Salas DCG3 preamp (line & headphone)

diyAudio Chief Moderator
Joined 2002
Paid Member
Listened to Peter Gabriel's Scratch My Back version of Heroes over the DCG-3 the other night.
It's a slowly sung song, seems close mic and it ramps up to a kind of chilling end with strings if you let it.
The lower mids of his voice never sounded so strong. The presence of him in the room was both strong, almost uninvited after a certain point. You may get this emotional reaction when you realize the sound can really have a visceral affect on you like that. This is one of the reasons I like audio - to get this experience from music. Sometimes you can only get that when the equipment is crisp and spot on. Like the low-noise of the folded, this pre allows you to squeeze a little more information out of your gear, and push it at you. It's not even a spectacular recording, but it's well delivered.

I felt like sharing this part of it. That the circuit may kick your butt if you let it.

I am very interested in what VGeorge and Salas have to say with their headphone shoot out.
Hey Mike, George visited yesterday afternoon. He will post at a point I believe but he got tired from the round trip I guess. Me I changed some components from generic to better grade today but not too fancy. Some signal path related dirt cheap generics to RN60 Dales (not even all of them - enough desoldering for one evening) and I also changed C5 MKS2 to MKP2 (it fits alright). Can hear little better smoothness, info, and space. Not a night and day difference but at this level worthwhile. It took an hour of warm up to settle because I changed the random local lytics to fresh Pana FCs too. I am listening to DSD "Any Major Dude Will Tell You" (Steely Dan - Pretzel Logic) in DoP right now. Reminds me of my old vinyl record enough. Here's a photo.
 

Attachments

  • DSDonCAN.jpg
    DSDonCAN.jpg
    840.2 KB · Views: 2,151
As teabag said, when over to salas two days ago and we had a headphone testing with his new headphone - line amp.
Dacs we used were his old and trusty 0404 EMU and his new Aune s16.
Headphones were my Sony MDRV-6 and AKG712 pro and salas AKG612, Grado SR60 and Sennheiser HD600.
We started with the EMU to get used to the sound of the headphones and continued with the Aune, which proved to be more revealing and better dac overall. I think that the EMU showed his age.
I also listened to the 712 and the HD600 direct from the Aune, as it has a dedicated headphone output. I preferred the Sennheiser, but it was obvious that the Aune could not properly drive the AKGs.
With the DCG3 now, there was a different game! It should be mentioned that it had no problem to drive any of these cans properly and at high level.
All of them sounded better through the DCG3. Everything was much more transparent and more musical.
The Sony and the Grado sounded very nice but the Sony were a little bit hot on the treble. Decent resolution for old designs.
The 612 and the HD600 were very similar in character, very neutral but I think that the 712 got the edge with a bit more musicality.
The DCG3 was powered by salas cap multiplier and used generic cables and components and nothing seems missing.It is really good! Only wonder could happen with dual BIBs and premium components.......
 

Attachments

  • IMG_20160914_202535[1].jpg
    IMG_20160914_202535[1].jpg
    632.9 KB · Views: 2,019
diyAudio Chief Moderator
Joined 2002
Paid Member
but it was obvious that the Aune could not properly drive the AKGs.
The Aune is proud of a very potent integral headphone amp. The weird sound thing when driving the 62 Ohm AKG712 PRO leads me to believe it also has high output impedance though. Maybe 30 Ohm and only properly compatible to 300 Ohm plus models because they always picture it with Sennheisers. Given it drives the HD600 alright it should always give better result with those at least than with sending the signal out to another amp through long wires and a pot. Well, not always :D

aune-s16-4.jpg
 
diyAudio Chief Moderator
Joined 2002
Paid Member
The 612 and the HD600 were very similar in character, very neutral but I think that the 712 got the edge with a bit more musicality.

The 612 is indeed great value for money and neutral in tone. Being much cheaper is bit coarser in the treble and bit opaque in the mids vs the two more expensive ones. The HD600 gives the impression of better clarity like having a generally faster resonance decay spectrum. The 712 is noticeably cleaner and softer than the 612 covering distance but still more lush than the 600. Both AKG are self adjusting and feeling like wearing a cloud when the Sennheiser clamps me and after a few hours it gives me a mild headache. If I had to have only one model for enjoying sweet but accurate enough music in comfort I would get the 712. The HD600 is more like a test instrument.
 
Listened to Peter Gabriel's Scratch My Back version of Heroes over the DCG-3 the other night.
It's a slowly sung song, seems close mic and it ramps up to a kind of chilling end with strings if you let it.
The lower mids of his voice never sounded so strong. The presence of him in the room was both strong, almost uninvited after a certain point. You may get this emotional reaction when you realize the sound can really have a visceral affect on you like that. This is one of the reasons I like audio - to get this experience from music. Sometimes you can only get that when the equipment is crisp and spot on. Like the low-noise of the folded, this pre allows you to squeeze a little more information out of your gear, and push it at you. It's not even a spectacular recording, but it's well delivered.

I felt like sharing this part of it. That the circuit may kick your butt if you let it.

I am very interested in what VGeorge and Salas have to say with their headphone shoot out.

Interesting!!! I ordered a couple dozen jets for this project already!! Never hurts to have extra!!
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Interesting!!! I ordered a couple dozen jets for this project already!! Never hurts to have extra!!
Any >3mA idss sample and preferably of no more than 2mA idss difference to use between channels will do fine. The LTP's tail CCS forces the units to bias at about 2.5mA Id per LTP side. There the Yfs is the same for any >idss unit because fixed lower on the curve. So no need to have many for producing careful matches.
 

Attachments

  • uPAyfs.png
    uPAyfs.png
    30 KB · Views: 1,702
Another Salas' interesting design! Thank you for sharing, it definitely deserves a try.

I have a few questions:

1. uPA68H (old, fat) or uPA68HA (new, slim)? You had the chance to test both, any prefrence? Preferred Idss grade should be K or L in this application, shoundn't it?

2. Series or shunt O/P relay: you stated "I avoided (shunt) because in case of strong transient the robust output stage is capable of passing strong current to the ground through the relay contacts." I prefer shunt. If the tiny contacts (1A) of the relay concern you, why not use them in parallel?
Or, maybe better, use two shunting contacts after R15/RZ. Yes, it costs few € more, but using two relays (and associated components) allows to equally load the PS, especially if dual mono. The same applies to U1 when using a single op-amp (see below).

3. DC Servo: a friend of mine (EE) claims that lowering the frequency (he goes down to 0.1Hz! :eek:) it becomes "less audible" (better phase response). Did you try going lower than 2.3Hz?
I read you tested several dual op-amp but the "classic" LF411 very often seen in many designs, including JC's. It's described in the datasheet as "low offset, low drift high speed integrator", it's single and much cheaper. Apparently perfect.
No doubt the AD823 suggested is a good choice from a reputed manufacturer, anyway. BTW, does it exist as "single op-amp" with a different designation?

4. Power supply: you proposed a dedicated series PS. No shunt regs this time (i.e. SSLV1.1)? :( Does the DCSTB sounds better with this circuit? If I would like to use the SSLV1.1 should I leave out C6/C7 or fit a couple of low value resistors before these caps?

Naste kalà! :wave2:
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Another Salas' interesting design! Thank you for sharing, it definitely deserves a try.

I have a few questions:

1. uPA68H (old, fat) or uPA68HA (new, slim)? You had the chance to test both, any prefrence? Preferred Idss grade should be K or L in this application, shoundn't it?

2. Series or shunt O/P relay: you stated "I avoided (shunt) because in case of strong transient the robust output stage is capable of passing strong current to the ground through the relay contacts." I prefer shunt. If the tiny contacts (1A) of the relay concern you, why not use them in parallel?
Or, maybe better, use two shunting contacts after R15/RZ. Yes, it costs few € more, but using two relays (and associated components) allows to equally load the PS, especially if dual mono. The same applies to U1 when using a single op-amp (see below).

3. DC Servo: a friend of mine (EE) claims that lowering the frequency (he goes down to 0.1Hz! :eek:) it becomes "less audible" (better phase response). Did you try going lower than 2.3Hz?
I read you tested several dual op-amp but the "classic" LF411 very often seen in many designs, including JC's. It's described in the datasheet as "low offset, low drift high speed integrator", it's single and much cheaper. Apparently perfect.
No doubt the AD823 suggested is a good choice from a reputed manufacturer, anyway. BTW, does it exist as "single op-amp" with a different designation?

4. Power supply: you proposed a dedicated series PS. No shunt regs this time (i.e. SSLV1.1)? :( Does the DCSTB sounds better with this circuit? If I would like to use the SSLV1.1 should I leave out C6/C7 or fit a couple of low value resistors before these caps?

Naste kalà! :wave2:
@ massimo

1. Anything testing 3mA idss and above. That includes the upper reach of K and all else. I did not see significant THD differences in circuit between the fat and the slim during breadboarding although in the end I plan to retest those subjectively too. Right now I run two slims.

2. I do not hear anything when powering on before the clicking as it is. When it clicks I hear a faint "plop" so I am happy with that. Tham reported the same on his high sens speakers. Normally nobody turns on the power amp first and only powers down a preamp last. One of the beta testers does not use the relay circuit at all because he always follows that sequence. But I think a delay should always be there since there is a DC settling cycle just to be safe of accidental misuse. You want to arrange it shunting, fine. Easy to re-arrange it under the relay pins. Just ground the second from the top and short the third with the fourth on each side. I will also recommend alternative 100k Ra to extend the wait time to 10sec now that I think of it because not all PSU types have the same ramp up profile and may promote longer settling.

3. That's nice theoretically and I am aware of it but with higher value 1uF C5 I wasn't happy with the DC offset chase speed and control. Too slow a control loop in this one. I started with higher value R14 also but I wasn't happy before 10K too. Many OAs look fine for the job on paper or work best in many designs. I just recommend what I can test best from what I got in this one. Freely test more and choose your own. AD825 is a nice single cousin but SMT unfortunately. Needs adapter.

4. That's for ease of heat with double mono PSUs and its the most benign series. Has no error amp, its not a real regulator, sounds good. Because there is PSRR in the main circuit. I can only tell you how near to using shunts when I will compare a fully done boxed preamp to my friend's that has the shunt. Anyway PSU as choice is individual, its a separate pcb audio circuit that can make use of anything. Can even test with an LM317-337. My bench supply wasn't making it sound bad for instance. Although the stabilizer was better. I have seen tendencies for instability without C6 C7 due to PSU incoming wiring. I arranged to two wire mode my friend's beta with the single common SSLV and it stopped. I had also seen it unsettling with long leads from my lab supply. You can use 85C Nichicon FG for those because not near the hot end. If stable rail on the scope in your wiring arrangement you may skip them or think of something fancy. But local reservoir near the Mosfets is a good thing in general.
 
Any >3mA idss sample and preferably of no more than 2mA idss difference to use between channels will do fine. The LTP's tail CCS forces the units to bias at about 2.5mA Id per LTP side. There the Yfs is the same for any >idss unit because fixed lower on the curve. So no need to have many for producing careful matches.
Q1 & Q2 are a mirror set up to pass unequal currents down to J1 & J2.
J3 sets the total tail current, it does not set the LTP halves to equal current. The current through J3 could be anywhere from 51% to 80% of the J3 tail current. That would leave anywhere from 49% to 20% to pass through J1.
I have not built this arrangement, but I can see that J1 and J2 will not have equal current and thus have non equal transconductance.

And there is no fixed resistor to show what current is passing through J1, J2

I think a fixed resistor in the emitter of Q2 would sort this non equality.
Maybe there is another reason that Salas has used a non-mirror to set the LTP currents.
 
Last edited:
diyAudio Chief Moderator
Joined 2002
Paid Member
Hence the word "about"

When matching of semis is perfect & VR1 is at zero, then the bias between J1-J2 will be 0.05mA different only. VR1 is to help bring offset close to zero initially. In practice it takes near to four Ohms of VR1 to balance the whole offset thing.
 
Hence the word "about"

When matching of semis is perfect & VR1 is at zero, then the bias between J1-J2 will be 0.05mA different only. VR1 is to help bring offset close to zero initially.
Yes. I agree, if both the BJTs and the jFETs are matched, then the currents in the two halves will be the same. The offset error should also be zero.

Getting perfect matching is nearly impossible.
So adding in an offset zeroing is wise.
I would expect to see a fixed resistor in the Q2 emitter lead. Maybe a value of half the VR. That way an inital pre-setting of the VR to 10ohms should give close to zero offset, if the four transistors are close.

But that still leaves the the non equal currents in the left and right halves due to non matching of the four transistors. And no method to check that in your layout. One would have to set the VR and measure it to then find out what current is passing using Ie=Vre/Re But as soon as you adjust the VR you lose that and you have to re-measure Vr again to check for balance.

A pair of low value source resistors and/or a pair of emitter resistors helps check the set up.
You can then use that to check the imbalance and set it deliberately to 1.5:1 or 2:1 if that is what is required, or check it is better than 1.05:1, if that is what you want.
 
Last edited:
I have done the 10 Ohm thing and it leads to little more THD and bigger initial DC transient. The set-up method I ended with is to let warm up and trim offset at the output via VR1 before putting U1 op-amp in its socket.
if you don't approve of source resistors then use emitter resistors in the mirror.
200r on both sides and then a 10r on Q2 side and the 22vr on the Q1 side.
This lets you dial out the offset and lets you see the balance of the LTP.

Big mirror resistors don't harm performance.