Electronics with passion: amplifiers and power supplies, FET's and vacuum tubes..

LM317 as a Constant Current Source in audio cirсuitry

LM317 appears to be the most appreciated CCS building block amongst solder-slingers (term (c) John Broskie).

Personally I never felt relaxed having such a complex OpAmp-based circuit in the audio-signal path. Nevertheless many routinely utilize this type of CCS in audio, including "pure" tube designs.

Recently I was playing happily around cascoding JFETs and MOSFETs for the CCS in my own design. You know, depletion MOSFET is a rare beast... Ok, ok, this topic merits another blog entry.

Suddenly one of my best friends complained that he disliked the big deal the sound produced with help of an "extremely simplistic, single-ended, fully class-A" headphone amplifier that he just built for himself. This person (that I respect a lot) he counts much on his hearing abilities and it used to be an instrument in his professional career. But somehow I never took his "sound-engineer" claims too seriously.

When we were looking at the schematic drawing of that headphone thing we both got a strange feeling about - you already know what - LM317 sitting there as a CCS load for the source follower. In a minute my friend grabbed his true and tried soldering iron and replaced that poor LM317 by a big power resistor. Done. He went on a prolonged headphone listening exercise not worrying about those abused voltage regulators any longer.

Here I encourage astute readers to take a moment and study LM317 datasheet from NS: LM117/LM317.

The picture above (c) National Semiconductor. Please let me know if it disturbs - I'll take it out along with all references to the manufacturer.

I won't repeat what's already been said by many about the fact that there's an OpAmp and 100% negative feedback loop, the propagation delay and other parameters not really specified for audio signal amplification... I just wanted to see how good or how bad LM317 is when being used as a CCS.

I started with testing it on DC. That showed pretty nice performance indeed! The internal CCS' resistance measured around 400KOhm, current regulation within 0.1% with the voltage change two-fold (from 15 to 30 Volts).

Then I have put this popular CCS into a reasonably basic schematic I had sitting on a breadboard on my desk. There was a current source controlled by voltage, loaded into 1.2KOhm resistance and "pulled up" by our CCS under test. The CCS had 12V DC across it, 2V (1.7mA) AC signal swing.

At first I applied 1600Hz sine wave - I could not notice any difference from what I had before with my own CCS (cascoded MOSFETs). No visible distortion on the analyzer's output, no extra noise. All clean and good. The 1.6KHz sawtooth signal, tried both polarities, has not revealed anything extra.

Only when I applied 100KHz meander (square-wave) signal to the construction - I started to see The Difference! Check it out for yourself on the pictures below.

LM317-based CCS @ 100 KHz
input signal at the bottom of the screen, output - above

How did you like that overshot from the negative feedback loop inside the chip? Ugly, wasn't it?
High load capacitance was not very attractive too.

Cascoded MOSFETs on the other hand did not bring anything special to the signal. Just some capacitive load. By a coincidence I had quite big devices sitting there; those were able to drive tens of amperes of the current - hence the higher capacitances; but still much lower that 1.5A-capable LM317 added.

Cascoded MOSFET CCS @ 100 KHz

Some may ask, and rightfully so - whether these artifacts that can only be observed on frequencies well above the audio range - whether these influence the quality of the sound produced by amps employing such solutions? It's not my battle to argue this. I am simply trying to avoid using the negative feedback loops in my designs.
Also from now on I will try to grub my friend with his sound-engineer ears  and put whatever audio equipment I produce through his thorough listening expertise. Just in case 😉

P.S. In case one wanted to build a CCS from two MOSFETS like shown above - please do not forget to attach gates via some resistance in order to avoid parasitic oscillations, 100 Ohm ... 1 KOhm will do. Regarding batteries - honestly, they are very handy. My favorites are lithium 3V "button" types and I do use them happily in prototyping.

Edit 2011.12.06

Reference Constant Current Source for DIY Audio and beyond
(and anything else that requires the really good CCS :)) - I have this project under development.
The "magic" CCS promises to have quite outstanding parameters:

  • I nominal as from tens of micro-Amperes up to 400mA (subject to some 50W power limit)
  • Very high dynamic resistance - hundreds of kOhms and up
  • Output capacitance around 10pF or less
  • Operational as from 2 to 3 Volts and up to 1kV (Yes: 1000 Volts! Same 50W limit applies)
  • high frequency capable - this is still to be measured...
  • Two-pins real CCS device: no batteries and no external PS needed, no feedback loops
  • Main power device in the industry-standard TO-220-3 case + a small PCB attached to it - easy to mount to the heat-sink of your choice (insulation washer and silicon thermal pad included in the kit)
  • Adjustable (in few ranges) and fixed current versions
  • Fixed versions are to be available in matched pairs and quads...

This little jewelry would serve as an excellent replacement in every place where people tend to use CCS based on adjustable power regulators. Tube gurus (or should I have said thermionic valve based designs) would benefit from such a high-voltage CCS as well - these should exhibit parameters better than a pentode-based CCS, while not eating any filament (or heater ?) current.

Should you be interested in purchasing such a device for your projects - please drop me a note!
I promise the prices will also be attractive! 🙂

1 Comment

  1. Pingback: LM317 как источник тока для аудио | MyElectrons.ru

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.