Nice Andrej, but K170 & J74 tend to positive TC in my experience when selecting them for IDSS. Also high Ciss. I have in hand a few BF245C and BF245B. Tested them, 10mA ones they go lower current about 0.3 mA relatively fast when left on 9V battery in G&S shorted IDSS mode. The delicate part is to find some already near target, because the source resistor (or trimmer) slows down the -TC time as I have seen with 100R. A 10mA one goes 7.7mA and moves negatively slower for just 0.1mA. Nice is that the C are almost 15mA all of them and the B are 7mA and 10mA groups. So a small trimmer can just hold the 10mA ones to trim around 8mA centre. We don't need PJFET, we can orient drain of NJFET towards the input transistor and keep them same die characteristics. Now if the steady current will be the main stabilizing force and not the supplementary TC tendency, maybe we can use many different JFET types and trimmers. Remains to be seen in practice. Soon.
Sometimes, aesthetic of the schematic can provide benefits to the sound. I understand-you well. Harmony is the target.
I mean, every *love* or pleasure you bring in a work will be given back a way or an other. Reason why no computer can design a "musical" power amp, neither a good text translation ;-)
More than that, under some reptilians reflexes we can have, we can discover sometimes some hidden intuitive and not analyzed knowledge witch will go to the surface later.
That makes the differences between our works, can bring to new ideas...
Well, i use to make symmetrical passive filters for my speakers. It first look stupid as it cost more components. it makes a difference, i tried to explain and now, i understand why and stay on it.
Last edited:
Member
Joined 2009
Paid Member
Ruwe, Apart from the unruly idle current how bad was the dc offset stability ? - did you physically tape together the two input devices ?
Thanks.
... If this solution would prove to be adequate, than you can simply omit j-fet from your VAS stage.
It will mean asymmetric cascode load resistors (need be trimmers). It will shoot fast the output Iq one way or opposite polarity if steady same resistors. Much as Ruwe had. Latfets have no symmetric Vgs and as the thing changes tempco current difference in the bridge branches it takes 4 hands to trim fast.
In practice the drivers common bias CCS kicks at about 150mV minus that homogenizes in the Latfes Vgs bracket enough.
Not sure. Intuition. At least not as good as a 4pF one for CCS PSRR. Choosing to test with the BF because I have both. Against DN price, rarity in Europe, size, big resistor to get down the Idss, probably noisier as Mosfet VS Jfet... But for basic current stability test it will do it too. As LM334 will and other we may have.
My question of speed was based on weak intuition. I guess that gate to source capacitance would affect high frequency PSRR. At lower frequencies PSRR would be just a function of transconductance? I asked about the DN part, as I have a bunch here (TO92 versions, so size not an issue). Would we bond these thermally to the input devices?
Sheldon
Bigun,
Not only I glued them together but I enveloped them in a bulb of RTV silicon for stable DC. Another factor improving DC stability was increasing the emitter resistors. I tried 22 ohms and it's definitely more stable. At turn-on the amp will go to about 60-70 mV and then slowly will stabilize around ±10mV.
To all of you guys, thanks for the ideas and the feedback!
To me the most important finding was to force more current from the output into the input. We call the circuit "current feedback", but we use it basically as a regular voltage feedback, with relatively high value resistors. I saw the feedback resistors in the F5. N. Pass is using 50 ohms/10 ohms for amplification of 5. To me, when you decrease the feedback resistor values you have better measuring and sounding amplifier. The rest is just to give enough current to the Zeners in order to keep them in stable regulation.
Andrej, I'll try the 1uF capacitor for stability. I kind of find it redundant, but you mentioned several times that it helps. So, I'll insert one and evaluate again the oscillation behaviour.
Thanks.
Yes, I was thinking of high frequency PSRR. They will run different dissipation to input transistors, better not bond.
Hey guys all these mV's are really nothing, they won't go anywhere unpredictably, they are just a sign that this amp really impatiently waits for the signal to come.
Achilles' heel of this design is bias thermal stabilization. Its source of instability is base-emitter diode of the input transistors, small voltage decrease of this diode cause smaller DC voltage on the feedback ground resistor (100 ohm) consequently this forces the rest of the supply current to emitter. So decreasing the supply injecting current to the bridge according to temperature variations would stabilize the quiescent current of the input stage. If we can find a simple solution it would suit all SSA versions like a glove.
Achilles' heel of this design is bias thermal stabilization. Its source of instability is base-emitter diode of the input transistors, small voltage decrease of this diode cause smaller DC voltage on the feedback ground resistor (100 ohm) consequently this forces the rest of the supply current to emitter. So decreasing the supply injecting current to the bridge according to temperature variations would stabilize the quiescent current of the input stage. If we can find a simple solution it would suit all SSA versions like a glove.
What you did to Crescendo is just simply amazing, I am impressed.
P.S. Google translate works just fine with your text here.
Could you explain the subjective sound when you achieved 1000V / uS slew.
I would also strive for high slew rate but in practice I have found that slower can sound more pleasant and realistic in some ways.
thx, mike
If you choose an high bandwidth, you will see rise in current in the driverstage before the the output.
Especially, if you have a zobel network in the output.
Without zobel, i have 60mA peak in my mirand amp by 140V. With Zobel, i am at 120mA --- Zobel is 22n + 2.2R
You have to check if it is within the limit of the transistors Collector and Basis current.
Especially, if you have a zobel network in the output.
Without zobel, i have 60mA peak in my mirand amp by 140V. With Zobel, i am at 120mA --- Zobel is 22n + 2.2R
You have to check if it is within the limit of the transistors Collector and Basis current.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.