Hi,
I can, except Altium quit on me after installing Win 7 SP1 and my ancient copy of PSpice quit working when I went to Win 7 64 Bit. SO I do it when I get everything reinstalled.
Untill then it will have to be over a pint of linn dubh (known to the ungodly as Guinness) or two on the back of a serviette (don't worry, I get drunk after one beer, I just can never remember if it is the 13th or 14th)...
Meanwhile, there was a schematic for the Creek OBH-8SE posted in this-here venue, courtesy of Alexander Nikitkin aka xpro (who's Idea I appropriated) which can illustrate the principle, if using mostly transistors suffering from split personality (bipolar disorder).
Ciao T
I can, except Altium quit on me after installing Win 7 SP1 and my ancient copy of PSpice quit working when I went to Win 7 64 Bit. SO I do it when I get everything reinstalled.
Untill then it will have to be over a pint of linn dubh (known to the ungodly as Guinness) or two on the back of a serviette (don't worry, I get drunk after one beer, I just can never remember if it is the 13th or 14th)...
Meanwhile, there was a schematic for the Creek OBH-8SE posted in this-here venue, courtesy of Alexander Nikitkin aka xpro (who's Idea I appropriated) which can illustrate the principle, if using mostly transistors suffering from split personality (bipolar disorder).
Ciao T
Hi,
The SE has a Jfet cascoded by the lowest NPN Transistor instead, but the principle remains...
Very smart it is too. It is not a SRPP BTW, more a Mu-Follower, but with a pentode analog in the lower position, meaning gain without the load is sky-high.
Ciao T
The SE has a Jfet cascoded by the lowest NPN Transistor instead, but the principle remains...
Very smart it is too. It is not a SRPP BTW, more a Mu-Follower, but with a pentode analog in the lower position, meaning gain without the load is sky-high.
Ciao T
Hi,
Yes, a J something...
Link to SE Schematic is in this post:
http://www.diyaudio.com/forums/analogue-source/151527-how-convert-creek-obh-8-mc-cartridges.html#post1983686
Ciao T
Yes, a J something...
Link to SE Schematic is in this post:
http://www.diyaudio.com/forums/analogue-source/151527-how-convert-creek-obh-8-mc-cartridges.html#post1983686
Ciao T
The Myth of matching paralleled JFETs for noise
Take two 2SK170GR (6-12mA) and run the pair at 8mA total
1) Match two at 6mA Idss, noise is about .49nV
2) Match two at 12mA Idss, noise is again .49nV
3 Take one at 6mA and one at 12mA worst possible spread, noise is now .50nV
This can easily be derived from the basic JFET equations.
Take two 2SK170GR (6-12mA) and run the pair at 8mA total
1) Match two at 6mA Idss, noise is about .49nV
2) Match two at 12mA Idss, noise is again .49nV
3 Take one at 6mA and one at 12mA worst possible spread, noise is now .50nV
This can easily be derived from the basic JFET equations.
Current hogging why? 6mA Idss in parallel with 12mA Idss, at 8mA total for both they share 6mA and 2mA. gm1 = 29.6mS gm2 = 17mS net 46.6mS. If you had matched 6mA fets the gm would be 48.3mS pretty trivial and by the time you are picking quads even less variance.
BTW that is the worst case 46.6 to 48.3 mS. The noise difference is less by the square root.
If you use graphical techniques to show what self biasing does, then the problems are increased, especially for a complementary push-pull design. I must agree that EXTREME matching does little for noise. However, running one junction at 100 degrees C and the other at 30 degrees C doesn't seem right either.
Dear Scott,
Based on the datasheets (and personal experience) your noise figures are too optimistic. The datasheet shows the noise reduction with increased Idss flattening out at around 0.82nV|/Hz. Two in parallel hence theoretically have 0.58nV|/Hz, or 3dB noisier than a 10 Ohm resistor, or equivalent to a 20 Ohm resistor.
In reality not all 2SK170 meet their "typical" noise spec, so you may be even worse off than this simple calculation depending on precise parts.
However, if "clone" the Vendetta with 4 pairs of 2SK170BL/2SJ74BL we have around 9dB improvement, so we get 0.29nV|/Hz theoretically. However if include some 15 Ohm source resistors per transistor to pad the Idss down to 6mA (so 24mA total current for the frontend transistors) and improve matching we add an extra 2 Ohm (0.18nV|/Hz) worth of noise, so we are back to 0.47nV|/Hz...
Of course we could try buying GR Grade Parts by the 1,000 (funny, I tried, only could get BL ones) and match them so we can run entierly without source resistor or simply accept the around 9mA Idss for the BL Devices...
To be honest, in practice, excluding the use with Ortofon MC2000 and Kondo Io I doubt that the noise differences between 0.4nV|/Hz, 0.5nV|/Hz and 0.7nV|/Hz Ein are material, unless you are a "technical reviewer" and you have an AP2 (well, so do I) and you are not afraid to use it...
I have build plenty of Phonos with single 2SK147, INF147 and even 2SK170BL WITH a fairly high source degeneration resistors (which adds noise - in my case often as much or more than the FET) and was well below the LP noise using 0.5mV-ish MC Pickups...
So while low noise is important in MC Phono, I think it can be overdone. I would aim for < 100 equivalent noise impedance (< 1.2nV|/Hz) and not loose sleep over noise below that, unless I intend to use 0.1mV and below MC Pickups.
In practice there are other factors that will make or break the sound quality of a Phono Stage to a by far greater degree than the difference between 0.4nV|/Hz Ein and 1.2nV|/Hz Ein.
Once we "get real" again instead of chasing spec's we can explore wider ranges of topologies and just generally have more fun.
Ciao T
Based on the datasheets (and personal experience) your noise figures are too optimistic. The datasheet shows the noise reduction with increased Idss flattening out at around 0.82nV|/Hz. Two in parallel hence theoretically have 0.58nV|/Hz, or 3dB noisier than a 10 Ohm resistor, or equivalent to a 20 Ohm resistor.
In reality not all 2SK170 meet their "typical" noise spec, so you may be even worse off than this simple calculation depending on precise parts.
However, if "clone" the Vendetta with 4 pairs of 2SK170BL/2SJ74BL we have around 9dB improvement, so we get 0.29nV|/Hz theoretically. However if include some 15 Ohm source resistors per transistor to pad the Idss down to 6mA (so 24mA total current for the frontend transistors) and improve matching we add an extra 2 Ohm (0.18nV|/Hz) worth of noise, so we are back to 0.47nV|/Hz...
Of course we could try buying GR Grade Parts by the 1,000 (funny, I tried, only could get BL ones) and match them so we can run entierly without source resistor or simply accept the around 9mA Idss for the BL Devices...
To be honest, in practice, excluding the use with Ortofon MC2000 and Kondo Io I doubt that the noise differences between 0.4nV|/Hz, 0.5nV|/Hz and 0.7nV|/Hz Ein are material, unless you are a "technical reviewer" and you have an AP2 (well, so do I) and you are not afraid to use it...
I have build plenty of Phonos with single 2SK147, INF147 and even 2SK170BL WITH a fairly high source degeneration resistors (which adds noise - in my case often as much or more than the FET) and was well below the LP noise using 0.5mV-ish MC Pickups...
So while low noise is important in MC Phono, I think it can be overdone. I would aim for < 100 equivalent noise impedance (< 1.2nV|/Hz) and not loose sleep over noise below that, unless I intend to use 0.1mV and below MC Pickups.
In practice there are other factors that will make or break the sound quality of a Phono Stage to a by far greater degree than the difference between 0.4nV|/Hz Ein and 1.2nV|/Hz Ein.
Once we "get real" again instead of chasing spec's we can explore wider ranges of topologies and just generally have more fun.
Ciao T
Interesting. I wondered about this. Had I taken the time to carefully review the equations/datasheets, I would still wonder if this is the case in practice.
.
Hi T,
You are so very right about noise. Some people are too obsessed with noise - sometimes influencing their design to the detriment of overall sound quality. 1 nV/rt Hz is fine. It will in theory provide an A weighted, RIAA-equalized S/N of about 78 dB with respect to a nominal MC cartridge output of 400uV.
Cheers,
Bob
Hi,
Incidentally, if I use the Vendetta Schematic as posted in this thread and my datasheets for K147/J72 I get 0.53nV|/Hz Ein.
Of course, we all know John used specially selected Fets, so he may have gotten slightly better results than that, however I doubt it would 3 - 6dB better.
BTW, while having a quantech noise analyser is nice, using the Wenzel Associates Low Noise Preamp and a normal soundcard can make a great noise analyser. Just make sure, when you build it using 2SK369 you test all your stock with the first build and the put the quietest ones into the low noise pre, so you have the best possible resolution...
Ciao T
Incidentally, if I use the Vendetta Schematic as posted in this thread and my datasheets for K147/J72 I get 0.53nV|/Hz Ein.
Of course, we all know John used specially selected Fets, so he may have gotten slightly better results than that, however I doubt it would 3 - 6dB better.
BTW, while having a quantech noise analyser is nice, using the Wenzel Associates Low Noise Preamp and a normal soundcard can make a great noise analyser. Just make sure, when you build it using 2SK369 you test all your stock with the first build and the put the quietest ones into the low noise pre, so you have the best possible resolution...
Ciao T
The 10 ohm spec. established by the Levinson JC-1, back in 1973 was established to properly amplify the Ortofon 50uV cartridge. When, soon after its introduction, Supex, FR came out with their 100uV output, and even Denon started selling their MC phono cartridges in the USA, the 10 equivalent noise spec still held well. When the Vendetta came out 8 years later, phono cartridges were gradually increasing in output, but a few Ortofons were still low in output, and some phono cartridges are still this low in output, today.
For most here, it does not matter, but some audiophiles want this range of low noise reproduction. In 1981, when I first made the prototype of the Vendetta, I only achieved about 20 ohms of noise, so I went on to parallel devices and made other changes, in order to reach 10 ohms of equivalent noise. This was made fairly easy with 2SK146V's and 2SJ72V's that were affordable and already matched in like pairs. The only matching requirement was the Idss of the N channels and the P channels. This was fairly easy, AND necessary.
Over the years, it has been harder and more expensive to make Vendetta like designs, but nobody wants to make a 'statement' product that is far off from what the Vendetta achieved 25 years ago, so we struggle on, and it is not easy, even with tens of thousands of devices, available to us.
We can sometimes get noisy devices, orphan devices (way out of range), and sometimes just no specific device in a given range is available.
This is why I tend to discourage people from taking on a Vendetta like design, lightly.
Can 1 nanovolt be OK? OF COURSE! For most MC phono cartridges, and the Parasound JC-3 is in this range, in fact it is even slightly noisier.
So go for it.
For most here, it does not matter, but some audiophiles want this range of low noise reproduction. In 1981, when I first made the prototype of the Vendetta, I only achieved about 20 ohms of noise, so I went on to parallel devices and made other changes, in order to reach 10 ohms of equivalent noise. This was made fairly easy with 2SK146V's and 2SJ72V's that were affordable and already matched in like pairs. The only matching requirement was the Idss of the N channels and the P channels. This was fairly easy, AND necessary.
Over the years, it has been harder and more expensive to make Vendetta like designs, but nobody wants to make a 'statement' product that is far off from what the Vendetta achieved 25 years ago, so we struggle on, and it is not easy, even with tens of thousands of devices, available to us.
We can sometimes get noisy devices, orphan devices (way out of range), and sometimes just no specific device in a given range is available.
This is why I tend to discourage people from taking on a Vendetta like design, lightly.
Can 1 nanovolt be OK? OF COURSE! For most MC phono cartridges, and the Parasound JC-3 is in this range, in fact it is even slightly noisier.
So go for it.
Your missing my point.
I offered a circuit that makes its own minus voltage so you don't need source resistors to bias at less than Idss, a cute trick (I saw it years ago at least 1976). I think Salas has carried on a long thread on another way to do it. If you parallel devices and set the drain current via another feedback loop as here matchiing buys you nothing and you don't need any source resistors.
I don't know why things get so sidetracked, I could care less about the last iota of noise.
EDIT - I've corresponded with Charles, ironicly nowhere does he say to match the 2 FETs in parallel on his amp (it does not matter).
Last edited:
Scott,
Sorry for missing that, maybe because I work mostly with tubes and this "trick" is quite ancient in the tube world (mains powered 1920's Radios and Theatrical Amp's often got their bias that way) and very common?
FWIW, several of my products and some DIY designs all across the board use this "trick". So sorry, for not picking up on your intention with this circuit, I just did not think it all that great an idea overall...
In this specifc circuit your trick imposes an AC coupling cap on the Fet Input (there is yet another trick to avoid that of course which was also common in 1920's Tube circuits, but it only shifts the cap to somewhere else).
At 1KHz your 0.33uF Cap has an impedance of 480 Ohm.
While in itself noiseless as a reactive element (just like a choke and the OLD MM Cartridges 0.5h INDUCTANCE IS MOSTLY NOISELESS), the impedance of this capacitor will be in series with the cartridge and hence will appear as impedance across the input of your Preamp, where you have large value resistor (at least 56KOhm to get around 8Hz -3dB point) to apply your bias voltage.
This resistors noise will now be shorted only by the Cartridge and the 0.33uF Cap, so at 1KHz your effective noise will equal sufficiently closely 482 Ohm + Cartridge Impedance as not to make a difference.
You need to increase the coupling cap 100 fold to 33uF in order to get around 5 Ohm Impedance at 1KHz and your noise at lower frequencies will still be high. Hence one frequently sees very large value coupling caps on MC Stage inputs, I should think a few 1,000uF would be needed before this cap does not severely degrades the SNR.
Hence my recommendation to direct couple the cartridge to the gate, to avoid this noise source entirely.
If we use a 2SK369V at Idss we will draw around 19mA (design center), so 9V batteries are out of the picture (they are 170mAH), but we will have a genuine 0.7nV|/Hz, 65mA/V transconductance and with a 150 Ohm Drain load a gain of around 20dB.
Supply should be around 11V, so for the "batteries is a must" faction, use a nice 12V rechargable.
This will make a cute enough headamp, add a few uF output coupling cap (to reduce the hum pickup and noise on the input of your MM Phono) and done.
Me, I'd probably use a 317 to make a higher voltage and then a few 150R/3,300uF Elna Silmic RC filter cells (around 3V per cell) to get the 317 noise down.
With 150R/3300uF the filter corner is 0.3Hz, so with 3 cells we get theoretically 120dB attenuation of the 317 Noise at 30Hz, practically less, but it should suffice, a well set up 317 is only around 2-3uV RMS.
The voltage from the 317 should be around 20V, we loose 9V in the filter-chain, I would probably trim the 317 Voltage for 8V on the 2SK369 Drain.
Lower noise is possible, but to get at least 6dB improvement on the single, source resistorless 2SK369V (or 2SK147V if you have them) is a major PITA and in my books less than 6dB improvement to a "pretty much good enough" circuit is not worth the hassle...
Yup, makes two of us, I just want low enough noise to listen to music...
Ciao T
Sorry for missing that, maybe because I work mostly with tubes and this "trick" is quite ancient in the tube world (mains powered 1920's Radios and Theatrical Amp's often got their bias that way) and very common?
FWIW, several of my products and some DIY designs all across the board use this "trick". So sorry, for not picking up on your intention with this circuit, I just did not think it all that great an idea overall...
In this specifc circuit your trick imposes an AC coupling cap on the Fet Input (there is yet another trick to avoid that of course which was also common in 1920's Tube circuits, but it only shifts the cap to somewhere else).
At 1KHz your 0.33uF Cap has an impedance of 480 Ohm.
While in itself noiseless as a reactive element (just like a choke and the OLD MM Cartridges 0.5h INDUCTANCE IS MOSTLY NOISELESS), the impedance of this capacitor will be in series with the cartridge and hence will appear as impedance across the input of your Preamp, where you have large value resistor (at least 56KOhm to get around 8Hz -3dB point) to apply your bias voltage.
This resistors noise will now be shorted only by the Cartridge and the 0.33uF Cap, so at 1KHz your effective noise will equal sufficiently closely 482 Ohm + Cartridge Impedance as not to make a difference.
You need to increase the coupling cap 100 fold to 33uF in order to get around 5 Ohm Impedance at 1KHz and your noise at lower frequencies will still be high. Hence one frequently sees very large value coupling caps on MC Stage inputs, I should think a few 1,000uF would be needed before this cap does not severely degrades the SNR.
Hence my recommendation to direct couple the cartridge to the gate, to avoid this noise source entirely.
If we use a 2SK369V at Idss we will draw around 19mA (design center), so 9V batteries are out of the picture (they are 170mAH), but we will have a genuine 0.7nV|/Hz, 65mA/V transconductance and with a 150 Ohm Drain load a gain of around 20dB.
Supply should be around 11V, so for the "batteries is a must" faction, use a nice 12V rechargable.
This will make a cute enough headamp, add a few uF output coupling cap (to reduce the hum pickup and noise on the input of your MM Phono) and done.
Me, I'd probably use a 317 to make a higher voltage and then a few 150R/3,300uF Elna Silmic RC filter cells (around 3V per cell) to get the 317 noise down.
With 150R/3300uF the filter corner is 0.3Hz, so with 3 cells we get theoretically 120dB attenuation of the 317 Noise at 30Hz, practically less, but it should suffice, a well set up 317 is only around 2-3uV RMS.
The voltage from the 317 should be around 20V, we loose 9V in the filter-chain, I would probably trim the 317 Voltage for 8V on the 2SK369 Drain.
Lower noise is possible, but to get at least 6dB improvement on the single, source resistorless 2SK369V (or 2SK147V if you have them) is a major PITA and in my books less than 6dB improvement to a "pretty much good enough" circuit is not worth the hassle...
Yup, makes two of us, I just want low enough noise to listen to music...
Ciao T
You do not mean selected ones ?
Of the J73/K146 or J75/K240 i have, a fraction has close Idss numbers, goes for Green, Bleu and Violet.
For 0.01mA accuracy, something as one in a hundred.
The monolythic duallies seemingly have a more favorable spread.
(typical Japanese rationale, J72/K147 and J73/K146)
On stout, i'd probably have twice the circumference of Mr Loesch within 6 months.
Last edited:
Its not only about noise IMHO. Its the Zo of a PSU and its shape which is very important too. That is why I opt for super shunt regulators. The input CCS that lends the 'super' term (by Stax in the 70's) its not the end all firewall at all for downstream gremlins from the rectification and general grounding, for those where in frequency and how they circulate in the loop areas it would not do more than a resistor in practice, but it lends the benefit of steady current draw from the feeding loop. That is the main benefit of CCSing the regs in my opinion. That opens the road for less concerned use of passive filtering elements in the main PSU also. Now for Vref considerations or a real regulator, not a filter, and noise:
IngemaR had simulated 25nVrtHz for a V1+EF LED string referenced 18V shunt reg with Toshiba Jfets and BJTs when the ref is filtered with 220uF or more. That will be equivalent to a 39K resistor and 3.55uV across 20KHz BW output noise. With that kind of psu noise the flat extended mOhm range Zo especially when the shunt element is buffer driven (+EF) becomes attractive IMO. The passives, Teddy Cmult, etc. pale for Zo spec and shape, which becomes increasingly problematic for interactions of low psrr ccts the more stages we want to feed from as simple a psu assortment as possible. The shunt route takes us a long way towards low impedance rails with just one central psu arrangement even. That is build economy. Remote sensing keeps the psu away enough safely for low Zo on another board so not to heat up the input devices and make them noisier by psu sinks proximity in a low MC phono. The subjective dynamic rendering and resolution gains had been reported back in a steady anecdotal vein by a multitude of happy campers if it matters at all. So its a matter of architectural approach mainly, many ways to skin the cat, but that is my preferred approach more analytically so to speak since I have seen it shortly discussed noise wise for passives VS actives in a previous post.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.