I am working on a differential tube preamp with a CCS for the LTP. I would like to design a low noise, high impedance current source with reasonable temperature stability and would like to understand the pros and cons of using a BJT vs JFET vs MOSFET. I am looking for 10-15mA and have about 25V to work with. I don't mind some circuit complexity (although I don't have tons of space) or spending some $$ to get the best performance.
I was considering an op-amp based circuit using a precision voltage reference driving a MOSFET, but I'm not sure if this would deliver the best performance.
Thanks for any assistance.
I was considering an op-amp based circuit using a precision voltage reference driving a MOSFET, but I'm not sure if this would deliver the best performance.
Thanks for any assistance.
Have you considered the LR8N3-G? It's a typical 3 pin regulator good to over 400v. TO-92. It will provide up to 30ma. I am using one now in a tube power supply. IMHO the regulation is not as good as the lower voltage 317 types, but with additional filtering stages I managed to get the ripple down to about 7mv -- on a breadboard! Hoping for 3-5mv on a pcb with a ground plane.
CCS are not really a noise source for a LTP since noise would be common. Besides, it is not in series with the signal, so not contribute it's noise in the signal path.
Op-amp and precision volt-ref is the best performance wise. You can also take advantage of zener diodes tempco together with bjt's Vbe and/or regular diodes tempcos and make a very low tempco CCS. A 6.2V zener has approx +2mV/C tempco, and together with a regular diode or Vbe junction you get close to zero tempco.
Op-amp and precision volt-ref is the best performance wise. You can also take advantage of zener diodes tempco together with bjt's Vbe and/or regular diodes tempcos and make a very low tempco CCS. A 6.2V zener has approx +2mV/C tempco, and together with a regular diode or Vbe junction you get close to zero tempco.
Assuming that you are planning to use the CCS in the tail circuit of the LTP, the LT3092 is worth a look. It's rated up to 200 mA and operates over a range of 1.2 to 40 volts. It's only available in SMT packages, however.
https://www.analog.com/en/products/lt3092.html
https://www.analog.com/en/products/lt3092.html
Thanks for the feedback.
@SemperFi - Regarding LTP noise source for an LTP, I can see how this would be true when there is no input signal, but wouldn't any noise be unbalanced once their was differential signal applied?
I'm also planning to use the same (or similar) circuit for a CCS for a balanced cathode follower stage.
Regarding the schematic, I was thinking of something along the lines of this one (grabbed from a different thread), but it would be operating between ground and -25V. I'd use a low-noise voltage reference (something like an LM329) instead of the 5V reg in this circuit.
Another alternative I was considering is using a MAX6126.
Thanks.
@SemperFi - Regarding LTP noise source for an LTP, I can see how this would be true when there is no input signal, but wouldn't any noise be unbalanced once their was differential signal applied?
I'm also planning to use the same (or similar) circuit for a CCS for a balanced cathode follower stage.
Regarding the schematic, I was thinking of something along the lines of this one (grabbed from a different thread), but it would be operating between ground and -25V. I'd use a low-noise voltage reference (something like an LM329) instead of the 5V reg in this circuit.
Another alternative I was considering is using a MAX6126.
Thanks.
@Ray Waters - The LT3092 looks interesting. I'll investigate this further. I don't mind some SMD. I'm doing a custom PCB.
Sorry I wondered what your preamp schematic was the LPT itself the cathode follower, the signals levels so you can work out what's needed rather than over-engineering. The LT3092 looks simple to use.
What about the LM334 with an external high voltage NPN? See the application notes for the circuit.
@baudouin0 - I'm building a version of Allen Wright's RTP3C line stage. The CCS for the LTP will see about 12V across it with very little variation. For the output stage, there will be about 28V across the CCSes with up to about 1V variation.
This is a fairly complex (and expensive) design, so I don't mind a little over-engineering 😉.
This is a fairly complex (and expensive) design, so I don't mind a little over-engineering 😉.
The LM334 is only rated at 10 mA and the OP indicated that he needs 10 to 15 mA. Within its current limits I've found the LM334 to work great. but I always use the temperature compensated configuration in tube circuits. That application is discussed in the datasheet.
The LT3092 has an impedance of only 10K ohm @ 20 Khz. according to the datasheet. For lower frequencies the numbers improve.
Or am I overlooking something?
Regards, Gerrit
Or am I overlooking something?
Regards, Gerrit
2SK246BL makes a very simple (jfet + 2 resistors) current sink that doesn't require extra power supply. The problem is low current (3-7 mA) for a single unit, so need to parallel 2-5 of them for 15 mA.
Walt Jung published an excellent two-part article in audioXpress magaine about various CCS configurations. Link below.
https://audioxpress.com/article/Sou...lator-Tests-for-High-Performance-Full-Article
In Part 1 he discusses the LM334 and includes a comment about its frequency-dependent characteristics. In particular, Walt states that "these impedance characteristics are exceptional, particularly at the lower frequencies where rejection is a few dB above the setup residual noise, even for the basic curve. This is indicative of an equivalent impedance of close to 10 Megohms, albeit with deterioration at the higher frequencies." In any case, I've always gotten good results with the LM334, especially given its simplicity.
Last edited:
There's a lot of options as described above. From a simple 6v2 zener which could power several NPN CCS's to some quite complicated stuff. There's also current mirrors which could be used too. The B+ supply is going to need equal attention. JFET's and MOSFETs have a big spread of current for different devices so OK for 1 off less good if you want consistency.
@baudouin0 - That's the RTP5 - a somewhat simpler circuit.
Here is the RTP3C.
I am building the line stage only. I have a -25V supply for the current sources. I plan to replace the FET circuit at the bottom of the output stage with whatever current source I end up with.
Here is the RTP3C.
I am building the line stage only. I have a -25V supply for the current sources. I plan to replace the FET circuit at the bottom of the output stage with whatever current source I end up with.
Yes I think a 2SK170/147 could be forward Vgs for some devices at 6.5ma. That output stage is quite mad.
I once used an LM317 based CCS in the attached differential amplifier, which operated at line level. However, I noticed that CCS too noisy.
Then I tested few simple alternatives and finally ended up with the CCS based on TL431 and NPN transistor.
The noise level decreased some 7 dB, which was at acceptable level. Dynamic resistance is more than 1M, which should be sufficient for most LTP's.
