One BJT line amp/buffer

Jump to Latest
G

ginetto61

Member
Joined 2009
Ionmw said:
... The simulations did not show these distortions...
Click to expand...
good morning I apologize if I insert myself brutally
I would like to know what difference you found between measured THD and THD from LTSpice
If it were a high value this would seem worrying to me.
If there is low correlation even for such a basic circuit i wonder what could happen with more complex ones
Did you use a model provided by a manufacturer and buy the specific component?
Could there be problems with the model?
 
D

dkfan9

Member
Joined 2016
1714958627905.png


I built this up in sim early in the thread but it didn't fit the criteria and then discussion shifted to the 317 circuit, and I never posted it. Using Barry Blesser's "Super Transistor" (h/t Scott Wurcer) and @MarcelvdG's idea of CRDs, I tried to stay within the bounds of the original criteria as much as possible. I think this would require one trace cut in the original circuit, at the original base. The top side (schematic-wise) of the cut would connect to the collector-base junction, the bottom side to the input emitter. I never saw details on the source and load impedances. I used 9.56k from original schematic as load R, and tested with 50, 1k, and 10k source impedances. I did not breadboard as I have no CRDs.

I used the Central Semi CRDs because their Spice models were easy to access. The CCL0130 models give 90uA rather than the expected 130uA. After checking the datasheet, 90uA looks about right for 27C, with 130uA at a higher temp (high, positive tempco--may require a different R5 value, a parallel resistor, or be unusable in real world application). Qspice sim attached, models embedded. For other programs, Central Semi CRD model file attached (CCL.txt), and BJTs are the Cordell models.

Sim results (when three values given, read as RS=50/1k/10k):
-Input impedance (ohms): 116K with variance <1k across frequency.
-Gain(dB) @ 1k: -0.005/-0.08/-.72
-Frequency Response: <-.04dB at 20hz
-Input referred noise @ 1kHz (nV/rtHz): 2.8/6.3/34.8
-Input referred noise @ 20Hz (nV/rtHz): 8.1/9.9/35.6
-SNR ref 1Vrms (dB): 128/122/108
-THD1k 1Vrms (dB): -119/-111/-96
-THD1k 5Vrms (dB): -78/-68/-52
-THD20k 1Vrms (dB): -102/-92/-76
-THD20k 5Vrms (dB): -78/-67/-51

FR can be flattened by increasing output C. LF Noise rise can be reduced by increasing input C. 47u for both caps gets rid of both issues.

THD is primarily 2H. Lowest load resistance that does not clip with 7.071Vpk input is 1.5k.

Input impedance is parallel of R5 and the super-transistor's "base" at emitter Q1. At LF, the impedance looking into Q1 emitter is approximately Beta(Q2)*RL, but at megaohms like you get with 10k load, it's less (presumably early effect and output impedance of current source).
 

Attachments

  • CCL.txt
    5.3 KB · Views: 5
  • Cordell-Models.txt
    19.6 KB · Views: 5
  • ST Line Buffer Central Semi CRD.qsch
    12.4 KB · Views: 6
  • Like
Reactions: 1 user
D

dkfan9

Member
Joined 2016
Last edited by a moderator:
  • Thank You
Reactions: 1 user
D

dkfan9

Member
Joined 2016
Sorry for the broken links above. The current regulating diodes act as two terminal current sources, so they send a (relatively) fixed current. Internally, they are self-biased JFETs. See attached app note for info on this technique.
 

Attachments

  • 70596.pdf
    79 KB · Views: 4
Last edited:
  • Thank You
Reactions: 1 user
D

dkfan9

Member
Joined 2016
They could be replaced by any current source. Resistors should work but the output transistor would then need a 900 or 1k resistor to maintain 10mA bias, so input impedance would decrease (maybe around 90-100k), and performance would probably be somewhat worse.
 
  • Thank You
Reactions: 1 user
You must log in or register to reply here.
Top Bottom