revised 5v regulator
attached is the +/-5.0Vdc regulator which, thanks to abraxalito's help uses fdn304p and fdn327n (seemed most complementary) as output shunt mosfets.
while the current schematic shows 4.7v zeners as the voltage references, i suppose i could use a tl431 (which adds adjustability and increased noise) or a string of leds. any ideas?
~ brad.
attached is the +/-5.0Vdc regulator which, thanks to abraxalito's help uses fdn304p and fdn327n (seemed most complementary) as output shunt mosfets.
while the current schematic shows 4.7v zeners as the voltage references, i suppose i could use a tl431 (which adds adjustability and increased noise) or a string of leds. any ideas?
~ brad.
revised 3.3v regulator
alright, these two pdfs wrap up the revision process. v1.pdf is the complete schematic document, all 12 pages (ouch!).
as always, i welcome questions, comments and suggestions, as i want to consider all avenues before embarking on one of them.
iko, if you're subscribed to this thread, how does our work relating to the shunt regulators look to you?
~ brad.
alright, these two pdfs wrap up the revision process. v1.pdf is the complete schematic document, all 12 pages (ouch!).
as always, i welcome questions, comments and suggestions, as i want to consider all avenues before embarking on one of them.
iko, if you're subscribed to this thread, how does our work relating to the shunt regulators look to you?
~ brad.
Shunt simulation
Yep, that's the guts of the TL431 according to TI's datasheet. It seems to give a fair impression of the impedance curve shown there - I ran it without the shunt at first to check its accuracy.
This FET selected for being the closest approximation to a bipolar I could find which was freely available and cheap. I ran it with a BC327 as well, that was at least 3dB poorer.
The additional caps are tweaks in an attempt to get the best flatness across the band - the low frequenciy impedance is dominated by the active components. Somewhere in the region around 10kHz, they hand over to the capacitors and I've worked to smooth out that transition. This circuit gives 5V output, there would need to be tweaks to cap values for higher output voltages.
Next stage would be to combine it with a real ccs (in this case the ccs is just the 1kV source and a 100k series resistor).
Yep, that's the guts of the TL431 according to TI's datasheet. It seems to give a fair impression of the impedance curve shown there - I ran it without the shunt at first to check its accuracy.
... with a logic-level-gate mosfet... i see no ccs, though i may be blind...
This FET selected for being the closest approximation to a bipolar I could find which was freely available and cheap. I ran it with a BC327 as well, that was at least 3dB poorer.
The additional caps are tweaks in an attempt to get the best flatness across the band - the low frequenciy impedance is dominated by the active components. Somewhere in the region around 10kHz, they hand over to the capacitors and I've worked to smooth out that transition. This circuit gives 5V output, there would need to be tweaks to cap values for higher output voltages.
Next stage would be to combine it with a real ccs (in this case the ccs is just the 1kV source and a 100k series resistor).
digikey and mouser both seem to still stock the pcm1704, though they cost a pretty penny...
~ brad.
PCM1704 = $45/ea. Hope your soldering skills are worthy...
Looks to be quite an improvement. But wait - what current are you running trhough the salas shunt? From memory I haven't seen it running at such low currents as I'm using here (10mA). Given that this MOSFET is only a tiny SOT23 package, its good for no more than 50mA at 5V operation. If you want to run it hot, I could look for a beefier FET...
Power supply inductors
Have you chosen a specific vendor's inductor? It might be prudent to check a couple of things: first, inductors saturate so you don't want to run too much DC current through. Secondly, with no damping you might experience a resonant circuit between your inductor and the following decoupling caps. In such cases, a parallel resistor might be useful to reduce the Q. Or instead of specifying an inductor, use a ferrite bead.
Have you chosen a specific vendor's inductor? It might be prudent to check a couple of things: first, inductors saturate so you don't want to run too much DC current through. Secondly, with no damping you might experience a resonant circuit between your inductor and the following decoupling caps. In such cases, a parallel resistor might be useful to reduce the Q. Or instead of specifying an inductor, use a ferrite bead.
heh... actually it would be the first project i've had to solder smd... though i trust that i can handle the task, given i've had my morning coffee at least an hour before the solder job, and perhaps a nice loupe or magnifying glass.
my 3.3v shunt is running 31.5mA shunt current, so roughly 100mW dissipation, which looks like a 28 degree increase... seems tolerable to me, though i'd be much more comfortable with something in a D2Pak that could handle the power a bit more... for some reason my search efforts on this matter are quite fruitless...
(edit: how about the ndb6020p and ndb6060l from fairchild? heh, if it's this much trouble finding even quasi-complementary logic-level fets i might just switch back to the BJT shunt transistors...)
~ brad.
Last edited:
I/V conversion critique
OK, having had a look at these I/V circuits, I'm gonna stick my neck out and say that from an engineering pov, they look way suboptimal I realise that Nelson's circuits have an excellent reputation for sound, so I'll just mention I'm not critiquing the sound quality, just the circuit design.
Mainly I have issues with the 10nF capacitor hung off the DAC output. It really cannot be helping to get an accurate I/V conversion since it shunts current away from the FET's source when the DAC transitions from one code to the next. Once shunted, this current can't contribute to the output voltage. A secondary point is to check the size of the current loop formed - the capacitor goes to the GND_30V rail. How far around the houses does it have to go before it gets to return to the DAC's power ground? Big current loops in high speed circuits are a no-no since they both radiate and act as efficient antennas.
I'm also a bit dubious that the IRF610 is the best choice for this particular circuit - mainly because its an old-style high voltage device and I wonder whether there aren't better (i.e. higher transconductance) parts available nowadays. But this is just my speculation - perhaps its there for its superb sound
OK, having had a look at these I/V circuits, I'm gonna stick my neck out and say that from an engineering pov, they look way suboptimal
I realise that Nelson's circuits have an excellent reputation for sound, so I'll just mention I'm not critiquing the sound quality, just the circuit design.Mainly I have issues with the 10nF capacitor hung off the DAC output. It really cannot be helping to get an accurate I/V conversion since it shunts current away from the FET's source when the DAC transitions from one code to the next. Once shunted, this current can't contribute to the output voltage. A secondary point is to check the size of the current loop formed - the capacitor goes to the GND_30V rail. How far around the houses does it have to go before it gets to return to the DAC's power ground? Big current loops in high speed circuits are a no-no since they both radiate and act as efficient antennas.
I'm also a bit dubious that the IRF610 is the best choice for this particular circuit - mainly because its an old-style high voltage device and I wonder whether there aren't better (i.e. higher transconductance) parts available nowadays. But this is just my speculation - perhaps its there for its superb sound
PCM1704 variants
Just noticed on Digikey that all the grades are the same price. Be sure to order the -K variant as its the highest spec
digikey and mouser both seem to still stock the pcm1704, though they cost a pretty penny...
Just noticed on Digikey that all the grades are the same price. Be sure to order the -K variant as its the highest spec
Why do you want complementary FETs ? There are no PNP based TL431s! The beauty of a shunt is it works the same on negative supplies as positive ones, it doesn't need to have a mirror image.
D2PAKs are considerably more expensive - if you're worried about power dissipation you can stretch it a bit by putting a resistor into the drain of the FET. I'm not sure how much could be included though - 100R didn't alter the impedance when I tried it, but this only absorbs 10mW at 10mA... Since you want 3.3V and 30mA I'll see how good it looks at that voltage/current combination. I agree, running around 30mA at 3.3V is definitely not a problem in heat terms.
Last edited:
you're absolutely right, removing the cap halved the THD in my spice sim. (down to 0.001%) looks like it's out.
thanks!I'm also a bit dubious that the IRF610 is the best choice ... perhaps its there for its superb sound
i understand your concern. actually, all my experiences with the '610 have shown it to have a very musical sound.
Just noticed on Digikey that all the grades are the same price. Be sure to order the -K variant as its the highest spec
PCM1704U-K is what i was planning on using. is that correct?
i understand what you mean, but it seems infeasible using the ccs-fed shunt topology i am working with, which specifically uses complementary devices...
~ brad.
hahaha, actually it reduced thd by a factor of 5... seems i'm a bit challenged with fractions today.
~ brad.
Simming a DAC
Really curious to know how you simmed the PCM1704 current output? Built up a piecewise linear current source? If so, that's quite a bit of work to get, say a single 1kHz cycle built up, 1.4uS at a time.. Then does the fourier analysis work on a single cycle?
hahaha, actually it reduced thd by a factor of 5... seems i'm a bit challenged with fractions today.
Really curious to know how you simmed the PCM1704 current output? Built up a piecewise linear current source? If so, that's quite a bit of work to get, say a single 1kHz cycle built up, 1.4uS at a time..
Then does the fourier analysis work on a single cycle?
i'm simulating it using a current source outputting a perfect sinewave in parallel with a 1k resistance to ground, so not the most accurate/realistic way i'm sure, but it seems to get the job done ok ... i let the fourier analysis run for ten cycles.
~ brad.
Last edited:
Sure, but who's to say that a more modern Toshiba part will sound any less musical?
Yep, that's the one. Incidentally I've just done a bit more in depth datasheet research to test my theory about running DACs faster
Comparing the THD+N figures between the PCM63-K (predecessor to PCM1704) and your chosen part, we find the PCM63-K delivers better performance at -20dB output. I reckon this is entirely due to the fact that its spec'd at 8fs and the PCM1704 is spec'd at 16fs. If not, then BB took a step backwards in releasing the PCM1704.i understand what you mean, but it seems infeasible using the ccs-fed shunt topology i am working with, which specifically uses complementary devices...
Guess you don't really understand then
Keep the ccs (series) part complementary but just copy/paste the positive side shunt reg to the negative side. Any clearer ?
Definitely not anywhere near accurate
The waveform out of the DAC is in the form of current steps, settling time 200nS. That's my main justification for criticising the presence of the 10nF cap - it rounds off the steps. But that's a real pig to simulate of course... So its really no wonder your sim results look potentially better than the DAC itself...I'm sure Nelson had a reason for including the cap, just can't figure out what it might be. Certainly it won't be necessary when simulating with a perfect sinewave
not sure, but i'd be willing to look into a specific part if you have one to recommend.
much, thank you. i'll have a go at it in the sim just to make sure i really do understand.
Definitely not anywhere near accurate
well what are simulations there for if not to get our hopes up?
you've got me interested now on how to simulate the actual pcm1704... time to do a bit of gear-spinning...~ brad.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.