ATL-04 Balanced headphone amp by al_tsankov

[Adhafera]

Hi al_tsankov,
I've just spend a little time to understand your schematic. I've understood most part of it, but some obscure points still remains. Could you help me to figure it out and check if I didn't misunderstand the schematic ?

So first of all, the simplest part (my two cents) is the output stage composed of a current source (J7/R67) and some current mirrors (Q2/Q20) which will duplicate the current reference (call it ib) to the diamond buffers via Q22/Q24 and Q1/Q19.
This diamond buffers (Q13/Q14 and Q15/Q16) will drive the output transistor (Q3/Q18 and Q4/Q17).
My question is how do you calculate the current biasing (class A) of these output transistor ? This current depends on R28, R17 and R21 (and respectively R19, R20, R22) isn't it ? Something like ib*R28/R21 (considering the Vbe of Q14 and Q3 are the same...) ?
Which value did you chose for Re (the emitter degeneration resistance depend of the biaising current of the diamond buffer i presume) ?
Why did you include 1ohm resistance R32 and R25 ? It increases the output impedance which was null (or very very small).
Last thing, i probably didn't search well but i don't find the datasheet of the J2SK117.

Now, the most tricky part : the input stage !
I'm not shure sure i understand this part well.
The IPS is composed of a floating complementary JFET-LTP (J1,J2,J3,J4,R3,R33). I'm not comfortable with the JFET because i never use it, so i don't exactly understand how it works. However, i suppose the N and P channel have to match to provide good performances. I mean the two halves have to match together. If not the gain of each half will be different and it will introduce harmonic distortion ? So to build this amp, we have to care about the JFETs matching ? Why not using only an N-channel JFET input ?

A last, the VAS. The signals arrive at the emitter of Q7, Q11, Q9 ans Q5 (with or without phase shit).
The voltage between D1/D2 and D3/J5 is fixed by the leds (J5/R54 imposes a constant current throwing the leds so the forward voltage is constant).
Edit : ahh ! just figure out the Q7/Q8 are juste a cascode stage ! I didn't recognize it ! Normaly, on traditional cascode, the signal arrives on the base of Q7.
Except these findings, i don't really understand how it works : what is R61 ? Same for R55/C3 : is it a feedback cap for stability ?

Last question : you say that for the moment the gain is 8. How to calculate it ? It depends on R16 and R16 (R52 and R24) ?


I know that it is a lot of questions, i'm sorry for that. But your amp is very interesting !
If you don't mind, could you share the LTSpice schematics with the components (JFETs and output BJT are not include with LTSpice) so i could simulate it to improve my understanding of the amp !

Thanks a lot for sharing your experience here !
Adhafera

[al_tsankov]

Hi Adhafera,
You are in the right direction .
According to diamond buffer - yes, Q1, Q2, Q22 and Q19, Q20, Q24 form a current mirrors. Their currents are given by the current source formed by J7, R67. J2SK117 is just a 2SK117 - the model in LTSpice is named J2SK117 and I simply use it as is. The output transistors standing currents are calculated according to formulae given in the Walt Jung excellent tutorial about Diamond buffers: Iq3 = Iq14*R28/R21. Of course, since the BJT's positive temperature coefficient, the resulting standing current is a little bit higher than calculated. For example, if J7, R67 are the current source with I=5.5 mA, R28=40 Ohm and R21 = 2.7 Ohm, Iq3 = 5.5*40/2.7 = 81.5 mA. Upon the time Iq3 rises to approx. 85 mA and stays stable since the local NFB imposed by R21.
As I've already wrote, 1 Ohms output resistors are added only for safety reasons (and may be with mind for another designs - non balanced, with TRS jacks, which can be prone to shorting). Of course, in the balanced design it's impossible and the output resistors can be omitted. OTOH, my headphones has 600 Ohm impedance and that's why I just can ignore them.
For the Input stage - it's the so called self biasing scheme which is rooted to Erno Borbely's designs, but is also used by many others - John Curl for example, Ti Kan's Beta22 etc. Of course, all 4 JFETs must be very close selected and it's the trickiest part. That's why the price of matched JFET quads is so high nowadays. Let's hope that Linear Systems will flood us with their LSK170/LSJ74 in the near future . Well, of course all design can be done with only one pair of n-JFETs but I prefer the symmetry .

The VAS is formed by folded cascodes (Q7, Q9, Q5, Q11), which in turn are cascoded by Q8, Q10, Q6, Q12. R59, R61 are introduced in order to form the OLG, since Q8, Q10, Q6, Q12 cascodes will rise it. The VAS currents are calculated as follows: Id2 - 0.65 = (Ij2 + Iq9)*R9
If Ij2 = 3.5 mA and R9=120 Ohm, it yields Iq9 = (Id2 - 0.65 - Ij2*R9 )/R9 = 6.5 mA.
R55, C3 and R43, C1 are frequency compensation networks which commonly are connected in parallel with feedback resistors R16, R52. However, I decided to take the signal from the VAS output, not from the OPS. The gain for this schematic is calculated simply dividing R16 to R15: 47/5=9.4 Well, I just put these values in the simulator. In the real world they are a little different, but it doesn't change the things.

Well, I'll send you the SPICE files, but I have to figure out how this can be done via the forum and beside that I'll have to see which models I should include.

Regards

P.S. - call me Alex if you don't mind .

[Adhafera]

Thank you Alex for yours explanations ! I get it
I love your design ! It gives me a lot of ideas.
I think i will built it in some months (when i'll have finished my on going projects...)

[GourouLubrik]

Did you finally compare your ATL-04 with Salas shunt PSU against your modified Sulzer PSU?

[flying baboon]

Finalement quelqu'un s'est lancé dans la construction de l'ampli ?

[flying baboon]

Bon aller je me lance je le prends, mais je vais faire ma feignasse et le commander pré-soudé

Du coup comme rectifier je pensais essayer ça : http://sjostromaudio.com/pages/index.ph ... ier-bridge (merci G600 pour ce site !)
Je vais demander à Alex ce qu'il a utilisé.

EDIT : en fait, l'ATL-04 est encore au stade de prototype.

[GourouLubrik]

Bon aller je me lance je le prends, mais je vais faire ma feignasse et le commander pré-soudé  :mrgreen:

Du coup comme rectifier je pensais essayer ça : http://sjostromaudio.com/pages/index.ph ... ier-bridge (merci G600 pour ce site !)
Je vais demander à Alex ce qu'il a utilisé.

EDIT : en fait, l'ATL-04 est encore au stade de prototype.

Tu as eu des nouvelles depuis ?
J'aimerai bien le tester aussi... pourquoi pas au fesses d'une balanced DAM :D

[flying baboon]

Non il faut que je recontacte Alex pour savoir si il a pu avancer sur l'ATL-04. Je lui envoie un mail

[DrizztDo]

Salut, finalement ce projet a aboutit ?

[corderaide]

Me semble qu'il est en chemin vers chez moi.

[DrizztDo]

Oh je suis intéressé par ton retour alors. Tu utilises l'alim inclue ou une custom ? Tu as pris le pcb déjà soudé + boitier ? J'ai besoin de me trouver une occupation en ce moment...

[frenchbat]

Me semble qu'il est en chemin vers chez moi.

C'est le ATL-03MKII qui est en route, Corde ;)

[GourouLubrik]

Non, pas de nouvelle sur l'ATL-04, par contre, il existe maintenant un prototype d'ATL-03 MK2 avec des modules JLsounds I2SoverUSB + DAC AK4490

@cordeuse forever

[Genius753]

Non, pas de nouvelle sur l'ATL-04, par contre, il existe maintenant un prototype d'ATL-03 MK2 avec des modules JLsounds I2SoverUSB + DAC AK4490

@cordeuse forever  :mrgreen:

Des news de l ATL-04 depuis?!il m interaisse trop...
A commencer d ici 3/4 mois donc si sa c'est toujours pas debloqué d ici la,je me dirigerais sur autre chose...malheureusement...

[Genius753]

Tjs aucune nexs gourou?