LTSpice works under WINE on Linux with no issues (well, minor ones, but well enough)
DIY Jfet phono buffer.
- Thread starter a.palfreyman
- Start date
PigletsDad
My intelligence test came back negative.
We can break the PSRR into two parts - symmetric and anti-symmetric.
In the symmetric case, both rails drop the same amount, and by symmetry, for well matched devices, the output doesn't change.
In the anti-symmetric case, one drops and the other rises, and you are now exploring how flat the line of drain current at constant VGS is, versus the device transconductance. Because J112 is specified as a switching device, the datasheet is no help. Squinting at the 2SK170 datasheet, on the VGS = 0.3V line round VDS = 8V, makes me think 1V change of VDS will cause maybe 0.2mA of change in 2mA, and you would have something like 20mS of transconductance. So I think that would give you something like 10mV of output shift - so maybe 100:1 or 40dB.
This is very, very rough, but if we run with it, for arguments sake, if you want the total noise to be less than say 3uV, you need the rails to be cleaner than 300uV. The NE5534 solution is way overkill but would certainly do that. Something simpler might do.
A proper simulation would give a better answer, rather than eyeballing graphs for a completely different device!
If you are on linux, there are some good native options, look at gEda which I think comes with gspice.
a.palfreyman
pfm Member
Some info on noise in J112:
You have to go to this to get much better:
Info courtesy of:
www.mvaudiolabs.com
I set one of the oddment J112s up (that was similar to my reasonably matched pairs) in the following configuration using my bench PSU:
I could then inject ac current (to modulate the supply voltage) at 100Hz, 1kHz and 10kHz. For ∆Vr of 10.0mV RMS, ∆Vz had to be set to 1.07V RMS. That gives impedance of near as damnit 250k with little variance, using the value of R = 2k4 to give 1mA standing current.
At some stage (if I can find that spare ZIF socket I have somewhere) I'm going to try to set up the Jfet pair as configured in OP but with the arrangement top of above image added to 'inject PSU noise' and see what happens...
You have to go to this to get much better:
Info courtesy of:
Modern JFET noise measurements - MV Audio Labs
JFET noise measurements of a modern and obsolete parts. As well as some other active semiconductor devices like LEDS, diodes.
www.mvaudiolabs.com
I could then inject ac current (to modulate the supply voltage) at 100Hz, 1kHz and 10kHz. For ∆Vr of 10.0mV RMS, ∆Vz had to be set to 1.07V RMS. That gives impedance of near as damnit 250k with little variance, using the value of R = 2k4 to give 1mA standing current.
At some stage (if I can find that spare ZIF socket I have somewhere) I'm going to try to set up the Jfet pair as configured in OP but with the arrangement top of above image added to 'inject PSU noise' and see what happens...
a.palfreyman
pfm Member
Ooh, sorry @PigletsDad somehow missed your post above. I'll re-read a couple of times to better understand what you have written.
Not too bothered if the NE5534 is overkill as that is a potential source of assymetric noise relative to output.
Thx,
Andy P
Don't have PC at home unfortunately (well, I do but it's almost as long in the tooth as I and not used in 15 years now...)
Not too bothered if the NE5534 is overkill as that is a potential source of assymetric noise relative to output.
Thx,
Andy P
Don't have PC at home unfortunately (well, I do but it's almost as long in the tooth as I and not used in 15 years now...)
a.palfreyman
pfm Member
Thanks, @PigletsDad, your post was very thought-provoking. As I'm less inclined to think in those terms (non-technical idiot...) we have, if I understand correctly:
So, as you say, PSRR is 40dB to plus rail and 40dB to virtual earth, but non-coherent so adds as root of sum of squares. At 20kHz BW we have noise contributions:
2k4: 0.9uV (two-off)
Jfet: 2nV root Hert = 0.3uV (two-off)
PSU: 1.4uV
(2 x 0.3^2 + 2 x 0.9^2 + 1.4^2)^0.5 = 1.9uV
I'll remeasure impedance with 1k2 as the set resistor is the biggest contributor whilst I'm certain that PSU figure is a gross overestimate too.
Last edited:
a.palfreyman
pfm Member
Right, I set the Jfet up with 1k2 source resistor giving Id of 2mA. impedance is now 128k or 106:1 so PSRR stays around 40dB.
Now, root 2 x 30uV / 106 is 0.4uV.
So if we can get PSU noise down to 30uV on plus rail and same on virtual earth we have (at 20kHz BW) the following noise contributions:
1k2: 0.6uV (two-off)
Jfet: 2nV root Hert = 0.3uV (two-off)
PSU: 0.4uV
(2 x 0.3^2 + 2 x 0.6^2 + 0.4^2)^0.5 = 1.0uV
More like it.
Now, root 2 x 30uV / 106 is 0.4uV.
So if we can get PSU noise down to 30uV on plus rail and same on virtual earth we have (at 20kHz BW) the following noise contributions:
1k2: 0.6uV (two-off)
Jfet: 2nV root Hert = 0.3uV (two-off)
PSU: 0.4uV
(2 x 0.3^2 + 2 x 0.6^2 + 0.4^2)^0.5 = 1.0uV
More like it.
Last edited:
PigletsDad
My intelligence test came back negative.
White noise level is fine, but 1/f is a bit harsh. Sadly, RIAA boosts low frequencies, by about 20dB at 50 Hz so 1/f noise may get a look in. The usual Fletcher-Munson curve makes the ear less sensitive to low frequencies, but the combination of bass boost and rising noise might be audible.
a.palfreyman
pfm Member
I did wonder about that. Double the typical at 50 Hz and double that again at 30Hz...and 2sk170 aren't cheap, if you can find real ones...
There was another good modern Jfet in the link, namely BF862, BUT its 8100dy SMD and tiny...
PigletsDad
My intelligence test came back negative.
What about LSK170, or LSK389 duals from Linear Systems? Or if you want a smaller device with less capacitance (which is where we came in), LSK489. The 1/f on these is very good, lovely modern processing. LSK170 available in TO92, the duals in a TO-71 package, as well as SMD if you want.
a.palfreyman
pfm Member
Bu99er. They are nice Jfets, but a bit spendy for a play project. 
At 30Hz, 8nV root Hertz is 1.1uV noise for 20kHz BW. Two together makes 1.6uV noise and dominates everything.
However, that's still about 66dB below a 3mV nominal cart, or am I missing something about spectral density? (Genuine question, not sarcasm). Always found the subject difficult, so tend not to think about it to much and just calculate as appropriate.
BUT, you can see the difference on a 'scope with / without an RC BW limiting filter.
Actually, a thought did occur. To save all the pra77ing around, what abt 2 x 9V batteries? Mid point is a nice conventional 0V plane that ONLY passes load current, then put a couple of hundred ohm to a 1000uF cap per rail. It's essentially two current sources eating a couple of mA, one of them modulated to maybe 10mV RMS absolute max which is <1uA load current into 47k/few hundred pF. So a stereo pair is 4mA standing current with 1uA modulation. (Could double the price too as it's super-duper battery power)
Edit: the way I see it is that the amplitude of the voltage from a transducer is approximately proportional to 1/f for sound/music. When you reverse RIAA that at about 4dB per octave (13-14dB per decade) you have a signal that is approximately flat amplitude wise (about -10 to 12 dB from low to high freqs.) so these J112 Jfets aren't that bad compared to the signal level passing through.
At 30Hz, 8nV root Hertz is 1.1uV noise for 20kHz BW. Two together makes 1.6uV noise and dominates everything.
However, that's still about 66dB below a 3mV nominal cart, or am I missing something about spectral density? (Genuine question, not sarcasm). Always found the subject difficult, so tend not to think about it to much and just calculate as appropriate.
BUT, you can see the difference on a 'scope with / without an RC BW limiting filter.
Actually, a thought did occur. To save all the pra77ing around, what abt 2 x 9V batteries? Mid point is a nice conventional 0V plane that ONLY passes load current, then put a couple of hundred ohm to a 1000uF cap per rail. It's essentially two current sources eating a couple of mA, one of them modulated to maybe 10mV RMS absolute max which is <1uA load current into 47k/few hundred pF. So a stereo pair is 4mA standing current with 1uA modulation. (Could double the price too as it's super-duper battery power
)Edit: the way I see it is that the amplitude of the voltage from a transducer is approximately proportional to 1/f for sound/music. When you reverse RIAA that at about 4dB per octave (13-14dB per decade) you have a signal that is approximately flat amplitude wise (about -10 to 12 dB from low to high freqs.) so these J112 Jfets aren't that bad compared to the signal level passing through.
Last edited:
a.palfreyman
pfm Member
Found this rather interesting paper on battery noise:
Basically says that for small discharge currents (ie load R >> internal resistance r) or at low charging currents, noise in NiCd batteries is limited to the internal resistance, making them better voltage reference sources than regulators. However, no idea if things have moved on since (doc not dated). So, I'm thinking a pair of NiMH pp3 fed by a voltage reg set to 19.5 to 20v followed by a fet ccs set to about 9mA. As voltage changes in batts, the voltage across the fet ccs will change, so it will deliver more current at lower batt volts, but less current as batt volts increase, so will find a nice equilibrium and can be 'fed' whilst in use. I'm thinking of also putting a good cap (pana FR 100uF/25V or I have some Oscon 100uF/16V SH) across the +/0/- rails too. Interestingly, I measured the FRs at lower leakage current than the SHs, but <10nA in either case.
Basically says that for small discharge currents (ie load R >> internal resistance r) or at low charging currents, noise in NiCd batteries is limited to the internal resistance, making them better voltage reference sources than regulators. However, no idea if things have moved on since (doc not dated). So, I'm thinking a pair of NiMH pp3 fed by a voltage reg set to 19.5 to 20v followed by a fet ccs set to about 9mA. As voltage changes in batts, the voltage across the fet ccs will change, so it will deliver more current at lower batt volts, but less current as batt volts increase, so will find a nice equilibrium and can be 'fed' whilst in use. I'm thinking of also putting a good cap (pana FR 100uF/25V or I have some Oscon 100uF/16V SH) across the +/0/- rails too. Interestingly, I measured the FRs at lower leakage current than the SHs, but <10nA in either case.
justshoemark
pfm Member
Very interested to see how this works out. I have a EAR 802 with mm phono input. And I am struggling to match a cartridge so a loading device is a great idea to sit before the phono input.
a.palfreyman
pfm Member
Here it is:
Not got the PSU yet and awaiting some green LEDs. It has 0.45uF polystyrene output caps.
Output impedance is about 2k4 so it is -0.5dB into 47k load at 1kHz. Bandwidth is 250kHz near as damnit.
Relative to 1kHz, driving 47k//330pF we have:
20Hz : -0.75dB
40Hz : -0.2dB
100Hz : -0.02dB
10kHz : -0.03dB
20kHz : -0.12dB.
Quite pleased.
Only one tiny remark: it has quite a turn on thump,
A neg spike of 1.5V which is very high compared to mV range. No such spike on turn off.
The reg is set to 19.2v and charging current tops out at 8mA via a Jfet ccs. When it's all balanced out (ie the batteries are just acting as a voltage source with current supplied by reg-ccs) it sits at a smidge under 18v.
Not got the PSU yet and awaiting some green LEDs. It has 0.45uF polystyrene output caps.
Output impedance is about 2k4 so it is -0.5dB into 47k load at 1kHz. Bandwidth is 250kHz near as damnit.
Relative to 1kHz, driving 47k//330pF we have:
20Hz : -0.75dB
40Hz : -0.2dB
100Hz : -0.02dB
10kHz : -0.03dB
20kHz : -0.12dB.
Quite pleased.
Only one tiny remark: it has quite a turn on thump,
A neg spike of 1.5V which is very high compared to mV range. No such spike on turn off.
The reg is set to 19.2v and charging current tops out at 8mA via a Jfet ccs. When it's all balanced out (ie the batteries are just acting as a voltage source with current supplied by reg-ccs) it sits at a smidge under 18v.
Last edited:
a.palfreyman
pfm Member
Noise is good too.
This is my 50dB gain, low noise amp with 2k4 on the input into my scope at 2mV/div:
Equivalent to 1uV RMS at 20kHz BW.
Here is the buffer (on battery only: awaiting PSU to arrive) with 390R on the input (to represent the DC impedance of a typical MM cart) into my low noise amp / scope, same settings:
Pretty much same level of white noise (so equivalent to that of 2k4 output impedance) with a tiny bit of flicker noise.
That's approx -70dB WRT a 3mV nominal cart output.
This is my 50dB gain, low noise amp with 2k4 on the input into my scope at 2mV/div:
Equivalent to 1uV RMS at 20kHz BW.
Here is the buffer (on battery only: awaiting PSU to arrive) with 390R on the input (to represent the DC impedance of a typical MM cart) into my low noise amp / scope, same settings:
Pretty much same level of white noise (so equivalent to that of 2k4 output impedance) with a tiny bit of flicker noise.
That's approx -70dB WRT a 3mV nominal cart output.