anatech said:
Hi Chris,
So my guess is the J-fets account for the lions share of the magic. I've seen the diamond arrangement mentioned but I'm not really sure what that means, but my real curiosity is the choice of J-fets (vp and IDSS range) for the front end of the MC amp and the power supply implementation. Interesting questions in my mind on both of these. Do you use a MC or MM cartridge, if I'm not getting too personal with that question?
I'll search for the Marantz 3650 schematc.
Regards, Mike.
Mike,
Given that most of us will never go to the trouble and expense inherent in a milled-out aluminum block, I am assuming that we will be dealing with chassis assemblies that aren't as good at blocking RF. With that in mind, it's not inconceivable that there might be parts of some circuits that could use a little localized help.
The capacitance objection is easy enough to deal with--drive the copper shield with a guard circuit. Yes, it could get messy if you tried to cover every component on the PCB with a shield, but that would be a dumb thing to do anyway. At that point you'd be better off adding another layer of shielding to the entire circuit.
I don't see this shield as necessarily being in physical contact with the component it's shielding unless you intend for it to act as a heatsink in addition to blocking RF. A couple of millimeters of air between the component and shield would go a long ways towards reducing potential capacitance problems.
I only see this as a possibility for one, two, maybe three components per circuit. If you started trying to cover every component, the PCB real estate would be out of control in no time. Trace length would increase considerably and you'd soon create more problems than you were solving.
Grey
Given that most of us will never go to the trouble and expense inherent in a milled-out aluminum block, I am assuming that we will be dealing with chassis assemblies that aren't as good at blocking RF. With that in mind, it's not inconceivable that there might be parts of some circuits that could use a little localized help.
The capacitance objection is easy enough to deal with--drive the copper shield with a guard circuit. Yes, it could get messy if you tried to cover every component on the PCB with a shield, but that would be a dumb thing to do anyway. At that point you'd be better off adding another layer of shielding to the entire circuit.
I don't see this shield as necessarily being in physical contact with the component it's shielding unless you intend for it to act as a heatsink in addition to blocking RF. A couple of millimeters of air between the component and shield would go a long ways towards reducing potential capacitance problems.
I only see this as a possibility for one, two, maybe three components per circuit. If you started trying to cover every component, the PCB real estate would be out of control in no time. Trace length would increase considerably and you'd soon create more problems than you were solving.
Grey
Hi Mike,
The diamond buffer design as I've played with has no distinct sound it seems. I think it was a very good choice.
J Fet input pairs have always equated to a very "open", non-harsh sound as long as they are built properly. They have in this preamp. I think the high sound quality is due to very good layout and circuit choices. The power supply is in the rear with shields between it and teh rest of the world.
-Chris
The diamond buffer design as I've played with has no distinct sound it seems. I think it was a very good choice.
J Fet input pairs have always equated to a very "open", non-harsh sound as long as they are built properly. They have in this preamp. I think the high sound quality is due to very good layout and circuit choices. The power supply is in the rear with shields between it and teh rest of the world.
-Chris
Would somthing likethis not be a good compromise between performance and price say for a MC-circuit used *inside* a good looking cabinet?
Rüdiger
Here you are with an english link
Rüdiger
Here you are with an english link
mike:
looks like 2SK136 Q (??) for the jfets in question (for the P400 phono board)
mlloyd1
edit:
it is Q grade by panasonic, datasheet here:
http://www.ortodoxism.ro/datasheets/panasonic/2SK136.pdf
go to page 3
looks like 2SK136 Q (??) for the jfets in question (for the P400 phono board)
mlloyd1
edit:
it is Q grade by panasonic, datasheet here:
http://www.ortodoxism.ro/datasheets/panasonic/2SK136.pdf
go to page 3
MikeBettinger said:
If you want to use diecast Alu boxes, Conrad has cheaper ones than Reichelt.
I myself get 1mm copper sheets from Baumarkt and make my own. Takes half an hour with a pair of tin snips and a big 100W soldering iron. If worried about oxidation, just spray with clear lacquer after cleaning.
Patrick
I myself get 1mm copper sheets from Baumarkt and make my own. Takes half an hour with a pair of tin snips and a big 100W soldering iron. If worried about oxidation, just spray with clear lacquer after cleaning.
Patrick
mlloyd1 said:
Thanks. From a simple inspection of the images on the link it looks like the MC stage is based on a single fet input and the RIAA has a single ended fet diff amp. I haven't taken it much further as I just realized there were backside images of the board a few hours ago.
Regards, Mike.
Better yet, Thanks a lot for posting the data sheets.
An interesting observation. On only the 2SK136 datasheet, on the graph for Id-Vds, if you look at the various Vgs voltages listed for the curves you will see all of the usual negative voltages, but it also shows a curve for positive .1 volts. I'm assuming that at these voltages all fets have a response, this is just the first time I've seen it noted on a specsheet.
With the low output signal of a MC cartridge the Vp and the response in this range might be significant. Actually once Vd get above a few volts the response between +/- .1v looks pretty good.
Thanks again. Mike.
An interesting observation. On only the 2SK136 datasheet, on the graph for Id-Vds, if you look at the various Vgs voltages listed for the curves you will see all of the usual negative voltages, but it also shows a curve for positive .1 volts. I'm assuming that at these voltages all fets have a response, this is just the first time I've seen it noted on a specsheet.
With the low output signal of a MC cartridge the Vp and the response in this range might be significant. Actually once Vd get above a few volts the response between +/- .1v looks pretty good.
Thanks again. Mike.
MikeBettinger said:
I thought maybe I farted in church again, so I got out a cartridge and double checked, yes it does put out a true ac signal. If I was to connect it directly to the gate of a fet it would see and need to respond linearly to a positive and negative input. Obviously there is something here everyone takes for granted.
Must be my inexperience with such things.
I reverse engineered the MC stage of the Marantz. Basically there isn't an audiophile approved thing in this part of the phono stage. And the cartridge is single endedly connected to the gate. Interesting (to me).
Mike.
I confess that I didn't follow your post. Did you think that phono cartridges put out DC? If so, via what mechanism?
It's not a good idea to push a JFET's Gate too positive (assuming N-ch). For one thing, they get nonlinear, fast. For another, it'll start drawing Gate current, which can complicate your life very quickly, indeed.
A tenth or a volt or two worth of bias will ensure that things run smoothly if you're running a moving coil. If you're using a moving magnet, you'll need more.
Grey
It's not a good idea to push a JFET's Gate too positive (assuming N-ch). For one thing, they get nonlinear, fast. For another, it'll start drawing Gate current, which can complicate your life very quickly, indeed.
A tenth or a volt or two worth of bias will ensure that things run smoothly if you're running a moving coil. If you're using a moving magnet, you'll need more.
Grey
Stand-up resistors
Personally, I wouldn't choose this construction technique for some mass produced product, or anything subject to the rigors of military-level field abuse. But for my own home audio projects, I'm not the least bit concerned about mechanical damage or leaning resistors. I'm a very diligent, careful builder - my resistors go in straight, and my equipment is gently used and well cared for. I'd rather have the board space savings and easy-access probe points. YMMV, so that's a personal choice.
Electrically speaking, for a large majority of audio circuits, I really don't see this construction technique making a whit of difference. Let me try and put the alleged detrimental effects into perspective with some hard numbers:
Let's start with inductance: Going from 4nH to 8nH is only going to make a very small change in the -3dB point, and for that kind of inductance, it's only going to barely start affecting signal bandwidth in the 100MHz range at node impedances of 10Ohms or less. Even at 1Ohm, the -3dB point is still out beyond 10MHz. For a high-speed opamp circuit, okay be careful. But for most discrete audio circuits, forget it, there's no real difference.
Likewise, a teensy bit of stray cap will only affect the highest-impedance nodes in a circuit... and in this case our dielectric is air, which is as good as it gets (next to vacuum). If we assume a very pessimistic stray cap of 0.5pF, you'll see effects starting at 300kHz for a node impedance of 1MOhm, which is about the upper limit of impedance in most tube audio circuits. 300kHz still gives us more than a decade of guard-band up from the highest audio frequencies. And, a typical solid-state circuit rarely has node impedances over 100kOhms, so we're really looking at >3MHz before we start seeing any significant bandwidth effect from this stray cap.
Antenna effect? Sorry, not below 1.5GHz. Antenna effects start to come into play when the physical geometry of a circuit approaches or exceeds 1/20th of a wavelength. For a 1cm resistor height, we're talking about a wavelength of 20cm, which is 1.5GHz. Audio frequencies are simply not radiating or receiving from such a structure, in the RF sense. In the audio band, and well above, we're talking strictly inductive and capacitive coupling.
So, I would submit that none of the above is going to make a difference in reality. But I know some people will still have lingering doubts, so let's be super- paranoid- pessimistic. Let's talk about coupling between circuits instead of bandwidth. Maybe we want to see less than -100dB crosstalk in the audio band between two components or circuit elements. Using a very sensitive capacitance meter, I just measured the stray capacitance between two 1cm lengths of wire, standing approximately 1cm apart, with one end against a ground plane, closely simulating a pair of vertical resistors above a circuit board's ground plane. I measure 19fF of stray cap, and the fixture residual was 6fF without the wires in place. So that's 13fF of stray cap between these two "standing resistors". It's an approximation, but good enough to work with for a ballpark estimate. I'll round up to 20fF for sake of pessimism. Next, we'll assume a typical circuit node impedance of 10kOhms on the 'victim' resistor's lead. The signal crosstalk picked up by the victim resistor will be -6dB in magnitude when the reactive impedance of the stray cap is equal to 10k... that's at a frequency of about 800MHz in this case. Falling at 20dB/decade, by the time we reach 20kHz, the coupling is at -98dB... pretty close to our target of -100dB. Further into the audio band, the coupling will be even less. Of course, if we know there are signals on adjacent components that we don't want coupled (assuming we couldn't work the placement better in the first place), it's easy enough to change the layout a bit, or add a little shielding in between them. I use small pieces of copper-clad FR4, brass, or copper sheet, soldered down to the ground plane, and this very effectively kills capacitive crosstalk between nearby circuit elements.
Alright, now what about these alleged thermal gradient effects? Well, I've never seen that make a difference either. Check out the UGS Adventures thread. There you'll see my C-UGS prototype, which is an all discrete preamp, constructed almost entirely with stand-up resistors (save for about a half-dozen resistors in the whole circuit). DC offset is typically under 100uV (yes, you read that right!) after trim, and remains stable - no servos, and no coupling caps. Over a 6-hour period connected to a 6.5digit bench meter, the min-max 1/f excursions were only +/-430uV from the average offset of 30uV. This is the kind of performance you can only get from some of the best IC opamps, and only when used in a properly designed circuit.
I didn't talk about magnetic coupling. These effects are a function of loop area and magnetic flux, and are a little more complicated to calculate than is appropriate for this already-looong post. But, the effects are similarly negligible, and are relatively easy to control through simple choices in component placement, shielding, and circuit geometry. It's a more complex topic, but if you want to learn, I highly recommend Henry Ott's "Noise Reduction Techniques in Electronic Systems". It's an excellent read (and essential, IMO, for those who are serious about audio circuit design).
All of the extremely marginal effects described above can easily be defeated or controlled through simple tricks and intelligent design choices - what's the best layout, where to apply different construction techniques, and where a little shielding is appropriate, etc. With a little thought, you can build a circuit, using your construction method of choice, which isn't susceptible to these problems. Even without a lot of thought, I think it's pretty safe to go ahead and use stand-up resistors or caps where needed, and sleep soundly, knowing the effect is nil in all but a few cases. All this paranoia about vertical or horizontal resistors is unwarranted, and most DIY builders would do well to ignore it and instead focus their attention on much more important matters such as basic circuit design and optimization, overall layout, and choice of components. Horizontal resistors cannot compensate for a poorly-conceived design.
Having said all that, it is a detailed understanding of the above effects and tradeoffs, and a meticulous attention to detail which really sets apart the cream-of-the-crop designs from the rest of the pack. It's not black magic, but it is art, and does take a lot of study and experience to get a handle on the many design issues and their relative importance, and which way to go with any given compromise. I've been designing audio circuits for over 15 years, and regularly work with 14-layer boards carrying 5GHz+ digital signals in my day job, and I'm still learning and getting better at audio design all the time... it's one of the things I love about this hobby, there's always more to learn and experience, and the balance of design choices is a complex and intricate intellectual challenge.
Personally, I wouldn't choose this construction technique for some mass produced product, or anything subject to the rigors of military-level field abuse. But for my own home audio projects, I'm not the least bit concerned about mechanical damage or leaning resistors. I'm a very diligent, careful builder - my resistors go in straight, and my equipment is gently used and well cared for. I'd rather have the board space savings and easy-access probe points. YMMV, so that's a personal choice.
Electrically speaking, for a large majority of audio circuits, I really don't see this construction technique making a whit of difference. Let me try and put the alleged detrimental effects into perspective with some hard numbers:
Let's start with inductance: Going from 4nH to 8nH is only going to make a very small change in the -3dB point, and for that kind of inductance, it's only going to barely start affecting signal bandwidth in the 100MHz range at node impedances of 10Ohms or less. Even at 1Ohm, the -3dB point is still out beyond 10MHz. For a high-speed opamp circuit, okay be careful. But for most discrete audio circuits, forget it, there's no real difference.
Likewise, a teensy bit of stray cap will only affect the highest-impedance nodes in a circuit... and in this case our dielectric is air, which is as good as it gets (next to vacuum). If we assume a very pessimistic stray cap of 0.5pF, you'll see effects starting at 300kHz for a node impedance of 1MOhm, which is about the upper limit of impedance in most tube audio circuits. 300kHz still gives us more than a decade of guard-band up from the highest audio frequencies. And, a typical solid-state circuit rarely has node impedances over 100kOhms, so we're really looking at >3MHz before we start seeing any significant bandwidth effect from this stray cap.
Antenna effect? Sorry, not below 1.5GHz. Antenna effects start to come into play when the physical geometry of a circuit approaches or exceeds 1/20th of a wavelength. For a 1cm resistor height, we're talking about a wavelength of 20cm, which is 1.5GHz. Audio frequencies are simply not radiating or receiving from such a structure, in the RF sense. In the audio band, and well above, we're talking strictly inductive and capacitive coupling.
So, I would submit that none of the above is going to make a difference in reality. But I know some people will still have lingering doubts, so let's be super- paranoid- pessimistic. Let's talk about coupling between circuits instead of bandwidth. Maybe we want to see less than -100dB crosstalk in the audio band between two components or circuit elements. Using a very sensitive capacitance meter, I just measured the stray capacitance between two 1cm lengths of wire, standing approximately 1cm apart, with one end against a ground plane, closely simulating a pair of vertical resistors above a circuit board's ground plane. I measure 19fF of stray cap, and the fixture residual was 6fF without the wires in place. So that's 13fF of stray cap between these two "standing resistors". It's an approximation, but good enough to work with for a ballpark estimate. I'll round up to 20fF for sake of pessimism. Next, we'll assume a typical circuit node impedance of 10kOhms on the 'victim' resistor's lead. The signal crosstalk picked up by the victim resistor will be -6dB in magnitude when the reactive impedance of the stray cap is equal to 10k... that's at a frequency of about 800MHz in this case. Falling at 20dB/decade, by the time we reach 20kHz, the coupling is at -98dB... pretty close to our target of -100dB. Further into the audio band, the coupling will be even less. Of course, if we know there are signals on adjacent components that we don't want coupled (assuming we couldn't work the placement better in the first place), it's easy enough to change the layout a bit, or add a little shielding in between them. I use small pieces of copper-clad FR4, brass, or copper sheet, soldered down to the ground plane, and this very effectively kills capacitive crosstalk between nearby circuit elements.
Alright, now what about these alleged thermal gradient effects? Well, I've never seen that make a difference either. Check out the UGS Adventures thread. There you'll see my C-UGS prototype, which is an all discrete preamp, constructed almost entirely with stand-up resistors (save for about a half-dozen resistors in the whole circuit). DC offset is typically under 100uV (yes, you read that right!) after trim, and remains stable - no servos, and no coupling caps. Over a 6-hour period connected to a 6.5digit bench meter, the min-max 1/f excursions were only +/-430uV from the average offset of 30uV. This is the kind of performance you can only get from some of the best IC opamps, and only when used in a properly designed circuit.
I didn't talk about magnetic coupling. These effects are a function of loop area and magnetic flux, and are a little more complicated to calculate than is appropriate for this already-looong post. But, the effects are similarly negligible, and are relatively easy to control through simple choices in component placement, shielding, and circuit geometry. It's a more complex topic, but if you want to learn, I highly recommend Henry Ott's "Noise Reduction Techniques in Electronic Systems". It's an excellent read (and essential, IMO, for those who are serious about audio circuit design).
All of the extremely marginal effects described above can easily be defeated or controlled through simple tricks and intelligent design choices - what's the best layout, where to apply different construction techniques, and where a little shielding is appropriate, etc. With a little thought, you can build a circuit, using your construction method of choice, which isn't susceptible to these problems. Even without a lot of thought, I think it's pretty safe to go ahead and use stand-up resistors or caps where needed, and sleep soundly, knowing the effect is nil in all but a few cases. All this paranoia about vertical or horizontal resistors is unwarranted, and most DIY builders would do well to ignore it and instead focus their attention on much more important matters such as basic circuit design and optimization, overall layout, and choice of components. Horizontal resistors cannot compensate for a poorly-conceived design.
Having said all that, it is a detailed understanding of the above effects and tradeoffs, and a meticulous attention to detail which really sets apart the cream-of-the-crop designs from the rest of the pack. It's not black magic, but it is art, and does take a lot of study and experience to get a handle on the many design issues and their relative importance, and which way to go with any given compromise. I've been designing audio circuits for over 15 years, and regularly work with 14-layer boards carrying 5GHz+ digital signals in my day job, and I'm still learning and getting better at audio design all the time... it's one of the things I love about this hobby, there's always more to learn and experience, and the balance of design choices is a complex and intricate intellectual challenge.
Actually, j-fets will run reasonably OK at OV bias, and you can forward bias them slightly without any real problem. This is because today's low noise fets have so much Gm that even .1V forward bias can a lot of current, in proportion to what is necessary. Another factor is that the junction in j-fets seems to be at a fairly hard junction to get current flowing, compared to some devices. This means that .1-.3V forward bias, in most situations will not change things much. This is NOT true with very high Z microphone inputs, etc. However, most sources are below 100K ohm.
GRollins said:
No I don't think a cartridge can output DC.
All of your comments about the nonlinearity of the gate is what my comments were based on. Here's a datasheet showing, in my interpretation, a device's performance into this positive range.
With a MC cartridge as an input (outputting a bipolar signal) I have wondered about the nonlinearity a normal fet might exhibit as the cartridges output swings both negative and positive.
I'm not sure I understand your comments on bias ensuring that things run smoothly.
I'm just attempting to fill in the blanks.
Mike.
HifiZen is on track due to his experience and education, I believe. It is HARD to do great analog. The tradeoffs are many.
When it comes to j-fet biasing, usually we like to operate BELOW Idss, which is the natural current that the fet will flow with 0 bias. Biasing can be inconvenient and noisy, and sometimes it is overused to be sure that the jfet NEVER exceeds Idss. This was OK about 40 years ago, but today, we allow a slight exceeding of Idss under worst case conditions, because it really doesn't do too much damage. This means that we can operate devices closer to Idss, at idle, and they work better under normal conditions.
When it comes to j-fet biasing, usually we like to operate BELOW Idss, which is the natural current that the fet will flow with 0 bias. Biasing can be inconvenient and noisy, and sometimes it is overused to be sure that the jfet NEVER exceeds Idss. This was OK about 40 years ago, but today, we allow a slight exceeding of Idss under worst case conditions, because it really doesn't do too much damage. This means that we can operate devices closer to Idss, at idle, and they work better under normal conditions.
- Status
- Not open for further replies.