Salas DCG3 preamp (line & headphone)

diyAudio Chief Moderator
Joined 2002
Paid Member
Hi All, I am really grateful to Salas for including me as a beta tester for DCG3 project.

I started sometime early June when I received the schematic. With a lot of help (and patience) from Salas, I have managed to put together my working DCG3 prototype and has been tweaking/listening to it the last month.

Let me state upfront that this is my most eagerly anticipated DIY project. I set out to build it double mono, power it with V1.2 main rail and with LDR volumn control, the result is a great sounding and most satisfying preamp in my system.

Let me attempt to describe how it sounds (subjectively in the context of my all DIY system); compared to my last preamp which is based on a famous buffer stage/LDR attenuation/V1.2 psu, it sounded faster, equal to if not more transparent, has better resolution, more detailed, better tone and image density, and exhibit tight control when the music gets complicated.

I strongly recommend everyone here to try this out in your system.

Tham hi, you are welcome. Beta testing feedback and implementation in self styled construction is always the proof of the pudding.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Thanks Salas. So the main objective is to increase the bandwidth???

We tried to reduce Vgd such that the effect of Cgd in limiting the bandwidth is also minimum? Or, you aim for some kind of a "sweetspot" Vgd or Vds for that particular hexfet?

M1 Vgd does not move at the real voltage gain rate of the duo so its Cgd does not appear multiplied. Because M2 shields M1 from Miller effect. The weak JFET needs to directly drive the big MOSFET's capacitance. Without help of another stage in this case so to keep things as straightforward as possible. By removing Miller the driver induced THD is lowered. Bandwidth is very good and we even spared some in the end but mainly we wanted to decrease THD.

Time for a circuit description:

The volume pot adds impedance to the RC input low pass filter (R1,C1). But when set at max volume the pot becomes just a contact and the circuit must remain stable. So R1's value is chosen to form an effective time constant with C1 for stability even in direct input conditions. The system is also closed loop stability compensated (C2 across feedback network's R6). With the volume pot at max and a 50R signal source impedance the slew rate of this circuit can be >25V/uSec. The exact figure depends on internal routing and bench test loop specifics too.

J1 & J2 form a long tail input pair actively loaded by Q1 & Q2 mirror. VR1 is used to null the DC offset before U1 DC servo is put in place. So to work very little, against thermal drifts only, not against a residual mismatch. J3 is a low capacitance RF type JFET used as a constant current source biasing the LTP.

The signal path directly continues from J1's drain to M1's gate. M2 cascodes M1. R9 R8 voltage divider biases M2 for half the positive rail across it. C3 filters at the divider's junction to keep gate DC voltage bias clean. Low capacitance is key so to can directly drive the output stage from just a small JFET with little distortion as already mentioned. Its a delicate situation here because there is no intermediate stage to boost drive current. The semiconductors hands on choices mattered for synergy.
With Fairchild's fastest TO-220 MOSFETS the circuit's THD wasn't as good as with Vishay IRF9610s for instance. Various JFET types changed the THD figure and its harmonic trend also. Higher rails voltage keeps on making the MOSFETS less capacitive up to 25V or so. Resulting to even less THD. This output stage is a voltage gain stage too. Due to dissipation reasons for all the semiconductors but especially for the high bias output ones, and not to add special regulators for U1 and the relay, a sweet spot was chosen.

M3 is a fast depletion mode (self biasing) MOSFET. Makes a constant current source for the output stage. Having CCS both on input and output stages not only fixes this circuit's total bias current against drifting better but also increases its PSU ripple rejection ability.
R10 sets M3's current draw. Bias >100mA can be chosen with M3's IDSS being the limit (listed at 150mA minimum). This mA region is also the M1 M2 zero tempco region. Bias stability tested solid indeed with moderate sinking. R6 R7 form a low impedance feedback network to the LTP's return input (J2). Low impedance is necessary not only for low noise but to also keep this particular circuit stable. In most cases the burden of driving that network is heavier than what the preamp will be called for to drive on its line output.
By the way, with the more capacitance of output interconnects this preamp sees, the more its phase margin. I.e. no worries for long or parallel output cables. Tested stable with just a 10nF capacitor as an output load even. The bandwidth would be seriously curtailed with such an extreme parasitic capacitance example of course. Coaxial signal cable has about 100pF per meter on average to keep things in perspective.

U1 op-amp senses the circuit's output around DC through the R12 C4 low pass filter. Then integrates time via C5 and drives counter-DC corrections through R14. Working as a DC servo at about 2.3HZ corner frequency proved responsive in this circuit. Many U1 types were tested to pick a BOM favorite. JFET input technology chips in general are most adequate for this job here because of the low pass filter's very high impedance they see.

The output from M2's drain is interfaced with a lab standard 50R to line output. Another route goes to headphone output. Zero additional Rz with using just a jumper is the technically neutral thing. Alternatively a judicious Rz value can be used to under-damp and loosen up some headphones which their users may describe as "dry". Caution though, anything higher than 10% of the nominal headphones impedance for an Rz resistor will interfere with their overall response more than +/- 1dB. Unless someone opts to intentionally color them even further.

Because no capacitors are in the signal path a relay system timed to 5 sec delay by Ra Ca lets no DC settling bump during power on to get out. It also disengages quickly during power off. Dc Db Da LEDS drop the rail voltage to 12V relay safe working range territory. And you know its on. One of them can be actually brought to the front panel as "ready" sign because they light up only when the relay clicks.

Adding the schematic again for ease of reference.
 

Attachments

  • DCG-3_Sch.png
    DCG-3_Sch.png
    26.7 KB · Views: 5,701
Count me in for the DCG3! I have your Simple Folded Phonopreamp and love it!

Hi all, this thread is about a JFET & MOSFET preamp I am cooking for sometime now. Its two stage single ended Class A circuit.
It was developed on the bench involving much experimentation with topology and a range of active parts. Simulations came second.

Some speakers are insensitive, some amps also, or a source is relatively low output like a phono stage. So a line preamp with gain is good to have.

Its named DCG3 because it has DC servo and three MOSFETS. Its standard gain setting is also times three. :D

The main schematic is attached

Some THD plots when driving 1.5kOhm line input and various headphones are also attached.
 
Member
Joined 2006
Paid Member
First round eval.

Let me attempt to describe how it sounds (subjectively in the context of my all DIY system); compared to my last preamp which is based on a famous buffer stage/LDR attenuation/V1.2 psu, it sounded faster, equal to if not more transparent, has better resolution, more detailed, better tone and image density, and exhibit tight control when the music gets complicated.


I am using the DCSTB power supply shown above, and a ALPS pot on a In-Select board. Although I want more time for evaluation in my best setup (I listened for about 3 hours this weekend). I can concur with the the 'apparent speed' aspect of it. I also noted that the precision of the image in the sound-field, at least from a left to right perspective is very tight. To get that tightness on a DCB1 for instance, in my system, it has to high biased on the hot-rod side. I am still studying image depth, and tonal qualities.
What I can say is that well done recordings, such as those from Mobile Fidelity Sound Labs, sound extremely transparent and detailed. It does not sound fatiguing, but it does sound intense. You then compare to say default web-browser HIFI Tidal running through USB, and it's not the same. Same with poor vinyl, it just shows the weaknesses or strength of the source recording. I have not done a lot of listening via the headphones at this point, as it just not my preferred listening style. But I do have another headamp to try that I hope to compare to later.
I try not to make attempt to oversell any boards that Salas and I may produce. But I will give my honest evaluation of it. And I need more time with it, more recordings. And more spoken voice, either through running movies through it, or more slowly sung material. To me this is the better test . A voice that you know well. A lot of gear will add a 'signature' sound to this. Really good gear should not add too much.
 
Just a bit of my listening experience with regards to imaging capability of DCG3...

I put on Amused To Death CD yesterday and the effect is akin to listening through a huge pair of headphone. A totally enveloping sound field; very good sense of depth, well separated images front and back and even above and behind the listening position, and all those (surround) sound effects contained in the recording are well spread and mostly discernible.

It is one of the most difficult test I have put the DCG3 through so far and I must say that this CD has never sounded better before.
 
Just a bit of my listening experience with regards to imaging capability of DCG3...

I put on Amused To Death CD yesterday and the effect is akin to listening through a huge pair of headphone. A totally enveloping sound field; very good sense of depth, well separated images front and back and even above and behind the listening position, and all those (surround) sound effects contained in the recording are well spread and mostly discernible.

It is one of the most difficult test I have put the DCG3 through so far and I must say that this CD has never sounded better before.
If there is such a big difference between then and now with a good intermediate stage between Source and amp, then that points me to conclude there is something lacking in the drive between your Source and your Power Amplifier.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
It could do the K1000's max with 20V rails, 150mA output stage bias, and gain 5.5 from a 2V RMS source (2.25K R6)

That would bring 6W standing dissipation per channel. Better be built in double mono for separate sinks with 15V reg chips for the op-amps and the relays also.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
And for Sennheiser H600 300 ohms?

No problem. HD 600 states 97dB/mW which is about 102dB/V for 300 Ohm. With 2.45VRMS and 20mW it will give 110dB. Don't listen so loud sustained to other than peaks in Rock concerts. I will sometime soon have HD 600s in my hands to tell you more about the quality of the combination too. K1000s are 74dB/mW and would do 100dB at 400mW (6.9V RMS at 120 Ohm) and only 104dB at its 1W max. So that's a non much useful destruction max. That inefficient headphones are simply quiet headphones and pushing them for bit more increases their distortion and probability of damage.