so the summary of the above is that it's the way it is because I don't really know what I'm doing.
I haven't had time to investigate the noise much yet. I think most is from at least one of the power supplies because when I unplug the circuit the noise stops instantly but the music continues as the capacitors hold the circuit up for a little time. Also if I pull the op-amp out of the socket the noise doesn't seem to change (again the capacitors keep the servo operating point for a short time).
It might be that after I fix the power supply noise I need to go back and redo the servo to get rid of some remaining noise.
Its clear from listening to the circuit settle at startup that there is a sweet spot in operation where a lot of the noise is canceled out and the servo seems to get pretty close. It might be that I could reduce the noise further with a better servo or perhaps with a servo on the output.
There is about 15mV of noise on each half of a balanced pair but they cancel out so the noise on the output is less than I can measure with my multimeter (below 1mV). 15mV is about what I get if I just hold the multimeter leads in the air and I realize I have left the whole thing floating. I'm wondering if I should maybe ground the negative of the supply for the current sink.
I plan to try grounding the current sink supply negative and also remove power from each of the supplies in turn to see if noise is from one in particular.
I also have a couple of big ferrite cores to try out as chokes on the power supply wires.
I took some measurements of noise and distortion with my sound card but they are not that accurate (sound card doesn't have balanced inputs and the resistors in my test lead were not well matched which I think would result in a high 2nd order distortion measurement). The results are currently too embarrassing to post.
Music sounds great though!
No, I did not say such a thing.
Hey, great, you caught me on a word! From the context it should be clear that I was talking about the S/N ratio of the complete I/V unit supplied by your power supply.
What is disappointing is that the people that actually (have) build these things are constantly countered with speculation.
But, again, build it and prove that your solution is the best. I mean, you DO build things, right? Being at diyaudio
The quality of power supply is determined by it's noise and Zout -- both over wide frequency band, from few Hertz to few Giga-Hertz. Those do not change with the application, though some applications are more demanding than others concerning the desired quality of the PSU.
I do build things, however at this stage I'm not interested in building Cen-Sen.
The Power supply design considerations don't dictate building any specific client for the PSU.
A filtered battery may be quieter than the PSU I suggested (and it may not), however I find using batteries inconvenient, therefore I prefer PSUs over batteries, as long as the performance isn't compromised practically.
Any comments on my actual suggested PSU design?
Any suggestions of possibly improving it?
If I may ask, why are you here then?
I also find batteries inconvenient, but found that the (few) alternatives I tried so far were compromised practically (i.e. sounding worse).
I am sure that multiple layers of well chosen transformers/regulators may ameliorate the noise problem and I would be curious to know the result of your experiments - when/if you do them.
Some threads I browse in order to see if I can learn anything, or contribute from my experience.
I will share my findings.