High End Praha 2010 - Výstava špièkové audiovizuální techniky
They will certainly prepare english version for 2011, I hope
Hi Joachim, the Vogel book shows the emitter resistor 3.3ohm. If you are using 1ohm then, while keeping the same gain and EQ, you need x3 the cap (that's 1.5uF). That's a lot of cap to load the output with, and I wonder how good the amp step response is. Care to address the other questions I asked in my post #1952? I really appreciate your inputs...
Karel Capek is one of the greatest Czech authors. Yes, he invented the name 'Robot"
RUR book (Rosum Universal Robots).
This is what i know. It is unfortunately in German:
http://de.wikipedia.org/wiki/Karel_Čapek
My only longer experience with Manley electronics is the Steelhead in Michael Fremers system. He kept it for a very long time and i do not know if he still uses it but it was hard to beat in it´s class over a long period. I am also friends with VTL and have heard some older designs from Lukes father. Lukes stuff is much more relyable though and performs much better in the lab but his father did the big breakthoughs i think. His book is an interesting read. I was particular surprised about his grounded grid phonostages.
http://de.wikipedia.org/wiki/Karel_Čapek
My only longer experience with Manley electronics is the Steelhead in Michael Fremers system. He kept it for a very long time and i do not know if he still uses it but it was hard to beat in it´s class over a long period. I am also friends with VTL and have heard some older designs from Lukes father. Lukes stuff is much more relyable though and performs much better in the lab but his father did the big breakthoughs i think. His book is an interesting read. I was particular surprised about his grounded grid phonostages.
No YWN, the gain goes up 4x and the EQ over 20kHz is even more accurate because the pole is shifted higher. I did not change the values of the RIAA components.
The buffer i use can swing 250mA clean and yes at very high frequencies it has to drive the 0.77 Ohm resistor. By then the amplitute is of cause very low. At 20kHz it amounts to ca.75 Ohm. To avoid trouble i added a safty resistor behind the buffer of around 150 Ohm.
That idea stolen from Mr.Vogel. Of cause i have no right to explain what i did because i have no idea how the circuit works acording to the experts.
I wanted to elaborate on the step response but again spare me aditional ambarrasment.
What i am willing to explain is the constat current source. I can not spot that in the bespoke circuit so i can safely asume that i am alowed to voice my limited experience.
The buffer i use can swing 250mA clean and yes at very high frequencies it has to drive the 0.77 Ohm resistor. By then the amplitute is of cause very low. At 20kHz it amounts to ca.75 Ohm. To avoid trouble i added a safty resistor behind the buffer of around 150 Ohm.
That idea stolen from Mr.Vogel. Of cause i have no right to explain what i did because i have no idea how the circuit works acording to the experts.
I wanted to elaborate on the step response but again spare me aditional ambarrasment.
What i am willing to explain is the constat current source. I can not spot that in the bespoke circuit so i can safely asume that i am alowed to voice my limited experience.
Thats not easy. One winding off can cause more then 2% change of value but Neumann has used LCR circuits too to encode the cutter machine. Typical Japanese thinking is to do the oposite in the RIAA stage to do philosophically speaking "the same in the oposite direction", lending it a measure of "authenticity".
We have investigated the famous Tango LCR and i was not totally impressed about the accuracy. I can post a photo of an LCR decoder that a friend of mine build. It can be done but it is expensive if done right. I for one did not try it but i could with the help of my friend.
We have investigated the famous Tango LCR and i was not totally impressed about the accuracy. I can post a photo of an LCR decoder that a friend of mine build. It can be done but it is expensive if done right. I for one did not try it but i could with the help of my friend.
Thanks, so you are using x4 gain compared to the Vogel book design. so about 60dB, right?
Am I right to believe there's a very poor PSRR in this Vogel design and the total noise is dominated by the power supply noise? Can't wait about your constant current source explanation. That would vastly improve the PSRR, but I don't see any simple way to make it working and avoid "current fighting" with the jfets.
Unfortunately i can not find any measurements of the Zanden on the web. If it has an inacurate RIAA and distortion over 0.1% then the magical sound could be explained somewhat. I can ask John Atkinson if he has kept some files. Other wise i am at a loss why that product is so well received by the ones that can afford it. It is certainy not something i could afford so i shoud shut up. What i will do is listen to the new Accuphase and Burmester creations. They are surprisingly aplenty here. I should visit Holger soon.
Will measure its RIAA curve for you if I will visit again my friend's system who owns one. Its he is living far and I got to carry the laptop and the Emu interface in the crammed 1 hour city train travel too. But I will remember it if a chance will be given.
YWN, yes, the gain is 60dB. Vogel has put a cap multiplyer in front of the resistor and his PSU is very elaborated, so yes, he has done an efford to make the PSU clean.
As far as measurements go he succeeded. They are published in his book. My choise of a constant current source was not only motivated because of the better PSU rejection but also because a wanted more gain. I had to make up at least 13dB or i end up with more distortion so what whould be the advantage ?
Back to busyness you see the most basic current source in the attachement, a bipolar transistor with some resistors. This is an NPN that is more common but a PNP source works the same with oposite polarity. Let us do some asumptions and simplifications first. We have to start somewhere. We asume that Iq is much bigger then Ib and that Ib is nearly zero. Iq is aproximately Ub devided by R1 + R2 and Iq x R2 is aproximately Ia x R3 + Ube. You can now find Ia by putting in ca. 0.7V for Ube, the Base-Emitter voltage. What is wrong with that ?
As far as measurements go he succeeded. They are published in his book. My choise of a constant current source was not only motivated because of the better PSU rejection but also because a wanted more gain. I had to make up at least 13dB or i end up with more distortion so what whould be the advantage ?
Back to busyness you see the most basic current source in the attachement, a bipolar transistor with some resistors. This is an NPN that is more common but a PNP source works the same with oposite polarity. Let us do some asumptions and simplifications first. We have to start somewhere. We asume that Iq is much bigger then Ib and that Ib is nearly zero. Iq is aproximately Ub devided by R1 + R2 and Iq x R2 is aproximately Ia x R3 + Ube. You can now find Ia by putting in ca. 0.7V for Ube, the Base-Emitter voltage. What is wrong with that ?
Yes, JC, a current source can inject noise but i was able to minimise that to a point where the circuit is dead quiet. I thought the circuit was dead also and now i am in the middle of it to elaborate. Well, if that is the wish of the comunity i surrender.
Yes Salas that whould be extremely interesting and if you can please also measure noise and distortion.
Yes Salas that whould be extremely interesting and if you can please also measure noise and distortion.
You are right, but the Vogel design noise performance is not better than the 3.3ohm emitter resistor noise. You are using 1ohm in the sources, so you are expecting much better noise, you need therefore a much better power supply.
Not sure why you need so much more loop gain, the input stage plus a good opamp can easily provide 60dB, enough to kill any trace of distortions.
The problem with the constant current source as you described it is in the voltage at the junction between the jfets and the pnp bipolar current source. That voltage is ill defined and most likely, the amp output will flip and stick to the power supply rail or to the ground.
Also I think John Curl is right, a current source will add to the noise.
Old Toshiba fets
Joachim,
The Toshiba 40ms fet actually came out first, as 2SK147 in TO92L (tall) package, I first used them in 1978 for MC input. There was also a p channel 2SJ72, with much higher capacitance and much lower gate current than the n channel and duals of both polarities composed of two TO92L enclosed in an aluminum can, 2SK146 and 2SJ73 that appeared to use the symmetrical nature of these fets as they were face to face but pins corresponded (gate in the middle). Later the same die were packaged in TO92 and TO92s with new numbers. If you are going to CES I will bring you some old ones. to compare with your newer parts.
Joachim,
The Toshiba 40ms fet actually came out first, as 2SK147 in TO92L (tall) package, I first used them in 1978 for MC input. There was also a p channel 2SJ72, with much higher capacitance and much lower gate current than the n channel and duals of both polarities composed of two TO92L enclosed in an aluminum can, 2SK146 and 2SJ73 that appeared to use the symmetrical nature of these fets as they were face to face but pins corresponded (gate in the middle). Later the same die were packaged in TO92 and TO92s with new numbers. If you are going to CES I will bring you some old ones. to compare with your newer parts.
I win around 0.2nVrHz by lowering the resistor to a bit less then 0.8 Ohm compared to Vogels choise. I loose around 1dB of SN if i can trust Sigurd Ruschkovkis simulation he did on another circuit that used current mirrors as CCSs with aproximately the same dimensioning.
It surprised me too but a very good compromise can be found. As far as i can see from the Data Sheet the Fets i use have a Gm of 50mS at 10mA so theoretically it could be slightly better then 0.48nVrHz but not much. What got me enthusiastic about this particular Fets is the very low noise they make under 1kHz. Subjectively i hear less noise at the listening space then my 4 plus 4 2SK170, 2SJ74 stages.
I will do measurements and compare. I have a lot stages lying around with other technology. Better still my yesterday "free your mind session" was rewarded with an idea for an MPP that solves the DC offset problem i think so i should build this for comparison. I actually wanted to use constant current sources there too but to give the discussion a measurement backoff i will build too versions and compare.
Back to the current source YWN, your thoughts are two steps forward from mine.
The problem with that simple circuit i see is that Ube is dependent on temperature by the formular 2mV per Kelvin devided by R3 ( aproximately, sorry that i repeat that so often). Temperature sensitivity is also in the cascode. The Fet i choose is rock solid though because it has a very high early voltage. That translates into a nearly horizontal line when you plot ID in mA against UDS in Volts, even at the rather high ID of 10mA per Fet i choose. The task now is to compensate the current source against the bipoar cascode transistor. I checked iddle current over the 0.8Ohm resistor and it was rock stable. I can do that with more resolution of cause. The trouble comes from the bipolar cascode transistor i think.
It surprised me too but a very good compromise can be found. As far as i can see from the Data Sheet the Fets i use have a Gm of 50mS at 10mA so theoretically it could be slightly better then 0.48nVrHz but not much. What got me enthusiastic about this particular Fets is the very low noise they make under 1kHz. Subjectively i hear less noise at the listening space then my 4 plus 4 2SK170, 2SJ74 stages.
I will do measurements and compare. I have a lot stages lying around with other technology. Better still my yesterday "free your mind session" was rewarded with an idea for an MPP that solves the DC offset problem i think so i should build this for comparison. I actually wanted to use constant current sources there too but to give the discussion a measurement backoff i will build too versions and compare.
Back to the current source YWN, your thoughts are two steps forward from mine.
The problem with that simple circuit i see is that Ube is dependent on temperature by the formular 2mV per Kelvin devided by R3 ( aproximately, sorry that i repeat that so often). Temperature sensitivity is also in the cascode. The Fet i choose is rock solid though because it has a very high early voltage. That translates into a nearly horizontal line when you plot ID in mA against UDS in Volts, even at the rather high ID of 10mA per Fet i choose. The task now is to compensate the current source against the bipoar cascode transistor. I checked iddle current over the 0.8Ohm resistor and it was rock stable. I can do that with more resolution of cause. The trouble comes from the bipolar cascode transistor i think.