MPP

This is a DC coupled lower noise version with input offset trim. It has several advantages.
Gm is double, noise is around 0.5nV/qHz so 3dB better, input DC offset can be lower. I use a simple Widlar mirror that can be substituted with more advanced schemes like the cascoded mirror or other more sofisticated mirrors i have shown here before. It has to be seen how good DC stability is but there are solutions to trim output offset or servo. An alternative whould be a folded cascode with or without current source. The current sourced folded cascode version could be configured as an All In One Go phonostage with the help of an output buffer.
 

Attachments

  • Simple BJT Cascode Pre-Pre Low Noise DC.TSC - TINA.pdf
    57 KB · Views: 221
The Simple Cascode Pre-Pre is playing. I am very pleased with the result considering the low complexity. Noise is decent low. At typical volumes i can not hear it at the listening place. I use the MM input of the JG-Self preamp. Poweramps are modified Lorycraft Zenith 601 monoblocks. I use them more often now then my LA Audio Tube Amps. SN is better, so is resolution of details. More power is on tab so it sounds more dynamic too. The Pre-Pre itself has a slightly mellow character that reminds me certainly of the Essex equalizer. Still there is a lot of transparency and space. I also have a discrete second stage on my bench so a fully discrete simple phonostage will be made.
 

Attachments

  • Simple Cascode Pre-Pre.jpg
    Simple Cascode Pre-Pre.jpg
    79.4 KB · Views: 470
Here is the basic draft of a fully discrete BJT-MOS phonostage. The 75usec is done transimpedance in the first stage and the rest is done with shunt feedback in the second stage. This stage is phase inverting so is my Baxandall line stage.
You may ask now why on earth does Joachim do another phonostage ? Part of it i have already answered. This new design uses only components that are abundant and cheep. There is a plethora of bipolars that can be used. Very low Rbe`devices are rare too but a lot of medium power transistors come close. There is still the 2N4401/4403. I can even remember that we made good sounding MC stages in the 80th with selected BC140/160. This stage i build with 2SC1815 / 2SA1015 that have an Rbe`of 50 Ohm. Using the Faulkner paralleling 2 x 4 BC550/560 do fine. Some people do not like constant current sources or current mirrors so i do not use them. Some even do not like cascoding so the input stage can be simplified even more by taking the cascode away at the price of less speed, more distortion and worce PSRR. There are even those that hate emitter degeneration so i do not use it in the secound stage and even the first stage could do without but i did not try this. The only thing that is not negotiable is the parallel symmetry out of the reasons i argued about. Feel free to do your own single ended version if you like more second harmonic, more caps in the signal chain and more noise. You can look at it also the other way. I already published a more complicated version of the input stage with less noise and current mirrors for potential DC coupling. Curent sources can be added to the second stage. So does cascodig of the MOSFETs and a servo.
 

Attachments

  • Simple Discrete BJT- MOS Phonostage.TSC - TINA.pdf
    66.2 KB · Views: 167
Hallo Joachim !

Thank you very much for sharing your design research with biplorar transistors.

After receiving 500 pcs sheety 2SJ74BL from a seller in China, I was loud swearing, like a trooper. Maybe it was so loud, that it comes to your ears in Brilon, and you started with bipolar design.:)

Your decision to use standard low noise parts is fine! Low Rbb- Transistor price is rising in the same way like p-channel jfets.

Is it possible to change your fantastic sound Hawksford-Buffer to bipolar design?

Best Regards

Sam
 
Yes, a Bipolar Hawksford-Buffer is posible. The input stage of the Simple BJT-Cascode Pre-Pre could also be configured with the Hawksford connection by connecting the Leds to the emitters and not to earth. The resistors that drive the Leds have then to be substituted with contant current sources or else the input transistors have to drive the low impedance resistors so distortion will suffer. Also the resistors that do the bias of the input stage could be substituted with constant current sources of preferably low noise. You see, as always this is only the beginning of a phantastic journey i hope. I am building the second stage right now. I got good friends with MOSFEts recently. I really start to like them and they are abundant too if not so much laterals. Yee, i know, the really good stuff is rare and expensive but there are some workarounds.
 
The simple BJT-Cascode-MOS phonostage is up and running. First i build the second stage with plus-minus 12V supply to test it with batteries. Later i will do a plus-minus 24V version. The resistors that set bias have to be scaled acordingly then. The MOSFets idle on around 45mA. The circuit needs some time to start up and the idle current is growing over the next few minuts until it reaches equilibrium. The plus-minus 24V version will need cooling of the MOSFets. Performance is very good despite the simplicity. Frequency response is 900kHz -3dB with a gain of 33dB at 1kHz. The circuit is stable without miller caps i think because i used gate stoppers. Then i added the input stage. I found another solution for the 75usec part. It is now a resistor in series and a cap to ground building a passive eqalizer. That saves 1 cap compared to the first solution. I also was afraid that noise could come into the circuit when i reference the caps against the PSU. I use a 100nF PHE450 from Rifa. It is a 5% component so i use a trimmer for HF adjustment. With the values shown i got plus-minus 0.15dB from 400Hz to 20kHz. The bass at 20Hz is a bit up by 1.5dB so i have to trim the shunt feedback components or enjoy a rather full sound. Gain is 66dB total so this is a bit too much for my Titan i. To adjust that i could raise the 2kHz series resistor but that gives more noise. The other option whould be to half the feedback resistors and double the feedback cap. I think i go that way. I tryed to put the output cap into the feedback loop but that did not work. The input stage needs some bias to flow. I then had the idea to split the 1MOhm resistor in two 2MOhms. One connected behind the cap. If that works i could also put the 91kOhm resistor and the 3.3nf cap behind the output cap. This project went smooth although i did not even put it on a groundplane PCB because i had rum out of PCBs.
 

Attachments

  • Simple BJT-Cascode-MOS Phonostage.jpg
    Simple BJT-Cascode-MOS Phonostage.jpg
    80.6 KB · Views: 398
  • Simple BJT-Cascode-MOS phonostage squarewave.jpg
    Simple BJT-Cascode-MOS phonostage squarewave.jpg
    74.4 KB · Views: 403
  • Simple Discrete BJT- MOS Phonostage 12V.TSC - TINA.pdf
    73.1 KB · Views: 125
Thinking about a BJT Hawksford buffer i did not come up with a DC coupled version so far.
I have being building Diomond buffers of considerable sofistication but the sound was never as elegant or transparent then the JFet Hawksford. I had succes with JFet-BJT buffers and i have published a Szikley design here that sounds rather fine. High Pinsh Off Voltage JFets are needed and they are less expensive and more abundant then the 2SK170/2SJ74 pair so i should be happy. Nevetheless i was looking for a circuit that uses only cheep BJTs and i think i found something i should try. It is a parallel symmetric variant of the current mirrored Schlotzaur buffer. D.Self has shown a single ended version on page 80 of his book "Small Signal Audio Design". The origins of the original circuit are somewhat obscure but i know that another frenshman that calls himself Forr here claims to have invented the current mirror trick. My contribution is only the parallel symmetry so the already very low distortion should go even down in second harmonic. Do you see what i see ? Yes it looks very similiar to the simple BJT-Mos phono so the current mirrors could enhance that circuit too. I also found another buffer i like. Looking for simplicity recently and starting to understand MOSFets i found the Millet AMB MOS Diamond buffer. I like that the signal only flows through one pair of transistors. It does not come much simpler then this when performance is on your mind.
 

Attachments

  • AMB Millet MOS Buffer.gif
    AMB Millet MOS Buffer.gif
    14.7 KB · Views: 408
  • Parallel Symmetric Current Mirror Schlotzaur Buffer.TSC .pdf
    39.5 KB · Views: 169
Here we go. I used 2mA green Leds. Red ones are theoretically less noisy so they should work too. The forward voltage with red Leds is more like 1.4V instead of 1.8V of the green ones so the resistors should be lowered a bit. The folmular is aproximately: Forward drop - Ube / I so 1.4V - 0.7V = 0.7V / 6mA = 116.66 Ohm but neither the Leds nor the idle current is very critical i think. The BJTs should be something like BC550/560 or the usual suspect Japanese one. Some think they sound better, others think that a German designer should use European transistors but that is more a matter of style. The original circuit uses BJTs with Hfe of 200 but i think this is also not critical.
 

Attachments

  • Parallel Symmetric Current Mirror Schlotzaur Buffer.TSC .pdf
    49.1 KB · Views: 200
Sam, be carefull. The insertion of the low value emitter resistors can bring the output trannies in trouble. It depends on the collector voltage of the driving transistor. If that is too high you will fry the output transistors. Start without the emitter resistors. That fixes
the bias voltage to Ube. You can then insert 50 Ohm emitter resistors and measure bias voltage again. If there is no low value emitter resistor posible without fying the output stage i could imagine to use MOSFets that have a much higher Ugs. The original circuit does not use emitter resistors. Is that a Class B ?
 
I did not simulate or measure it but having two constant current sources in series it should. There are of cause a lot of other ways to make a current source and the simple BJT-Led variant should be more then good enough.Output impedance of that simpler circuit is already more then 500kOhm. I will build this buffer first simply with resistors as current sources. The current mirrors act as current sources too anyway.