Salas DCG3 preamp (line & headphone)

Cold, hard facts

100KHz into 33R load
Top trace = in
Bottom trace = out

Horiz scale = 2us/div
Vert scale(in) 1V/div
Vert scale(out) = 2V/div
 

Attachments

  • IMG_100KHz.JPG
    IMG_100KHz.JPG
    177 KB · Views: 312
  • IMG_100KHz_NoAmbient.JPG
    IMG_100KHz_NoAmbient.JPG
    171.5 KB · Views: 319
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
By the way if you feed 1Vpk-pk sinewave to your build (which has fixed gain of 3) and you up the generator's frequency until DCG3's output goes from 3Vpk-pk down to 2Vpk-pk you find its F3 upper limit. How many hundred kHz is that? Because passing such a good looking 100kHz squarewave surely takes a very high F3.
 
Hi Salas -

Since I am using the UBibs to power my DCG-3s, do you think I'd be better off swapping out C6 and C7 for smaller caps - perhaps higher quality such as Elna Silmic II RFS in 33uF or 47uF?

Also, have you found that increasing the bias current in either the UBibs or DCG3s helps SQ at all? My FETs are not getting overly warm (<40C) right now so I have some room.

Thanks,
Jay
 
diyAudio Chief Moderator
Joined 2002
Paid Member
C6 C7 100uF Panasonic FC seem to blend very well with DCG3 but they are not written in stone either. They are in local decoupling caps role and something should be there nonetheless. I have seen oscillations with some wiring schemes if without. Yes you can definitely try Silmic II RFS in their place. Although not advertised as low Z HF caps I have measured very low 100kHz ESR in them as well. 47uF should be sufficient. Pana FC are still longer lasting when located in high temperature spots.

UBiBs are not as much benefited by high spare current as the previous generation BiB shunts. 100-150mA headroom is good enough. You may increase their CC up to 45C on their sinks I guess. In DCG3 itself you may see in post#1 that 150mA bias has measurably better THD than 100mA on 32 Ohm headphones but did not differentiate for higher than 60 Ohm impedance headphones or for line level use. Couple of builders thought 150mA bias was better subjectively for line level too but others didn't seem to notice.

Remember, you have a differential doubled up build with four times the mW output power because any load you use it for is bridged between two voltage amps for XLR phases. P=V2/R. And you also put individual PSUs for everything. It has more than a regular DCG3's juice reserve in all accounts. Its like a two-man saw.
 
Thanks for the reply. I'm going to give my preamp a few more weeks to settle in before I decide whether to try making any more changes.

I had read elsewhere that shunt regulators (and particularly your design, but perhaps referring to the older version) tend to work better without as much decoupling capacitance since they are so fast to respond and some of that speed is lost with larger decoupling caps.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
That's true in general. Shunts can also absorb kickback energy that series regs can't. In UBiB there is a specific ESR 33uF C3 output capacitor needed for stability. Because of much higher open loop gain than in previous gen BiB. You may safely bypass UBiB's C2 Vref filter cap with 0.1uF MKP underneath for a change in tone though. And see how you like it.
How much capacitance you use for local decoupling down the road on the client circuit its not only what minimum it needs for itself. But also a matter of what type of PSU is combined with the wiring length and dressing between. Using headphones or not and their impedance may play a role too in what value for C6 C7. Because they are also energy reservoirs near to the output Mosfets. I haven't tried too small value C6 C7 along with shunt regs but something minimum must be there for peace of mind against rails HF instability.
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Interestingly my original simulation plot's indication and your fresh build's F3 response measurement result are close within experimental error. Therefore from prototype and theory to current parts batches and boards, all stick to spec very well after five years. Satisfactory consistency.

*J1,J2=LSK389C Cordell Spice model, signal source Z=50R, output load=33R (parallel to 100pF for 1 meter coax)
 

Attachments

  • 2021-11-25_191019.png
    2021-11-25_191019.png
    23.7 KB · Views: 350
Gain setting

Dear Salas

I also would like to pair this pre amp with a First Watt F6. As in post #4845 mentioned it appears that a higher gain setting may be favourable.
Any recommendation for R6 to set gain with a TDK 2CP-2511 pot; 1K will be just fine with most use cases?
For comparison in my system if I connect a Logitech SBT directly to the F6 I can get decent volume levels. Not sure how much gain the SBT delivers.

Thanks!
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Hi

With 1k R6 and a TKD the unity gain state should be met when turned up to about one o'clock. There it passes the SBT's output equal when you use it as a source in the preamp's input. Larger preamp output signal starts from that point on until a three times stronger voltage level than SBT's own is reached at max volume knob setting. That is what a log pot in a standard gain setting (3x=~10dB) DCG3 will do.

Do you think this is alright for you? If yes, some higher gain setting will only be needed if you plan to use significantly weaker sources than the SBT. Like a phono stage for instance. In any case you can boost the standard setting situation by a further 6dB when replacing R6 with a 2.5k resistor. In other words you set for 6x gain now. Half way there (+3dB) would be to alternatively use 1.8k R6.

Can be good with F6 and medium sensitivity speakers but loud on usual sensitivity headphones when with strong sources though. Shortening the volume knob's comfort zone travel for most headsets.
 
Do you also know how to find V/uS slew rate?

Actually, I have not done this on any build. I've always assumed that slew rate was good enough.:sleep:

But now I am curious. I would guess that a 1MHz square wave (1us/division) would be very convenient, at say 5V or maybe 10V P-P? Or should this be measured at some frequency within the audio spectrum, say 10KHz and then scaled?

I'll bet there is a better way. What would you recommend?
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Large signal domain is using 10Vpp square. Small signal 100mVpp square. Audio test traditional frequency is 1kHz. Octave wise near the center. 1uS horizontal scale is helpful for counting divisions about slew rate. To watch the rising slope of a square closely will take fast divisions anyway.