I am currently designing a low voltage (+/- 3V rails) amplifier circuit. I would like to have opinions on what the best low voltage constant current source design is for supply currents in the micro amps.
The dual transistor has among the lowest voltage of the CCSs.
I wonder if germanium transistors could offer even lower voltages for this?
I wonder if germanium transistors could offer even lower voltages for this?
If the Voltage changes, the current changes.
If the current changes, then the impedance changes.
You have gained nothing.
Using the modified circuit where the mirror has a CCS setting the reference current you do gain something.
If the current changes, then the impedance changes.
You have gained nothing.
Using the modified circuit where the mirror has a CCS setting the reference current you do gain something.
sure you have gained something... more voltage headroom over the CCS, you could make a dual mirror and insert a jfet CCs in the driving string. this will give you the largest possible voltage swing over the mirrored CCS output. this is how I most often make housekeeping in my pre-amplifier designs.
Last edited:
Is this for a diff amp? Sounds like you could just use a resistor at those currents. A couple of volts and a couple of microamps mean a megohm resistor.
Last edited:
A non-obvious problem with low voltages is choice of actual BJTs. Look at low voltage HF devices, because regular Vceo>=60V or so devices operating points will be well into the region where the early effect seriously impacts performance (like having much lower and non-linear beta than expected). This is alo very important with CCSs as they will get progressively worse as you approach the rail voltage, with effective impedance lowering by an order of magnitude or more.
Hi Brandon,
I recommend you "squander" an extra 35 millivolts or so, and use a pair of matched 1% resistors in the emitters of a current mirror. This will let you use individual BJTs (which are easy to buy) instead of duals (which are hard to buy), and compensate for small amounts of VBE mismatch.
I also recommend you build the current mirror out of BJTs that are known to exhibit excellent ICE flatness at very low VCE. What you really do NOT want are transistors that suffer from Quasi-Saturation. The ones favored here on diyAudio are the BC337-40 (NPN) and BC327-40 (PNP). They have reasonably high beta and excellent ICE flatness at very low VCE, see attached photo.
DigiKey, Mouser, Farnell_UK, and many other distributors sell these; they are in production today from several manufacturers.
Attached is an example schematic showing a very low voltage drop current source, applying these suggestions. To get the other polarity just turn everything upside down. Note that the LM334 has about 5 volts across it; that's waaaay more than enough to get waaaay out on the flat part of its I-V curve (see Fig.11 of the LM334 datasheet)
I recommend you "squander" an extra 35 millivolts or so, and use a pair of matched 1% resistors in the emitters of a current mirror. This will let you use individual BJTs (which are easy to buy) instead of duals (which are hard to buy), and compensate for small amounts of VBE mismatch.
I also recommend you build the current mirror out of BJTs that are known to exhibit excellent ICE flatness at very low VCE. What you really do NOT want are transistors that suffer from Quasi-Saturation. The ones favored here on diyAudio are the BC337-40 (NPN) and BC327-40 (PNP). They have reasonably high beta and excellent ICE flatness at very low VCE, see attached photo.
DigiKey, Mouser, Farnell_UK, and many other distributors sell these; they are in production today from several manufacturers.
Attached is an example schematic showing a very low voltage drop current source, applying these suggestions. To get the other polarity just turn everything upside down. Note that the LM334 has about 5 volts across it; that's waaaay more than enough to get waaaay out on the flat part of its I-V curve (see Fig.11 of the LM334 datasheet)
Attachments
Here's a simulation using BC337-40 transistors whose VBEs differ by 2.0 millivolts, typical of the variation seen in volume production. The LM334's current I_LEFT was set to 100 microamps, per the requirements in post #1.
The error caused by mismatched VBEs is plotted on the vertical axis; it is the percent difference between I_LEFT and I_RIGHT. As the voltage across the emitter resistors rises, mismatch error shrinks. You can see there's nothing particularly special or magic about V(Re) = 35 millivolts; it's simply one point on a smooth curve. Error is reduced 2.4X when V(Re) is 35 mV, from 8% to 3.3%.
_
Attachments
Firstly, I would like to thank you all for the great ideas and secondly I would like to ask Mark Johnson why a BJT current source and mirror? Why not a FET current source/current mirror configuration? Also how big does the capacitor have to be?
A JFet CCS is the best for low voltage designs, It's depletion and has the highest output impedance of any type CCS'S, but it needs trimming as the component spread is "big".
For mirrors I would suggest doubles or quads, there are many good suitable doubles, I have had good success with the 3904/3906 quads for cascoded mirrors, even with quads I degenerate the mirrors with some resistance.
For mirrors I would suggest doubles or quads, there are many good suitable doubles, I have had good success with the 3904/3906 quads for cascoded mirrors, even with quads I degenerate the mirrors with some resistance.
If I may answer this...
JFET CCS require a relatively high voltage to work properly, although this depends heavily on the choice of JFET. In any case it is desirable for the voltage across the JFET to be well over Vp (the pinch-off voltage), which is often a couple of volts. If not, the JFET will be working in it's 'triode region' or very close to the knee of the Id characteristic curve, hence more like a resistor than a CCS (or rather a bad CCS).
So, in principle, with around 5.5V or so available, you could use a JFET CCS carefully choosing the JFET. Also, once you do, keep in mind that Idss and Vp have significant variation between different instances of the same type part - often +-25%, hence more selection is needed or a means to adjust the CCS.
Fooling around with the JFET models that are presupplied when you download a free copy of LTSPICE, I built a few 100uA current sources and looked at their low voltage behavior. The basic idea is shown in Figure 1 below: a JFET (Pchannel in this case) is connected with a resistor in its source leg, and the gate is connected at the far end of the source resistor. Unfortunately the 2N5460 makes a current source which is unsuited to low voltage operation; the simulated current-versus-voltage behavior of Figure 1 is plotted in Figure 2. Bottom Line: using a 2N5460 to make a 100uA current source, you must apply more than 3 volts across the current source. And that is just too much.
So I tried five other JFET part numbers: another Pchannel, and then 4 Nchannels. These are plotted in Figure 3. None of them are anywhere near as low-voltage as you can achieve with a BC337-40 bipolar NPN transistor.
However, a painstaking search through a big collection of JFETs, just might stumble across one whose IDSS and VP gave a wonderfully low voltage 100uA source. It's not impossible, merely daunting. Looking at Figure 3, it seems clear that some JFET must exist which gives an I-V curve to the left of the red curve; the only trick is to find it!
_
So I tried five other JFET part numbers: another Pchannel, and then 4 Nchannels. These are plotted in Figure 3. None of them are anywhere near as low-voltage as you can achieve with a BC337-40 bipolar NPN transistor.
However, a painstaking search through a big collection of JFETs, just might stumble across one whose IDSS and VP gave a wonderfully low voltage 100uA source. It's not impossible, merely daunting. Looking at Figure 3, it seems clear that some JFET must exist which gives an I-V curve to the left of the red curve; the only trick is to find it!
_
Attachments
Last edited:
Lower Gm devices? KSK595 is N-ch, Idss is from 200uA to 500uA, Vp is typically 0.3V to 0.6V. Then there are the real low Gm, 2N4117/8/9 N-ch J-fets. 2N4117 Idss (Vgs=0V) is typically around 50uA. Use two in parallel for 100uA CCS. Very low noise this way, operating at Idss.😉
The 5460/1/2 P-ch J-fets tend to have a large Vp vs Gm. 2N177/6/5/4 P-ch have lower Vp seem to have a lower Vp vs Gm.
The 5460/1/2 P-ch J-fets tend to have a large Vp vs Gm. 2N177/6/5/4 P-ch have lower Vp seem to have a lower Vp vs Gm.
When you mirror in the currents, you gain several benefits, first you have plenty voldtage for the driving CCS, second you get an sink/source current-balance and last you consume the lowest possible voltage headroom
Somebody in this discussion thread mentioned something about a "JFET current mirror" .
Current mirrors are usually implemented with bipolar transistors, or with enhancement mode MOS Field Effect Transistors. But you very seldom (never) see current mirrors implemented with Junction FETs. So that's a fun little design challenge.
I will define "JFET current mirror" to mean a circuit which
A simulation plot is shown in Figure 2. Although this is not an especially low-voltage current source, it is a current source -- when Vapplied > 1.4 volts.
Perhaps this schematic shows why you seldom (never) see a "JFET current mirror"
Current mirrors are usually implemented with bipolar transistors, or with enhancement mode MOS Field Effect Transistors. But you very seldom (never) see current mirrors implemented with Junction FETs. So that's a fun little design challenge.
I will define "JFET current mirror" to mean a circuit which
- has an input port and an output port, AND
- current Iin is forced into the input port (by external circuitry), AND
- the "JFET current mirror" generates a current Iout (Iout=Iin) which flows into the output port, AND
- the input port and output ports are both drain nodes of Junction FETs, J1 and J2.
A simulation plot is shown in Figure 2. Although this is not an especially low-voltage current source, it is a current source -- when Vapplied > 1.4 volts.
Perhaps this schematic shows why you seldom (never) see a "JFET current mirror"
Attachments
It would be a good idea to provide some context to the question: does this CCS really need an extra high impedance, like as the load for a VAS, does its value need to be accurately and independently defined, or is something proportional to Vsupply acceptable, etc
If you post the schematic where this CCS is going to be used, it will probably answer most of the important questions
- Status
- Not open for further replies.