The Son of Dork (SOD) project is the result of the diyAudio.com multi-channel preamp initiative (http://www.diyaudio.com/forums/showthread.php?threadid=3517). The project has been split into threads dedicated to discussion of the preamp's different functional units:
- General Discussion (project status, functional requirements)
- Active Circuitry
- Attenuator
- Power Supply
- Connections and Layout (including input switching)
- Chassis and Construction
- Hardware Logic, Controls, Display
- Microprocessor (programming, implementation)
I will be managing the project as a whole, as well as initially overseeing individual threads. For a background on the project, please see the original thread referenced above. A website dedicated to the project will be located at http://dorkus.org/diy/sod.
Let the fun begin!
- General Discussion (project status, functional requirements)
- Active Circuitry
- Attenuator
- Power Supply
- Connections and Layout (including input switching)
- Chassis and Construction
- Hardware Logic, Controls, Display
- Microprocessor (programming, implementation)
I will be managing the project as a whole, as well as initially overseeing individual threads. For a background on the project, please see the original thread referenced above. A website dedicated to the project will be located at http://dorkus.org/diy/sod.
Let the fun begin!
power supply?
so, each preamp module (which includes input switching, attenuation, and gain stages) will have local regulation on-board near the signal circuitry. the type of regulation is up for debate, but i sense a general distaste for active regulators, so probably some sort of passive/shunt regulation will be used.
as for the main DC supply, i think a single very large supply feeding all preamp modules (up to 6) would work fine. some people may object and want separate transformers, rectifiers, etc. for each channel, but that would be a huge pain with 6 channels. besides, i think a single supply with 6x the VA transformer and 6x the supply capacitance with 6x the load has some advantages over 6 individual, proportionately smaller supplies (we can lower output impedance and less ripple for one).
i would like the main supply to be choke-regulated, probably in a traditional choke-input filter config. it has some desirable characteristics (reduced turn-on surge, reduced transient currents on transformer, longer capacitor life), and as long as we can ensure min. current draw for the given inductance and a low-enough resonant frequency i think it should work ok. the superior regulation over a regular capacitor-input filter also will allow for much more efficient operation of the down-stream shunt regulators. thoughts?
so, each preamp module (which includes input switching, attenuation, and gain stages) will have local regulation on-board near the signal circuitry. the type of regulation is up for debate, but i sense a general distaste for active regulators, so probably some sort of passive/shunt regulation will be used.
as for the main DC supply, i think a single very large supply feeding all preamp modules (up to 6) would work fine. some people may object and want separate transformers, rectifiers, etc. for each channel, but that would be a huge pain with 6 channels. besides, i think a single supply with 6x the VA transformer and 6x the supply capacitance with 6x the load has some advantages over 6 individual, proportionately smaller supplies (we can lower output impedance and less ripple for one).
i would like the main supply to be choke-regulated, probably in a traditional choke-input filter config. it has some desirable characteristics (reduced turn-on surge, reduced transient currents on transformer, longer capacitor life), and as long as we can ensure min. current draw for the given inductance and a low-enough resonant frequency i think it should work ok. the superior regulation over a regular capacitor-input filter also will allow for much more efficient operation of the down-stream shunt regulators. thoughts?
Power supply considerations
Shunt regulators are inherently inefficient as are heavily biased Mosfet stages. Efficiency of the unreglated power supply seems like a moot point in consideration that most of power dissapation will be in the gain modules and local supply regulation when using series or shunt regulators. Your desire to drive 600 ohm loads will call be some fairly highly baised mosfet drivers in the gain modules. I would plan on some very good heatsinking and ventilation.
H.H.
Shunt regulators are inherently inefficient as are heavily biased Mosfet stages. Efficiency of the unreglated power supply seems like a moot point in consideration that most of power dissapation will be in the gain modules and local supply regulation when using series or shunt regulators. Your desire to drive 600 ohm loads will call be some fairly highly baised mosfet drivers in the gain modules. I would plan on some very good heatsinking and ventilation.
H.H.
i would like the circuit to be capable of driving 600 ohm loads, but in practice it will not be driving anything near that low a load (at least in my system). my front amp has 100k input impedance, my rear amp 10k, and i doubt i will be hooking up anything lower than that. i just feel that low impedance drive capability translates to better dynamic performance into normal impedances. let's assume the final target load is 10k however.
Given that this is a preamp design, my feeling is that a watt or two here and there aren't going to be a problem.
By all means, burn an extra watt if it makes it sound better.
For a power amp...that's another question entirely. Yes, I'm willing to go the extra bit to do class A with all the inherent waste heat, but it may not be to everyone's liking.
I know Jam is hot (pun intended) on the shunt regulator idea, but I have no opinion, myself, not having tried one against a series regulator. If the active circuitry ends up being discriminating enough, it might provide an interesting test bed for shunt vs. series regulation, but that would imply a separate regulator (not necessarily at great distance; surely within the same chassis) that could be swapped in and out.
That goes back to how modular you want this thing to be.
Grey
By all means, burn an extra watt if it makes it sound better.
For a power amp...that's another question entirely. Yes, I'm willing to go the extra bit to do class A with all the inherent waste heat, but it may not be to everyone's liking.
I know Jam is hot (pun intended) on the shunt regulator idea, but I have no opinion, myself, not having tried one against a series regulator. If the active circuitry ends up being discriminating enough, it might provide an interesting test bed for shunt vs. series regulation, but that would imply a separate regulator (not necessarily at great distance; surely within the same chassis) that could be swapped in and out.
That goes back to how modular you want this thing to be.
Grey
To obtain a truely modular system, the power supply could be variable to allow for those who wish to (say) try out op-amp circuits as well as higher voltage discrete circuits (wouldn't be me, but there are obviously those who might).
Perhaps a variac in there? Wouldn't have to be too expensive, as small surplus variacs are widely available very cheaply. How that would fit with a regulated supply is moot, though (but presumably would just require some tweaking of component values).
It would have the benefit (if it is possible to still achieve the desired other characteristics) of giving an adaptive platform for future "plug and play" experimentation with gain stages.
Just a suggestion. Seems like if the rest of the pre-amp is going to be as ambitious as it seems, it would be a waste not to ensure that the all the peripheral (control, etc) design and build effort could still be used if the "newest and bestest" gain stage uses a different supply voltage (within reason, of course).
Jake
Perhaps a variac in there? Wouldn't have to be too expensive, as small surplus variacs are widely available very cheaply. How that would fit with a regulated supply is moot, though (but presumably would just require some tweaking of component values).
It would have the benefit (if it is possible to still achieve the desired other characteristics) of giving an adaptive platform for future "plug and play" experimentation with gain stages.
Just a suggestion. Seems like if the rest of the pre-amp is going to be as ambitious as it seems, it would be a waste not to ensure that the all the peripheral (control, etc) design and build effort could still be used if the "newest and bestest" gain stage uses a different supply voltage (within reason, of course).
Jake
10k load indeed.
... at least that's what my system requires.
the 600 ohm thing was just sorta part of the wish list. but that is a lot more current to source than with 10k, and the added requirements are probably not worth it in practice. so yes, 10k target. do any of you drive 600 ohm loads? i guess some pro baluns and such have that sort of input impedance, but in a home system it is probably rare (i know of only one amp that has such an input impedance, and it is using an input xformer).
so ok, 10k load, but should still have good distortion performance down to about, oh, i dunno, 4k load or so. just so we know we are not operating on the margin. oh yeah, and 24V or 25V rails seem about right. i don't really see any need for anything above 30V, unless there is a clear-cut performance advantage. i guess the JFETs may have slightly lower distortion at those higher-voltage operating points but the difference is probably negligible.
... at least that's what my system requires.
the 600 ohm thing was just sorta part of the wish list. but that is a lot more current to source than with 10k, and the added requirements are probably not worth it in practice. so yes, 10k target. do any of you drive 600 ohm loads? i guess some pro baluns and such have that sort of input impedance, but in a home system it is probably rare (i know of only one amp that has such an input impedance, and it is using an input xformer).
so ok, 10k load, but should still have good distortion performance down to about, oh, i dunno, 4k load or so. just so we know we are not operating on the margin. oh yeah, and 24V or 25V rails seem about right. i don't really see any need for anything above 30V, unless there is a clear-cut performance advantage. i guess the JFETs may have slightly lower distortion at those higher-voltage operating points but the difference is probably negligible.
I think that you should be able to run your output stage at a high bias point. At what voltage level should you be able to drive 600Ohm??
The internal stages (before the attenuator) does not need to drive less than 2 - 10KOhm.. So at this stage this would not be a problem.
Harry : you have wrote "Your desire to drive 600 ohm loads will call be some fairly highly baised mosfet drivers in the gain modules"
What do you mean by that mA's or A's
Sonny
The internal stages (before the attenuator) does not need to drive less than 2 - 10KOhm.. So at this stage this would not be a problem.
Harry : you have wrote "Your desire to drive 600 ohm loads will call be some fairly highly baised mosfet drivers in the gain modules"
What do you mean by that mA's or A's
Sonny
choking on the input
Dorkus, (and everybody else)
I think a choke input filter is an excellent idea for all of the reasons you mentioned and then some. In my system(s) I use choke input filters. If you want REAL quiet, use a two section LC filter. The second inductor is not critical, but the more L the better as long as you do not get too much DC drop across it. Then forget the active (noisy) regulator. One possible problem with not having a regulator is that the DC rail voltage varies with changing input AC voltage. But the choke input filter is a low pass filter and will filter out fast voltage changes (noise and transients). One way around the varying DC voltage problem is to use push-pull circuitry so that the varying DC voltage cancels out. The next way is to use enough L and C in your PSU so that any change in the DC voltage is subaudible and use blocking capacitors between stages that will not pass the subaudible stuff. (I use both both methods) Then I suggest that you put another inductor between each module and the PSU. Each module would have it's own shunt cap after the inductor. The latter L and C will keep the stages from "talking to each other". By using this approach your DC line will probably be far quieter (in the audio range) than if you use active devices. If you do need an active regulator, I suggest that you put it after the PSU and before the last mentioned L and C section for noise reduction. I also suggest using "voltage refernces" which are usually quieter than zeners.
I dont't think a choke input supply makes down stream regulators more efficient. Efficiency is power (voltage in this case since current in = current out) out over power in.
Hammond sells chokes and chokes sometimes show up on auction sites like ebay.
Good luck and keep us posted.
Dorkus, (and everybody else)
I think a choke input filter is an excellent idea for all of the reasons you mentioned and then some. In my system(s) I use choke input filters. If you want REAL quiet, use a two section LC filter. The second inductor is not critical, but the more L the better as long as you do not get too much DC drop across it. Then forget the active (noisy) regulator. One possible problem with not having a regulator is that the DC rail voltage varies with changing input AC voltage. But the choke input filter is a low pass filter and will filter out fast voltage changes (noise and transients). One way around the varying DC voltage problem is to use push-pull circuitry so that the varying DC voltage cancels out. The next way is to use enough L and C in your PSU so that any change in the DC voltage is subaudible and use blocking capacitors between stages that will not pass the subaudible stuff. (I use both both methods) Then I suggest that you put another inductor between each module and the PSU. Each module would have it's own shunt cap after the inductor. The latter L and C will keep the stages from "talking to each other". By using this approach your DC line will probably be far quieter (in the audio range) than if you use active devices. If you do need an active regulator, I suggest that you put it after the PSU and before the last mentioned L and C section for noise reduction. I also suggest using "voltage refernces" which are usually quieter than zeners.
I dont't think a choke input supply makes down stream regulators more efficient. Efficiency is power (voltage in this case since current in = current out) out over power in.
Hammond sells chokes and chokes sometimes show up on auction sites like ebay.
Good luck and keep us posted.
My take on this
My take is that you probably need at least +-30V power supply. You also need the capability to drive not 600 Ohms, but 100 Ohms, but only to a a few volts -- say 4 which should give you a little bit of headroom.
If you are in class A mode you only need about 40mA of bias to do this. For your 600 Ohm version, you are under 10mA .... Not too much to ask ....
An alternative is to use JFET's all around and go with lower voltages which is also a most valid approach.
Yet another alternative is higher voltage with tubes. The 6H30P-DR is a pretty cool customer
If you go with a Pass-X stage you need less voltage since this is a bridged unit.
I would feel comfortable with any of these approaches but feel that progress on topology needs to be made first.
As for using a Variac -- that is fine, but you still need a regular transformer for isolation since the Variac does not isolate, nor will it give you a good center tap if that is important to you. If using a +- supply, you really need two separate windings so that you can use double rectification and avoid ground currents as you would with a center tap unit. Inexpensive and effective.
Petter
My take is that you probably need at least +-30V power supply. You also need the capability to drive not 600 Ohms, but 100 Ohms, but only to a a few volts -- say 4 which should give you a little bit of headroom.
If you are in class A mode you only need about 40mA of bias to do this. For your 600 Ohm version, you are under 10mA .... Not too much to ask ....
An alternative is to use JFET's all around and go with lower voltages which is also a most valid approach.
Yet another alternative is higher voltage with tubes. The 6H30P-DR is a pretty cool customer
If you go with a Pass-X stage you need less voltage since this is a bridged unit.
I would feel comfortable with any of these approaches but feel that progress on topology needs to be made first.
As for using a Variac -- that is fine, but you still need a regular transformer for isolation since the Variac does not isolate, nor will it give you a good center tap if that is important to you. If using a +- supply, you really need two separate windings so that you can use double rectification and avoid ground currents as you would with a center tap unit. Inexpensive and effective.
Petter
when i said it would make down-stream regulators more efficient, i meant because it keeps the supply pretty well regulated around .9*RMS, whereas a cap filter is 1.4*RMS with no load, but drops towards RMS with load... so with a choke regulator you can use a smaller voltage margin to the down-stream passive regulators, thus decreasing power dissipation in the regulators themselves. with a cap input filter, you would need to be sure there is enough margin at RMS voltage but then at light loading the regulators will need to drop that additional .4*RMS voltage. just less wasted power with a choke.
I am not happy about all this LC or PI filter things....
The reason is that such a filter will have a peaking around ~ 1/(2*pi*SQ(L*C)) (not the right formula but close to the frequency where the peaking is) if the Q too high ... To lower Q the series resistance of the inductor has to be rather high.
So if the inductors and caps are of wrong values this peaking is inside the audible area!!
Don't get me wrong i do use ferritbeads and inductors . The do work fine ... But you have to be carefull!
I would prefer a shunt regulator or some of the noise cancel psu circuit shown on this site.
SonnyA - So shure, Are you
The reason is that such a filter will have a peaking around ~ 1/(2*pi*SQ(L*C)) (not the right formula but close to the frequency where the peaking is) if the Q too high ... To lower Q the series resistance of the inductor has to be rather high.
So if the inductors and caps are of wrong values this peaking is inside the audible area!!
Don't get me wrong i do use ferritbeads and inductors . The do work fine ... But you have to be carefull!
I would prefer a shunt regulator or some of the noise cancel psu circuit shown on this site.
SonnyA - So shure, Are you
100 ohms??? this is starting to sound more like a headphone amp than a preamp!
i do not think we need this sort of drive capability. Class A at 5V RMS into 4kohm is plenty enough juice, let alone into 10kohm. most likely this thing will be operating with peak voltages no greater than 3V or so into loads of 10k or greater.
given that, i think all-JFETS is a distinct possibility. i am going to prototype the borbely "super buffer" circuit soon, which is the basis of his latest line amp design (the complementary dc-coupled one), the main difference being the super buffer uses JFET followers whereas the line amp uses MOSFETS (it was designed to drive headphones as well). but that is a discussion for the Active Circuitry thread!
sonny - do you think the resonant frequency of the filter needs to be below the audible range? i suppose you're right. most people say it just needs to be below the ripple frequency, but it's quite possible that the dynamic load of a circuit could induce resonance. if we have to tune the filter to, say, 15hz or so... that is a mighty big choke.
ugh... now i am starting to see why people use batteries, pain that they are...
cheers,
marc
i do not think we need this sort of drive capability. Class A at 5V RMS into 4kohm is plenty enough juice, let alone into 10kohm. most likely this thing will be operating with peak voltages no greater than 3V or so into loads of 10k or greater.
given that, i think all-JFETS is a distinct possibility. i am going to prototype the borbely "super buffer" circuit soon, which is the basis of his latest line amp design (the complementary dc-coupled one), the main difference being the super buffer uses JFET followers whereas the line amp uses MOSFETS (it was designed to drive headphones as well). but that is a discussion for the Active Circuitry thread!
sonny - do you think the resonant frequency of the filter needs to be below the audible range? i suppose you're right. most people say it just needs to be below the ripple frequency, but it's quite possible that the dynamic load of a circuit could induce resonance. if we have to tune the filter to, say, 15hz or so... that is a mighty big choke.
ugh... now i am starting to see why people use batteries, pain that they are...
cheers,
marc
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.