Good, as you do not use the output stage from the AD844. I suggest a simple JFet Buffer as an output though.
Borbely style, the one with only N-Channel Fet's is similar to Pass B1.
Attachments
At times there can be a difference between measures results and sound quality. Not everything that measure betters also sounds better.
Dear SoNic,
Here the tubes output are not the weakest point of the chain. Did you've ever built a FM stereo tuner made of 100% tubes?, I've did it some years ago, and believe me when I tell you that it was the most quiet and wonderful FM receiver I've never heard in all my life. This only probes that the tubes still has a lot to offer......even in 21st century with all micro mini devices and all kind of wonderful gadgets.
Kind regards.
Here the tubes output are not the weakest point of the chain. Did you've ever built a FM stereo tuner made of 100% tubes?, I've did it some years ago, and believe me when I tell you that it was the most quiet and wonderful FM receiver I've never heard in all my life. This only probes that the tubes still has a lot to offer......even in 21st century with all micro mini devices and all kind of wonderful gadgets.
Kind regards.
Hi, Remembering the inductance of the S&B the cap value needs to be >3.3uF. In that value (4.7uF actually) Musicap, Auricap, Motor-run by themselves and then bypassed with 0.1uF Vcap. I'm ok with the Solen tin foil caps, but not here in that sort of value.
PS 6CJ3 , 10H to 52uF, CCS fed VR Shunt 0A2 with/out RC bypass.
Like I said, it was good, just not as direct sounding. Could live with it if I hadnt heard the OPA. Then again, maybe the 6072 is a better tube than 396A/5670 or 6922. Still interested in 6SL7 too. Agree on the low capacitance CCS. I like Mr Pimms self biased circuit. The Mu out sounds the same to me direct plate connection (with a reasonably low Rp tube), no reason to assume the Mu out is not completely acceptable here with high Rp tubes.
AFAIK his OPA861 cct wasnt in public domain. I'd suggested contacting him through http://audialonline.com/ and asking him directly.
I did own some 25-30 years ago a tube radio (with FM). Call it "quiet" compared with a DAC is just hillarious...
Now is not 25-30 years ago, and own is not the same as build. Anyway, I think this thread is about any good TDA1541A kit, not tubes vs op amps or anything else.
Kind regards.
Hi,
Let us think about this a little, shall we?
The TDA1541 will swing between 0ma and -4mA output current, with a "digital silence" centre-point of -2mA.
So with a 33 Ohm I/V conversion resistor the output of the TDA1541 and thus the gate voltage of the 6072A will swing between 0mV and -132mV, with a centre-point of -66mV.
Now we MAY have a miniscule grid current flowing with Vgk = 0V, however, the gridvoltage can never become positive, so no appreciable grid current will happen.
This approach ONLY works for the TDA1541/TDA1541A and no other DAC known to me and ONLY if the output current offset is NOT compensated. It is not a generic Circuit applicable to a wide range of DAC's.
Ciao T
Let us think about this a little, shall we?
The TDA1541 will swing between 0ma and -4mA output current, with a "digital silence" centre-point of -2mA.
So with a 33 Ohm I/V conversion resistor the output of the TDA1541 and thus the gate voltage of the 6072A will swing between 0mV and -132mV, with a centre-point of -66mV.
Now we MAY have a miniscule grid current flowing with Vgk = 0V, however, the gridvoltage can never become positive, so no appreciable grid current will happen.
This approach ONLY works for the TDA1541/TDA1541A and no other DAC known to me and ONLY if the output current offset is NOT compensated. It is not a generic Circuit applicable to a wide range of DAC's.
Ciao T
Hey T,
As I said they will "clip" at the same time. Actually it will work for the ancient TDA1540 too(joke).
I was more thinking of any unlinear effects when biasing a tube so close to Ug=0V. I have actually seen a few russian tubes in Klausmobiles tests beginning to draw gridcurrent just above Ug=0V. So let´s stay away from them.......
As I said they will "clip" at the same time. Actually it will work for the ancient TDA1540 too(joke).
I was more thinking of any unlinear effects when biasing a tube so close to Ug=0V. I have actually seen a few russian tubes in Klausmobiles tests beginning to draw gridcurrent just above Ug=0V. So let´s stay away from them.......
I've found a HP 4194A to play with and a 47uF case D polymer in combination with a few comparatively low value ceramics is looking very nice when it comes to impedance flatness atleast up to 40Mhz wich is the upper measurement limit of the analyzer. (I've only done some trial measurements thus far making sure the calibration is decent, not added any DC bias)
Since it's easier to make the impedance overal flat if it magnitude is kept in the region of 100-200mohm this is because the inductive behavior of the regulator at high frequency gets more damped if it has a larger resistive part in its complex output impedance. This is also helped by capacitors not having extremely low esr or high parasitic inductance. Which also helps in reducing the Q factor's of the many different tank circuits that is formed by them, the regulator and the many parasitcs.
So the question becomes what to prefer, impedance flatness or lower overall level accepting greater ripple in the impedance magnitude response? Maybe its comparatively easy to answer for the 1541' since it (according to thorstenL) does not create much disturbance above the highest clockrate it sees anyway, if so maybe we can almost get rid of the ceramics altogether if we do a good enough job with the layout (and use modern smd polymer caps with lowish values).
Another aproach perhaps more suitable for the cmos circuits needed could be to try and tune the tank circuit for minimum impedance at the fundamental of the clock freqency and maybe its 3:rd 5:th, 7:th and even higher overtones.
Hi,
Well, I would not call it clipping. Also, the source impedance is very low. This limits any grid current effects. Yes, TDA1540 is also applicable. In this case oversampling must be retained though.
But we do not go above 0V and we can look at it from another angle.
By using essentially zero bias (the -66mV will provide a minimal bias, but MINIMAL) we operate the tube in the most linear region (look at Rp vs. I and u vs. I), plus we dispense with any biasing components, which especially in the case of cathode bias are not very inaudible.
In fact, I have become quite fond of near zero bias operation of tubes. I would not use if for commercial designs, as there is enough variation in production tubes that some may just not work, but for a DIY project it makes an immense amount of sense...
It allows us to draw down the circuit to the essential, that is an input voltage that never exceeds 0V and only swings negative, a triode, a source of anode current and an output capacitor.
Of course, we can do this only because our input voltage is very precisely and unambiguously defined, as said, it is a very specific solution, not a generic one.
Ciao T
Well, I would not call it clipping. Also, the source impedance is very low. This limits any grid current effects. Yes, TDA1540 is also applicable. In this case oversampling must be retained though.
But we do not go above 0V and we can look at it from another angle.
By using essentially zero bias (the -66mV will provide a minimal bias, but MINIMAL) we operate the tube in the most linear region (look at Rp vs. I and u vs. I), plus we dispense with any biasing components, which especially in the case of cathode bias are not very inaudible.
In fact, I have become quite fond of near zero bias operation of tubes. I would not use if for commercial designs, as there is enough variation in production tubes that some may just not work, but for a DIY project it makes an immense amount of sense...
It allows us to draw down the circuit to the essential, that is an input voltage that never exceeds 0V and only swings negative, a triode, a source of anode current and an output capacitor.
Of course, we can do this only because our input voltage is very precisely and unambiguously defined, as said, it is a very specific solution, not a generic one.
Ciao T
I did some LTSpice sims a couple of years ago that might interest you on this. I added a helper MOSFET to lower the impedance of the TL431:
http://www.diyaudio.com/forums/digital-line-level/161968-my-complete-solution-dac-rfc-2.html#post2106563
With the 33 ohms I/V, I believe the tube "sees a voltage source" with output impedance near 33ohm, so its a bit analogous to simpleSE driving class A2 (over 0V grid voltage) with mosfet help.
Also remember it was common to self bias 12Ax7's with the cathode tied to ground and using a large grid leak resistor the grid current developed a negative -0.1V or so, this was used in the old days.
I say anytime you can get rid of a cathode bypass cap, its worth trying.
But Ove's experiements show much lower 2H distortion simply by canceling the -2mA offset across a passive I/V using a simple jfet ccs. His page of actual measurements points toward the TDA1541 not favoring DC or AC developing betwen i-out and aground.
Hi,
20uA would be an error of around 1% for 2mA Iout.
However, if you look at something like the 6N1P at 4mA (which is close to 6072A) or the 6N23 at 15mA (which is a 6922 or ECC88 etc.) you will see that gridcurrent remains < 5uA.
It is not precisely "0V Bias". Another application is an MC Cartridge input stage, even MM can be sucessful, as signal levels are only a few mV...
Ciao T
20uA would be an error of around 1% for 2mA Iout.
However, if you look at something like the 6N1P at 4mA (which is close to 6072A) or the 6N23 at 15mA (which is a 6922 or ECC88 etc.) you will see that gridcurrent remains < 5uA.
It is not precisely "0V Bias". Another application is an MC Cartridge input stage, even MM can be sucessful, as signal levels are only a few mV...
Ciao T
Hi,
Ove's experiment shows that for the value I/V resistor he selected AND for the use of the ECC88 (especially) he achieved a certain result.
It is probably not a good idea to carelessly generalise from this.
Also, personally I rarely worry about a bit of 2nd HD...
Ciao T
Ove's experiment shows that for the value I/V resistor he selected AND for the use of the ECC88 (especially) he achieved a certain result.
It is probably not a good idea to carelessly generalise from this.
Also, personally I rarely worry about a bit of 2nd HD...
Ciao T
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.