Constant Current Source (CCS) For Audio Applications

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I used this setup to measure the current noise of the conventional DMOS cascode current source. The amplifier is the Colin design with noise which ranges 1nV/Rt Hz to a bit lower. The CCS was set for 1mA DC. Frequencies examined 1kHz and 10kHz.

Measured voltage out is 37.2nV/Rt Hz == which if you apply the formula is 3.5pA/RtHz
 

Attachments

  • CurrentNoise_DN2540.png
    CurrentNoise_DN2540.png
    18.8 KB · Views: 1,330
  • Like
Reactions: 1 user
I used this setup to measure the current noise of the conventional DMOS cascode current source. The amplifier is the Colin design with noise which ranges 1nV/Rt Hz to a bit lower. The CCS was set for 1mA DC. Frequencies examined 1kHz and 10kHz.

Measured voltage out is 37.2nV/Rt Hz == which if you apply the formula is 3.5pA/RtHz

I should have mentioned, the switched is closed prior to test. This drains the DC voltage to zero at the preamp input. 10uF is overkill as the f3 is ridiculously low.

The best way to operate is to allow the preamp to settle for 4 or 5 minutes, run a noise test with R4 shorted. S1 closed.

Easier to download the data into Excel than write an AP-Basic Macro!
 
Member
Joined 2011
Paid Member
The LM334 with the addition of a resistor and diode (resistor is 10X Rset, 1N4148) will give you a zero tempco current source.

What!??!?!

The LM334 datasheet takes great pains to instruct you NOT to use a gold-doped superfast diode like the 1N4148, whose Vfwd and whose tempco dVfwd/dT do not track the (not gold doped) junctions inside the LM334 IC. Instead, the LM334 datasheet tells you to use the (not gold doped) 1N457 diode. Figure below.

~
 

Attachments

  • datasheet.png
    datasheet.png
    16.9 KB · Views: 1,188
Although you won't see any improvement in sims with perfect dc rails, in a real world circuit with noise this can have a major improvement in psrr.

thanks for this tip jerulwoo, here is a graphical comparison showing the improvement vs a couple of other circuits (flatter slope == better)
 

Attachments

  • Screenshot_2020-10-05_04-26-12.png
    Screenshot_2020-10-05_04-26-12.png
    42.6 KB · Views: 662
  • bjt-current-sources.asc
    2.9 KB · Views: 71
Member
Joined 2011
Paid Member
In many cases one needs a constant current from a rail, then biasing becomes straightforward and there is no advantage to stick to the dogma of a two terminal current source.
There are truck loads of these CCS inside ICs, that do not float

Member cellularmitosis agrees with this; and when you scrutinize the attachments to his post #69, you will see that he is applying and measuring more-than-two-terminal current sources. Study the plot legend in panel 2. There it is.
 
mchambin you are right, it is slightly confusing the way these are drawn, as they aren't true "two terminal" sources (because the load is inside the two terminals). I drew them in this way to make it more obvious where the boundaries between each circuit where. Perhaps it would have been better if I had drawn boxes around them instead.

Relatedly, I have started working on a directly-coupled buffer for my headphones which uses this current source: directly coupled class A buffer for low-voltage headphones (work-in-progress)
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.