OCD log – Part II
Part I – DC servo
Existing JFET input buffers in various amplifiers, that use transformer as VAS, are very good. But, more capable buffer brings some benefits as shown here: LuDEF
Deciding should design of new buffer be based on discrete components or integrated circuits (What? ‘Evil’ ICs instead of ‘magical’ JFETs?) I asked myself: what is inside DAC box, that is considered the source of reference sound magic? Right.
Many DACs contain worse output components than used in this design. For input is used one of the best operational amplifiers in existence, with true JFET inputs. One mandatory checkbox ticked. Transformer driving force is provided by specialized buffer IC that can deliver 170 mA rms with 2000 V/us slew rate and 0.00003 % distortion, at best conditions.
Buffer is supplied by on-board super regulators that provide high PSRR, excellent transient response and low noise from any +-18 to +- 60 V source. It is unimportant how good or bad is amplifier’s power supply. Buffers will always have optimal power supply.
PCB has 4 layers with 80 x 80 mm dimensions. Front and back copper areas are used as heatsinks and one layer is used as Zero Signal Plane, increasing immunity to EMI. Whole PCB is mounted in place using RCA socket. Input socket can be placed in two ways and PCB can be rotated in desired position. I find this format convenient for quick various front end/buffer interchange.
Possible benefits from this buffer are:
Extended frequency response – very low output impedance and high current capability can better drive amplifier output MOSFETs input capacitances. In example, LuDEF frequency response is – 3 dB at 38.500 Hz. With this buffer, it is extended to 87 kHz.
Less distortion – standard JFET buffer contributes with measurable distortion, more at high frequencies. Distortion by this buffer is too low to be reliably measured by hobby equipment. Looking at the following measurement, I’m confident that it achieves datasheet declared performance of 0.00003 %. Buffer frequency response is DC to 2.5 MHz at gain 2.5 X and to 3 – 4 MHz at less gain. You can forget on intermodulation distortion and transient intermodulation distortion with such high bandwidth. RF and EMI incursion is limited with low pass filter at input (R1 – C5).
This distortion measurement was made at Focusrite output voltage that provides lowest self -introduced distortion.
100 kHz square wave response, made without input RC filter, is perfect.
Increased input sensitivity and total amplifier gain – required input voltage for full output power is 3.7 Vrms in case of LuDEF or SissySIT. Buffer gain can be easily changed from 0 dB to 12 dB or from 1 to 4 x enabling to drive amplifier directly from DAC with volume control or from Korg B1 preamplifier which has 1 % distortion at 1 V output, but 6% at 4 V output.
Better channel separation – as every buffer channel has its own PS, left and right channels are perfectly isolated from any influence to sensitive VAS stage by single power supply providing power to both amplifier output boards.
Buffer DC offset at output is less than 100 uV.
How does it sound? Absolutely transparent and detailed with perfect soundstage.
As reference was used (to my ears/brain deeply burned-in sound of) Salas DCG3 preamplifier with dual mono shunt regulators, Toshiba 2SK170 JFETs at input and Muses 72323 volume control, directly driving Cinemag input transformers. Sound provided by directly driving VAS transformer like that is considerably better than with original JFET buffer.
Introducing in between, at amplifier input, new buffer didn’t change that exquisite sound in any perceivable or measurable way! Enough for me.
If anyone is FAB enough to consider building this, I’ll provide gerbers and documentation.
EDIT 30.01.2022. - project documentation added (complete as one zip file and separate items)
EDIT 05.03.2022. - added gerber files with PCB R3 having mounting holes
Mouser shared basket
Part I – DC servo
Existing JFET input buffers in various amplifiers, that use transformer as VAS, are very good. But, more capable buffer brings some benefits as shown here: LuDEF
Deciding should design of new buffer be based on discrete components or integrated circuits (What? ‘Evil’ ICs instead of ‘magical’ JFETs?) I asked myself: what is inside DAC box, that is considered the source of reference sound magic? Right.
Many DACs contain worse output components than used in this design. For input is used one of the best operational amplifiers in existence, with true JFET inputs. One mandatory checkbox ticked. Transformer driving force is provided by specialized buffer IC that can deliver 170 mA rms with 2000 V/us slew rate and 0.00003 % distortion, at best conditions.
Buffer is supplied by on-board super regulators that provide high PSRR, excellent transient response and low noise from any +-18 to +- 60 V source. It is unimportant how good or bad is amplifier’s power supply. Buffers will always have optimal power supply.
PCB has 4 layers with 80 x 80 mm dimensions. Front and back copper areas are used as heatsinks and one layer is used as Zero Signal Plane, increasing immunity to EMI. Whole PCB is mounted in place using RCA socket. Input socket can be placed in two ways and PCB can be rotated in desired position. I find this format convenient for quick various front end/buffer interchange.
Possible benefits from this buffer are:
Extended frequency response – very low output impedance and high current capability can better drive amplifier output MOSFETs input capacitances. In example, LuDEF frequency response is – 3 dB at 38.500 Hz. With this buffer, it is extended to 87 kHz.
Less distortion – standard JFET buffer contributes with measurable distortion, more at high frequencies. Distortion by this buffer is too low to be reliably measured by hobby equipment. Looking at the following measurement, I’m confident that it achieves datasheet declared performance of 0.00003 %. Buffer frequency response is DC to 2.5 MHz at gain 2.5 X and to 3 – 4 MHz at less gain. You can forget on intermodulation distortion and transient intermodulation distortion with such high bandwidth. RF and EMI incursion is limited with low pass filter at input (R1 – C5).
This distortion measurement was made at Focusrite output voltage that provides lowest self -introduced distortion.
100 kHz square wave response, made without input RC filter, is perfect.
Increased input sensitivity and total amplifier gain – required input voltage for full output power is 3.7 Vrms in case of LuDEF or SissySIT. Buffer gain can be easily changed from 0 dB to 12 dB or from 1 to 4 x enabling to drive amplifier directly from DAC with volume control or from Korg B1 preamplifier which has 1 % distortion at 1 V output, but 6% at 4 V output.
Better channel separation – as every buffer channel has its own PS, left and right channels are perfectly isolated from any influence to sensitive VAS stage by single power supply providing power to both amplifier output boards.
Buffer DC offset at output is less than 100 uV.
How does it sound? Absolutely transparent and detailed with perfect soundstage.
As reference was used (to my ears/brain deeply burned-in sound of) Salas DCG3 preamplifier with dual mono shunt regulators, Toshiba 2SK170 JFETs at input and Muses 72323 volume control, directly driving Cinemag input transformers. Sound provided by directly driving VAS transformer like that is considerably better than with original JFET buffer.
Introducing in between, at amplifier input, new buffer didn’t change that exquisite sound in any perceivable or measurable way! Enough for me.
If anyone is FAB enough to consider building this, I’ll provide gerbers and documentation.
EDIT 30.01.2022. - project documentation added (complete as one zip file and separate items)
EDIT 05.03.2022. - added gerber files with PCB R3 having mounting holes
Mouser shared basket
Attachments
Last edited:
Moderator edit: This schematic is reported to contain an error. The correct one is attached to the first post.]
Yes, this is just my kind of game. I have LuDEF boards in route, and have already been convinced that quality audio opamps can sound wonderful. See my comments on the Marauder FE for the VFET lottery amp.
BTW, notice that I have used both the OPA552 and OPA445 opamps as the output drivers in the Marauder. The input buffer / inverters are the OPA2134. Power rails boosted by on-board DC-DC converter.
BTW, notice that I have used both the OPA552 and OPA445 opamps as the output drivers in the Marauder. The input buffer / inverters are the OPA2134. Power rails boosted by on-board DC-DC converter.
I’ve seen your comments and those by chede, voicing the same opinion. It is of notice, that Mark Johnson considers Norwood as best M2x FE (from the first batch). That were supporting points to go this route.
It can be used no problems, but there is already excellent FE of same type - Norwood.
Ah! Thanks for that tip.
My M2x is stil WIP, so I'll get to the Norwood variant eventually.
Thank you again for a new project to consider. Most of (OK ALL) of the aspects for design consideration are well above my head. However, both this and the "replacement R in CRC" projects fascinate me. The other project is delayed due to my sitting on my hands until parts were no longer easily available.
Given my enjoyment of the Norwood IPS for my M2x along with some early interest from TA (who's listening experiences sometimes mimic my own), this seems like a very interesting project. I have a LuDEF ready to go on the bench, and I've got another set of SissySIT FE boards of another type that I may compare with.
When I started into this hobby, I often wondered how people would not get around to finishing things and/or be delayed. Yeah, now I understand.
Just saying thanks and following. Won't add to the project list just yet.
Given my enjoyment of the Norwood IPS for my M2x along with some early interest from TA (who's listening experiences sometimes mimic my own), this seems like a very interesting project. I have a LuDEF ready to go on the bench, and I've got another set of SissySIT FE boards of another type that I may compare with.
When I started into this hobby, I often wondered how people would not get around to finishing things and/or be delayed. Yeah, now I understand.
Just saying thanks and following. Won't add to the project list just yet.
OT: MQB for audio
This trend in modularization reminds me of the VW MQB system.
While my WIP (M2x) will be an all-in-one, including the power supply, my future builds will seek to separate power to a separate unit.
Kind regards,
Drew
This trend in modularization reminds me of the VW MQB system.
While my WIP (M2x) will be an all-in-one, including the power supply, my future builds will seek to separate power to a separate unit.
Kind regards,
Drew
tombo56 nice project , I don't know if you read the last article by Walter Jung, it could be applied to your C3 and C4 in the regulators. An Improved Reference Filter for Audio Regulators
I also used OS-CON in that position until I made measurements, now I use UKZ , OS-CON is practically unusable there due to high leakage current. The Pan FC(25V) and Ryb ZLJ(35V) series also showed good results in my measurements. Further improvements can be made by replacing R1 and R2 with simple CCS, for this I use BSS159 or CPH3910 depending on the voltage.
I also used OS-CON in that position until I made measurements, now I use UKZ , OS-CON is practically unusable there due to high leakage current. The Pan FC(25V) and Ryb ZLJ(35V) series also showed good results in my measurements. Further improvements can be made by replacing R1 and R2 with simple CCS, for this I use BSS159 or CPH3910 depending on the voltage.
Hi grunf,
Everything written by Walt Jung is mandatory literature, so I did.
All that you suggest is in place and welcome advice.
However, Panasonic SEPC series organic polymer capacitors are far better in leakage department than datasheet specified value by two orders of magnitude. I was surprised. Anyway, they are on my boards because I had them in the drawers and they don’t limit noise or PSRR performance of this small super-regulator.
As you suggest, on another project and in same voltage reference filter function, Panasonic FM capacitor provided best output noise profile and PSRR was not influenced by it.
This buffer uses opamps that both have high PSRR on their own, so it was not required to make power supplies better in that regard. Primary design goal was perfect transient response and wide power bandwidth, which they provide both. Noise is very low as well.
Everything written by Walt Jung is mandatory literature, so I did.
All that you suggest is in place and welcome advice.
However, Panasonic SEPC series organic polymer capacitors are far better in leakage department than datasheet specified value by two orders of magnitude. I was surprised. Anyway, they are on my boards because I had them in the drawers and they don’t limit noise or PSRR performance of this small super-regulator.
As you suggest, on another project and in same voltage reference filter function, Panasonic FM capacitor provided best output noise profile and PSRR was not influenced by it.
This buffer uses opamps that both have high PSRR on their own, so it was not required to make power supplies better in that regard. Primary design goal was perfect transient response and wide power bandwidth, which they provide both. Noise is very low as well.
Last edited:
