What people typically have been referring to as the "X resistor" is set to zero ohms. That is probably what you are not seeing. Ordinarily there would have been two current sources in the middle, not one. If you look carefully at current levels (reads the middle bottom resistor at half the value of the two surrounding it), you will see that the one current source does the duty of two such units.
Now what happens when the resistor that has been created so much furore is zero?
a) You get more gain.
b) I question whether there are other effects (other than trimmability)
Since JFET's are used, their current gain is much smaller than that of MOSFET's so I don't have a problem with this. In fact, I have my misgivings about the need for a non-zero resistor in there anyway. How small can a resistor be before we are no longer "X-ing" seems a mute point.
Again, nice work, Uli. I am looking forward to impressions.
Petter
Now what happens when the resistor that has been created so much furore is zero?
a) You get more gain.
b) I question whether there are other effects (other than trimmability)
Since JFET's are used, their current gain is much smaller than that of MOSFET's so I don't have a problem with this. In fact, I have my misgivings about the need for a non-zero resistor in there anyway. How small can a resistor be before we are no longer "X-ing" seems a mute point.
Again, nice work, Uli. I am looking forward to impressions.
Petter
X idea
Hi
when you take a look at the AX (or XA?) designs you see this
resistor being zero.
IMHO the X does not depend on this resistor.
NP stated somewhere that this resistor is just to control too
high open loop gain which can cause instability.
The X comes from the double feedback R17 & R18.
When you feedback the signal from drain to gate (what happens
here) the errorsignal can be seen at the source which couples
this errorsignal into the source of the second fet thus creating
a double (in phase) mirror image at the drain of the second fet!
Actually you normally need to scale down this coupling to control
mirror chamber effects as seen in the X-Amps. As the admittance
of the JFETs is fairly lo there is no need to scale down!
This is not a standalone design but meant to be
put in as a replacement of existing gainstages!
This means you have to apply the same kind of external circuitry
as if it were an OpAmp.
Thanks
Uli
rtirion said:
Hi
when you take a look at the AX (or XA?) designs you see this
resistor being zero.
IMHO the X does not depend on this resistor.
NP stated somewhere that this resistor is just to control too
high open loop gain which can cause instability.
The X comes from the double feedback R17 & R18.
When you feedback the signal from drain to gate (what happens
here) the errorsignal can be seen at the source which couples
this errorsignal into the source of the second fet thus creating
a double (in phase) mirror image at the drain of the second fet!
Actually you normally need to scale down this coupling to control
mirror chamber effects as seen in the X-Amps. As the admittance
of the JFETs is fairly lo there is no need to scale down!
This is not a standalone design but meant to be
put in as a replacement of existing gainstages!
This means you have to apply the same kind of external circuitry
as if it were an OpAmp.
Thanks
Uli
Again, excellent work, Uli. I have som suggestions which may improve the circuit:
1. Gain is ideally R17/R15 per half. Thus, there seems to be a mistake with both the ratio AND the input resistance of 100 Ohms. I would like to have an indication of how the circuit works, but even with the low current gain of the input JFET's, I would expect the gain to be far too high for my taste. May I suggest 10K or 20K in both positions for a nice gain of 1 which is useful for modern digital equipment. The way it is now, according to theory, the input impedance should be about 100 Ohms per half. I admit it will be more with current setup but still why not stick to X basics?
2. R11 and R13 should not really be there .... As I see it, they kind of ruin everything X since there are so many volts dropped across them (almost 5 volts for each) which means that they assist in keeping the voltage at "zero", but I don't understand the rationale for having them unless it is to keep power dissipation of the transistors around it down.
3. I would change both the red LED's and the 9V1 Zeners with 5.6V zeners or thereabouts. This would make the current sources better performing and more stable over time/temperature as well. As we know, it is important to have stability in these quarters to allow for non-drift of the "collector output" output voltage (less item 2 above). In fact, it is possible to dispense with the bottom BC549 cascodes completely ..... which is what I will likely do now that I keep looking at it. Your mileage may vary.
4. I personally prefer to have the 9.1 Zener bypass caps go all the way to the rails. If one is concerned with zener noise, that is easily taken care of by small caps in addition if deemed necessary.
5. When I build this (because I likely will, thanks Uli), I will likely have more current passing everywhere, and particularly in the outer sections. Probably try to use 10mA for the center and 30 or so for the outer sections. This will of course require proper selection of the input transistor current group.
6. Is there a typo for the input transistors? Should it be 2SK389?
If this has been built as proposed, I would be very interested to hear what the gain was, and how it sounded.
Petter
1. Gain is ideally R17/R15 per half. Thus, there seems to be a mistake with both the ratio AND the input resistance of 100 Ohms. I would like to have an indication of how the circuit works, but even with the low current gain of the input JFET's, I would expect the gain to be far too high for my taste. May I suggest 10K or 20K in both positions for a nice gain of 1 which is useful for modern digital equipment. The way it is now, according to theory, the input impedance should be about 100 Ohms per half. I admit it will be more with current setup but still why not stick to X basics?
2. R11 and R13 should not really be there .... As I see it, they kind of ruin everything X since there are so many volts dropped across them (almost 5 volts for each) which means that they assist in keeping the voltage at "zero", but I don't understand the rationale for having them unless it is to keep power dissipation of the transistors around it down.
3. I would change both the red LED's and the 9V1 Zeners with 5.6V zeners or thereabouts. This would make the current sources better performing and more stable over time/temperature as well. As we know, it is important to have stability in these quarters to allow for non-drift of the "collector output" output voltage (less item 2 above). In fact, it is possible to dispense with the bottom BC549 cascodes completely ..... which is what I will likely do now that I keep looking at it. Your mileage may vary.
4. I personally prefer to have the 9.1 Zener bypass caps go all the way to the rails. If one is concerned with zener noise, that is easily taken care of by small caps in addition if deemed necessary.
5. When I build this (because I likely will, thanks Uli), I will likely have more current passing everywhere, and particularly in the outer sections. Probably try to use 10mA for the center and 30 or so for the outer sections. This will of course require proper selection of the input transistor current group.
6. Is there a typo for the input transistors? Should it be 2SK389?
If this has been built as proposed, I would be very interested to hear what the gain was, and how it sounded.
Petter
Petter said:Again, excellent work, Uli. I have som suggestions which may improve the circuit:
THANKS!
1. Gain is ideally R17/R15 per half. Thus, there seems to be a mistake with both the ratio AND the input resistance of 100 Ohms. I would like to have an indication of how the circuit works, but even with the low current gain of the input JFET's, I would expect the gain to be far too high for my taste. May I suggest 10K or 20K in both positions for a nice gain of 1 which is useful for modern digital equipment. The way it is now, according to theory, the input impedance should be about 100 Ohms per half. I admit it will be more with current setup but still why not stick to X basics?
Those arent input resistors but gatestoppers!
This circuit must be looked at like an opamp.
You need some external circuitry!
2. R11 and R13 should not really be there .... As I see it, they kind of ruin everything X since there are so many volts dropped across them (almost 5 volts for each) which means that they assist in keeping the voltage at "zero", but I don't understand the rationale for having them unless it is to keep power dissipation of the transistors around it down.
R11 and R13 are in fact bypasses for gate-source of an optionally added sourcefollower. the layout give you the possibility to split
the circuit in those places
3. I would change both the red LED's and the 9V1 Zeners with 5.6V zeners or thereabouts. This would make the current sources better performing and more stable over time/temperature as well. As we know, it is important to have stability in these quarters to allow for non-drift of the "collector output" output voltage (less item 2 above). In fact, it is possible to dispense with the bottom BC549 cascodes completely ..... which is what I will likely do now that I keep looking at it. Your mileage may vary.
IMO the emitterresistor of a BJT CCS should be kept as lo as
possible to avoid noise!Just a bit higher than BE drop to assure
thermal stability. The zener can easily be swapped to 5.6 or so
no problem.
PS: a red LED is far superior than any zener in terms of noise.
4. I personally prefer to have the 9.1 Zener bypass caps go all the way to the rails. If one is concerned with zener noise, that is easily taken care of by small caps in addition if deemed necessary.
ok for me.
5. When I build this (because I likely will, thanks Uli), I will likely have more current passing everywhere, and particularly in the outer sections. Probably try to use 10mA for the center and 30 or so for the outer sections. This will of course require proper selection of the input transistor current group.
Simply the reason is that I have lots of 100R 0.1% resistors lying around
6. Is there a typo for the input transistors? Should it be 2SK389?
No the 2SK369 is an ultra lo noise JFET, except "matching" it is
better than 2SK389 although the PCB accepts both.
If this has been built as proposed, I would be very interested to hear what the gain was, and how it sounded.
As I use it in my pre (for which Dale and Craig are doing a
superjob customizing their APOX system!) I have this design
embedded with external DC-servo and complementary source
followers (IRF620, IRF960 running at 40mA).
Therefore the openloop gain is somewhere between
44 and 50 dB. As I use them with feedback the stage before
volume has input resistors of 20k and fb resistors 50k,
the outputbuffer has 10k input and 40k fb.
Petter
sorry for adding text above
Uli
Uli
Thanks for the explanations. I get it now. It is only healthy that we agree on most things and have different opinions of a few other.
If you are using this as an op-amp building block, I also assume you are taking the feedback around the output stage using R17 out to the end (my preference unless much lower gain is used at the "op-amp core"), or did you prefer to use the high gain op-amp circuit in the traditional op-amp way? Or did you simply add a large input resistor??????
Hmmmm
Petter
Thanks for the explanations. I get it now. It is only healthy that we agree on most things and have different opinions of a few other.
If you are using this as an op-amp building block, I also assume you are taking the feedback around the output stage using R17 out to the end (my preference unless much lower gain is used at the "op-amp core"), or did you prefer to use the high gain op-amp circuit in the traditional op-amp way? Or did you simply add a large input resistor??????
Hmmmm
Petter
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.