Correct.
For CEN you can half the value, as the signal goes through both arms in parallel.
Another argument for CEN.
🙂
Patrick
For CEN you can half the value, as the signal goes through both arms in parallel.
Another argument for CEN.
🙂
Patrick
Last edited:
Well Patrick, with some help from Nic and Paul I got my first dirty prototype dual mono Sabre and Sen up and running and pleasing it is too.
A lot of work yet to do on the implementation but for now I have a building block and it has sold me on Sen.
Thanks
A lot of work yet to do on the implementation but for now I have a building block and it has sold me on Sen.
Thanks
Hi Patrick
Did you experiment with different electrolytic coupling types, and, if so which type did you prefer?
Paul
Did you experiment with different electrolytic coupling types, and, if so which type did you prefer?
Paul
No. We do not consider it a critical component.
Nichicon & Elna sounded similar.
Of course you can try Black Gate N. If you can still get them.
But I am not going to recommend obsolete components unless we have tins of them.
😉
Patrick
Nichicon & Elna sounded similar.
Of course you can try Black Gate N. If you can still get them.
But I am not going to recommend obsolete components unless we have tins of them.
😉
Patrick
An alternative, even if you are using ES9018 as 4mA per channel (i.e. 680R R_iv) would be, in parallel:
220uF (or 470uF if you have to) bipolar electrolytic
2.2uF WIMA MKS2, or equivalent
Panasonic 0.1uF 16V PPS
The 2.2uF MKS will take care of signals from 106Hz, and the PPS from 2.3kHz.
So the electrolytic is only effective for signals below 100Hz.
Of course if you use lower R_iv, then you need to make a new PCB to allow bigger caps.
Patrick
220uF (or 470uF if you have to) bipolar electrolytic
2.2uF WIMA MKS2, or equivalent
Panasonic 0.1uF 16V PPS
The 2.2uF MKS will take care of signals from 106Hz, and the PPS from 2.3kHz.
So the electrolytic is only effective for signals below 100Hz.
Of course if you use lower R_iv, then you need to make a new PCB to allow bigger caps.
Patrick
Hi all, I get a Buffalo II and I make the I/V conversion with a Borbely I/V converter, but I'm disappointed about the impact in the bass . I read on this thread that the ESS9018
(configured in stereo mode) need high grade Idss JFET to work correctly. I'm interested on Sen I/V, but also TPA' Legato3.1 is worth of interest,cause it's dedicated to Sabre DAC.
Did someone compared these I/V convs?
Thanks.
(configured in stereo mode) need high grade Idss JFET to work correctly. I'm interested on Sen I/V, but also TPA' Legato3.1 is worth of interest,cause it's dedicated to Sabre DAC.
Did someone compared these I/V convs?
Thanks.
Since you disabled all messages and email to you, would you kindly send me a email via the forum ?
Thanks,
Patrick
Thanks,
Patrick
Why does Sen sound so good?
Anyone have any ideas why Sen sounds so good?
We can take low noise and distortion for granted but they don't explain the difference between Sen and an OP AMP I/V.
Here are some candidate theories:
A) Does a good job of rejecting RFI coming in from the output cable?
B) Distortion profile rising with frequency reintroduces harmonics stripped out by digital mastering (sound is more accurate than the digital master)?
C) Harmonic distortion profile happens to be pleasing (sound is more musical than the digital master)?
F) Low IMD?
Anyone got any more ideas?
Anyone have any ideas why Sen sounds so good?
We can take low noise and distortion for granted but they don't explain the difference between Sen and an OP AMP I/V.
Here are some candidate theories:
A) Does a good job of rejecting RFI coming in from the output cable?
B) Distortion profile rising with frequency reintroduces harmonics stripped out by digital mastering (sound is more accurate than the digital master)?
C) Harmonic distortion profile happens to be pleasing (sound is more musical than the digital master)?
F) Low IMD?
Anyone got any more ideas?
Hi,
well, not exactly an academic explanation but an OPAmp is like a feature and gadget filled over-motorized luxury car driven by pushing the pedal to the metal and breaking as hard as possible at the same. Sure one way to race at incredible 10mph. But will it blend?
The simple discrete circuit is like the F1-racer stripped down to the minimum.
Gas or break as needed
But then, probabely just magic 😉
jauu
Calvin
well, not exactly an academic explanation but an OPAmp is like a feature and gadget filled over-motorized luxury car driven by pushing the pedal to the metal and breaking as hard as possible at the same. Sure one way to race at incredible 10mph. But will it blend?
The simple discrete circuit is like the F1-racer stripped down to the minimum.
Gas or break as needed
But then, probabely just magic 😉
jauu
Calvin
I don't believe in magic. I'm guessing A and C.
I think my recent Sen variant fell down on point A - though I'm having a go at fixing it with an extra capacitor.
Your A/B testing of the cascoded Sen variant could provide some evidence about point C.
Some other reasons for Sen to work well compared to alternatives:
The JFET is intrinsically less prone to envelope detection of out-of-band (RF) signals than are bipolars, especially when the latter are biased at typical levels in the first stages of opamps.
The Sen's input is a common-gate wide-channel JFET (Scott Wurcer remarked that the chip is larger than some entire video amps from AD). The input resistance approaches the reciprocal of the device transconductance, i.e., it's pretty low to begin with. The traditional opamp approach relies on high negative feedback to reduce the initially highish input impedance to a lower value, and depending on opamp open loop gain the resulting impedance looks resistive at low frequencies but soon begins to look inductive. The overall opamp circuit must be designed carefully to take into account the capacitance of the source. Otherwise oscillatory instability will result. And the capacitance may be code-dependent.
The floating supply of the Sen when provided by batteries achieves good isolation from mains noise, and as noted the Sen has fairly high power supply rejection anyway, even in its simplest form.
The noise associated with the current sink of the matching device adds 3dB to the equivalent input voltage noise, which in turn determines what the current noise will be based on the output impedance of the DAC. But SK170s are rather low noise to begin with, quite a bit better than the vast majority of IC opamps.
When fed from a high output impedance DAC the distortion of Sen is low at low frequencies, since the only loss of current is to gate leakage. At higher frequencies the device capacitances begin to affect the output current. When the DAC output impedance is lower other distortion mechanisms appear, and depending on the DAC may have code-dependencies. Whether or not these are pleasing to the ear will be, necessarily, a matter of personal opinion.
The JFET is intrinsically less prone to envelope detection of out-of-band (RF) signals than are bipolars, especially when the latter are biased at typical levels in the first stages of opamps.
The Sen's input is a common-gate wide-channel JFET (Scott Wurcer remarked that the chip is larger than some entire video amps from AD). The input resistance approaches the reciprocal of the device transconductance, i.e., it's pretty low to begin with. The traditional opamp approach relies on high negative feedback to reduce the initially highish input impedance to a lower value, and depending on opamp open loop gain the resulting impedance looks resistive at low frequencies but soon begins to look inductive. The overall opamp circuit must be designed carefully to take into account the capacitance of the source. Otherwise oscillatory instability will result. And the capacitance may be code-dependent.
The floating supply of the Sen when provided by batteries achieves good isolation from mains noise, and as noted the Sen has fairly high power supply rejection anyway, even in its simplest form.
The noise associated with the current sink of the matching device adds 3dB to the equivalent input voltage noise, which in turn determines what the current noise will be based on the output impedance of the DAC. But SK170s are rather low noise to begin with, quite a bit better than the vast majority of IC opamps.
When fed from a high output impedance DAC the distortion of Sen is low at low frequencies, since the only loss of current is to gate leakage. At higher frequencies the device capacitances begin to affect the output current. When the DAC output impedance is lower other distortion mechanisms appear, and depending on the DAC may have code-dependencies. Whether or not these are pleasing to the ear will be, necessarily, a matter of personal opinion.
Very well thought out analysis, I couldn't agree more about the oscillation opamp issue.
Also I agree about the importance of a output impedance with a current output DAC, it is often overlooked. This has been compensated with modern DAC's by jacking up the output current levels, in otherwords one can get away with lower output impedance via higher output current. But especially with higher resolution DAC's where the LSB is very very low the output impedance causes the level to to be even lower from resistive division, burying it in noise but the higher the input impedance is of the I/V the bigger the issue.
So we have two completely different approaches to dealing with an i-out DAC.
1. High current, low output impedance feeding a high performance opamp I/V with a low "on paper" input impedance, gives great classic TDN+N numbers.
2. Lower current, higher output impedance DAC feeding a SEN with higher imput impedance, but higher stability and more consistant thru the bandwidth.
That's why IMHO to get the most out of the Sen, using an ESS9018 is kind of a mismatch. Not saying the SEN isn't one of the best choices for the ESS9018, rather that there are better choices for the digital side.
Evil mentioned concern with offering a PCM1704 digital section because of Ti marking it as planned obselesent. But the instumentation industry still buys the PCM1704 over the modern choices, every batch sells out each time Ti doubles the price for the past several years, there are important reasons for this that I hope the developers of the Sen don't forget that.
My name is EUVL, and I hope not too evil. 😉
We use PCM1704 or AD1865 ourselves, as mentioned in the original article.
But there are people who bought ES9018s. So we only did it by popular demand.
Good that an independent source confirms our choice. 🙂
Patrick
We use PCM1704 or AD1865 ourselves, as mentioned in the original article.
But there are people who bought ES9018s. So we only did it by popular demand.
Good that an independent source confirms our choice. 🙂
Patrick
One additional advantage I didn't mention: the low pinchoff voltage JFETs like the SK170 and BF862 have a rather low temperature coefficient of gate-source voltage around Idss. Of course upper and lower devices (or multiples of devices in those cases) are to be matched anyway. But the effect of signal swings on power dissipation will still be there even with matching. So having each device have an intrinsically lower sensitivity to self-heating is helpful to reduce thermally-induced distortions. These are very messy to model, so most of the time our simple simulators don't bother.
If one doesn't mind a little negative feedback, what does look promising, from a consideration of dealing with higher output current DACs especially, is to put the input common-gate device(s) in a tight fast loop. The amplifier involved can be a rather simple topology as it only has to drive one or more JFET gates, and can itself be made out of JFETs. The power supply for the ancillary amp can be referenced to ground, that is, in a conventional non-floating arrangement. The basic Sen core should still use the floating supply.
Simulation results are encouraging, but I haven't built it yet for test and auditioning. BTW this is not the same as the circuit I showed recently in a new thread, using a bipolar Boxall pair as the input device, which was designed to perform a current inversion and offer one approach to differential-to-single-ended conversion, but simply an augmentation of the basic Sen or cascoded Sen.
Brad
If one doesn't mind a little negative feedback, what does look promising, from a consideration of dealing with higher output current DACs especially, is to put the input common-gate device(s) in a tight fast loop. The amplifier involved can be a rather simple topology as it only has to drive one or more JFET gates, and can itself be made out of JFETs. The power supply for the ancillary amp can be referenced to ground, that is, in a conventional non-floating arrangement. The basic Sen core should still use the floating supply.
Simulation results are encouraging, but I haven't built it yet for test and auditioning. BTW this is not the same as the circuit I showed recently in a new thread, using a bipolar Boxall pair as the input device, which was designed to perform a current inversion and offer one approach to differential-to-single-ended conversion, but simply an augmentation of the basic Sen or cascoded Sen.
Brad
Why use negative feedback when you get better than -110dB without ?
It is so simple as it is, and yet so high performance.
There are people who tried passive IV of 33R with the PCM1704 with success.
So Zin is not an issue as far as I am concerned.
🙂
Patrick
It is so simple as it is, and yet so high performance.
There are people who tried passive IV of 33R with the PCM1704 with success.
So Zin is not an issue as far as I am concerned.
🙂
Patrick