I forgot to mention :
All BJTs are BC550C / BC560C matched.
Current sources are E153 matched.
If you want more drive at the output, you can parallel multiple transistors at the output stage.
Thermal coupling still crucial.
Supplies +/-12~15V.
Total consumption about 75mA as shown (single ended).
Will probably need a better heatsink than our normal Quad heatsink.
Note also phase reversal as compared with CEN / SEN.
Patrick
.
All BJTs are BC550C / BC560C matched.
Current sources are E153 matched.
If you want more drive at the output, you can parallel multiple transistors at the output stage.
Thermal coupling still crucial.
Supplies +/-12~15V.
Total consumption about 75mA as shown (single ended).
Will probably need a better heatsink than our normal Quad heatsink.
Note also phase reversal as compared with CEN / SEN.
Patrick
.
Last edited:
Jan,
You are absolutely right.
Pin 2 to i_in, Pin 3 to Vref, Pin 5 to Riv & Civ, done.
Small problem -- Zin = 65R.
Have to do discrete, I am afraid.
Of course you can also put 65x AD844 in parallel.
Patrick
Agreed. This is one area where your discrete solution outperforms the intergrated chip.
jan
The AD844 works but the circuit you show is simpler and better sounding i assume.
I have no objection whatsoever that a PCB gets designed. I would ad the following improvements : Input bias cancelation, better suppression of bias noise to the input transistors, a better current mirror. This is for the input stage and i build circuits like that,
that sounded excellent as an MC phono stage. I could draw a circuit if you wish. There are more elaborate versions of the diamond buffer and other alternatives but what i like with your circuit is that it shows the principles very clear in it´s most basic form. The main advantage is that a bipolar has about 10x Gm with the same idle current then a J-Fet so input impedance of that circuit is very low. That is good for some DA converters.
I have no objection whatsoever that a PCB gets designed. I would ad the following improvements : Input bias cancelation, better suppression of bias noise to the input transistors, a better current mirror. This is for the input stage and i build circuits like that,
that sounded excellent as an MC phono stage. I could draw a circuit if you wish. There are more elaborate versions of the diamond buffer and other alternatives but what i like with your circuit is that it shows the principles very clear in it´s most basic form. The main advantage is that a bipolar has about 10x Gm with the same idle current then a J-Fet so input impedance of that circuit is very low. That is good for some DA converters.
Here is the input stage that i would use. Forget the AC coupling and the second stage.
R1 and P1 build a local feedback loop for trimming input DC offset. It works very stable and has to be adjusted ones the circuit is in the chain. Of cause you can use this trim also to insert a certain positive or negative DC if you wish. The input transistors are biased by fets that supply current to a current mirror. Depending on the fet you can arrange the idle at will by adjusting the source trimmers. This adjustment gives you some flexibility over the input impedance too. The value of 3.5mA was chosen for a Denon DL103 that has a resistive impedance of 40 Ohm. More idle gives you less input impedance. From the current mirror you see an RC filter supplying the input transistors.
This gives much better PSR then going from the transdiodes direct. 10.000uF is a bit extreme but it should have at least 2200uF. The sound benefits much from a big cap. The transistors chosen work extremely well in this circuit although with 30 Ohm Rbee they do not have the lowest value of base spread resistance in the business. Nevertheless the circuit is dead quiet. People that build it first thought that the circuit did not work so quiet it is.
R1 and P1 build a local feedback loop for trimming input DC offset. It works very stable and has to be adjusted ones the circuit is in the chain. Of cause you can use this trim also to insert a certain positive or negative DC if you wish. The input transistors are biased by fets that supply current to a current mirror. Depending on the fet you can arrange the idle at will by adjusting the source trimmers. This adjustment gives you some flexibility over the input impedance too. The value of 3.5mA was chosen for a Denon DL103 that has a resistive impedance of 40 Ohm. More idle gives you less input impedance. From the current mirror you see an RC filter supplying the input transistors.
This gives much better PSR then going from the transdiodes direct. 10.000uF is a bit extreme but it should have at least 2200uF. The sound benefits much from a big cap. The transistors chosen work extremely well in this circuit although with 30 Ohm Rbee they do not have the lowest value of base spread resistance in the business. Nevertheless the circuit is dead quiet. People that build it first thought that the circuit did not work so quiet it is.
Attachments
Good article and great thread.
I have a couple of AD1862 lying around.. it seems to be a perfect fit for your SEN/CEN circuit.
Thanks
Valeriano
Last edited:
Here is the input stage that i would use. Forget the AC coupling and the second stage.
R1 and P1 build a local feedback loop for trimming input DC offset. It works very stable and has to be adjusted ones the circuit is in the chain. Of cause you can use this trim also to insert a certain positive or negative DC if you wish. The input transistors are biased by fets that supply current to a current mirror. Depending on the fet you can arrange the idle at will by adjusting the source trimmers. This adjustment gives you some flexibility over the input impedance too. The value of 3.5mA was chosen for a Denon DL103 that has a resistive impedance of 40 Ohm. More idle gives you less input impedance. From the current mirror you see an RC filter supplying the input transistors.
This gives much better PSR then going from the transdiodes direct. 10.000uF is a bit extreme but it should have at least 2200uF. The sound benefits much from a big cap. The transistors chosen work extremely well in this circuit although with 30 Ohm Rbee they do not have the lowest value of base spread resistance in the business. Nevertheless the circuit is dead quiet. People that build it first thought that the circuit did not work so quiet it is.
R1 and P1 build a local feedback loop for trimming input DC offset. It works very stable and has to be adjusted ones the circuit is in the chain. Of cause you can use this trim also to insert a certain positive or negative DC if you wish. The input transistors are biased by fets that supply current to a current mirror. Depending on the fet you can arrange the idle at will by adjusting the source trimmers. This adjustment gives you some flexibility over the input impedance too. The value of 3.5mA was chosen for a Denon DL103 that has a resistive impedance of 40 Ohm. More idle gives you less input impedance. From the current mirror you see an RC filter supplying the input transistors.
This gives much better PSR then going from the transdiodes direct. 10.000uF is a bit extreme but it should have at least 2200uF. The sound benefits much from a big cap. The transistors chosen work extremely well in this circuit although with 30 Ohm Rbee they do not have the lowest value of base spread resistance in the business. Nevertheless the circuit is dead quiet. People that build it first thought that the circuit did not work so quiet it is.
Joachim
did you used that schematic http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-11.html#post2703178
did you used that schematic http://www.diyaudio.com/forums/digi...minimalistic-iv-converter-11.html#post2703178
Joachim,
The ES9018 has a low output impedance of about 800R.
So let's go as far as we can and aim to do Zin < 1R.
Since it is CEN, the PNP & NPN are in parallel for the signal.
That is why I chose E153 as current source (15mA bias), so that Zin is about 0R8.
This has the same function as your BF245 with trimmer.
E153 are widely used in Japan for powering LEDs.
I am sure Davide would be able to get them if necessary.
I was hoping to get zero DC offset by perfect matching.
Anyhow, the schematics demonstrates how it can fulfill all the functional requirements.
Patrick
The ES9018 has a low output impedance of about 800R.
So let's go as far as we can and aim to do Zin < 1R.
Since it is CEN, the PNP & NPN are in parallel for the signal.
That is why I chose E153 as current source (15mA bias), so that Zin is about 0R8.
This has the same function as your BF245 with trimmer.
E153 are widely used in Japan for powering LEDs.
I am sure Davide would be able to get them if necessary.
I was hoping to get zero DC offset by perfect matching.
Anyhow, the schematics demonstrates how it can fulfill all the functional requirements.
Patrick
EUVL, the is always a small difference in Ube in PNP and NPN bipolars that give an offset in a parallel symmetric BJT transimpedance input. Unfortunately Ube is a week function of idle current so you can not trim it out by asymmetric bias. You could of cause select a pair of transistors that have the same Ube at the required idle current. Good luck. Actually i do not think that this ( small ) DC offset is a problem but on my thread the consensus was that even a small DC offset is not allowed, that in the context of amplifying an MC phono cartridge. I solved that problem with the local loop but curiously the same people that bugged me did not react. No response, nada. That happens on the web quite often. When the readers spot a "problem" it is beaten to death and when the solution comes than it is "obvious". That does not frustrate me in any way, quite the contrary. Without pushing me i get lazy. I will do some calculations how much input impedance you get with how much idle but your estimate sounds about right to me. As i said, low input impedance is the forte of a BJT common base stage with high idle.
Were they not getting scarce as proverbial hen's teeth, I'd recommend the Toshiba low rbb' parts, the 2SA1316 and 2SC3329. The blue grades claim 350 - 700 beta for both parts, which if one is lucky enough to get toward the high end of the range is pretty decent. My friend Ben G. bought bags of them some years ago and I don't believe ever used more than a few.
He did report being disappointed that they still had appreciable current noise, but digging a bit deeper it seemed to be well in line with expectations based on the operating point and the typical behavior of bipolar base current. They are not JFETs!
This reminds me of an anecdote, but I've muddied this thread enough already
He did report being disappointed that they still had appreciable current noise, but digging a bit deeper it seemed to be well in line with expectations based on the operating point and the typical behavior of bipolar base current. They are not JFETs!
This reminds me of an anecdote, but I've muddied this thread enough already
I also do not think the small offset is an issue.
An easy way to solve it is by trimming the resistor network that provides the Vref offset at the base of the input biasing pair.
ES9018 needs an offset of +1.65V (or 1/2 of +Vs analog).
So one can take the voltage off the regulator, divide by 2 with a potential divider, and trim one of them till the input voltage is zero at ideal.
I know it is better to solve the problem where is arises.
But solving the problem without adding 5 resistors is also quite attractive.
Don't you think ?
Patrick
An easy way to solve it is by trimming the resistor network that provides the Vref offset at the base of the input biasing pair.
ES9018 needs an offset of +1.65V (or 1/2 of +Vs analog).
So one can take the voltage off the regulator, divide by 2 with a potential divider, and trim one of them till the input voltage is zero at ideal.
I know it is better to solve the problem where is arises.
But solving the problem without adding 5 resistors is also quite attractive.
Don't you think ?
Patrick
I was re-reading one of Linsley-Hood's books the other day, and he makes the controversial statement that P and N JFETs are much better complements than PNPs and NPNs. Well, I guess it all depends on what parameters one is concerned about. The mismatch in Vbe is certainly a nuisance for bipolars in many configurations. But the drastic difference in mobility of holes and electrons in majority-carrier devices leads to huge differences in the capacitances associated with comparable-transconductance JFETs. I'm still very glad we have both polarities available, but they are not very complementary.