When the output voltage swings to one rail or the other, there is a voltage required for the cascode to be in opeation.
You subtract the cascode voltage from the rail and the voltage. That is what the output can swing.
You subtract the cascode voltage from the rail and the voltage. That is what the output can swing.
Last edited:
My intent with the LU1014D Modulated Cascode, which perhaps I failed to make clear, is to create a three terminal circuit that preserves the triode-like properties of the LU1014D over a more useful voltage range. This post gives some more detail about the circuit.
The first image defines the triode equations of the circuit. Sorry ZM.
The second image shows the quiescent state DC conditions and how Vcasc voltage is derived.
The first image defines the triode equations of the circuit. Sorry ZM.
The second image shows the quiescent state DC conditions and how Vcasc voltage is derived.
Attachments
Until I obtain valid spice parameters for simulation, I postpone it as my results lead nowhere.
Curious to see the output characteristics to 'swap' from convex to concave when Rs>0.
So the LU1014 is a SIT - I didn't realise that. It's a Nch j-fet I suppose? Is a Pch available too?
I've no experience with Pass-designs, although I've seen some single active device amps, and they proof very good according to many postings here.
Curious to see the output characteristics to 'swap' from convex to concave when Rs>0.
So the LU1014 is a SIT - I didn't realise that. It's a Nch j-fet I suppose? Is a Pch available too?
I've no experience with Pass-designs, although I've seen some single active device amps, and they proof very good according to many postings here.
SIT or not (my twisted logic sez not, but lazy to consult books from yore and actual datasheet), but it is certainly sand triode only when helped with everything explained here
call it Schade Feedback ( in broadest term), call it Schade modulated Cascode, call it Modulated Cascode, whatever
as shown in tube era with Schade Feedback, also later developed Ultralinear arrangement, each different tube type demands establishing own set of ideal parameters
same here with LU1014, "all" we need is to establish best compromise of operating parameters, to have SIT for masses
now, my personal logic would be to make this sans source resistor under LU, if possible
call it Schade Feedback ( in broadest term), call it Schade modulated Cascode, call it Modulated Cascode, whatever
as shown in tube era with Schade Feedback, also later developed Ultralinear arrangement, each different tube type demands establishing own set of ideal parameters
same here with LU1014, "all" we need is to establish best compromise of operating parameters, to have SIT for masses
now, my personal logic would be to make this sans source resistor under LU, if possible
During my curve tracer adventures I confirmed that the cascode mirrors the curves of the LU, just with higher Vds (no surprise with that). The cascode voltage has only a small effect on the point where the curves turn concave/convex. This means you need to look ag the LU curves to see the transition from convex to concave. This is very variable between different LU/LD1014 parts. Here are some examples: https://www.diyaudio.com/community/threads/ld1014d-lu1014d-matched-by-vgs-and-curve-tracing.381338/
ZM wrote:
That is exactly what is shown in post #63. The idea is a 3 terminal device containing the LU, the mosfet, 3 resistors (or 2 resistors and a zener), and a capacitor.
There is a surprising result in the gm[casc] equation in post #63 which is verified in simulations. I previously thought that the cascode FET needed have a very high transconductance, more than a single IRFP240 could provide.
In the equation gm[casc]=1/(1/(μ[J1]gm[M2])+1/gm[J1]), gm[M2] is multiplied by μ[J1]. At 1.5A the IRFP240 gm is about 5S and the LU has a μ[J1] of about 8.5, resulting in gm[M2]μ[J1]=42.5S. The LU has a gm[J1] of about 27S at 1.5A and Vds[J1]=5.1V.
The final result is gm[casc]=16.5S, which is excellent.
In the equation gm[casc]=1/(1/(μ[J1]gm[M2])+1/gm[J1]), gm[M2] is multiplied by μ[J1]. At 1.5A the IRFP240 gm is about 5S and the LU has a μ[J1] of about 8.5, resulting in gm[M2]μ[J1]=42.5S. The LU has a gm[J1] of about 27S at 1.5A and Vds[J1]=5.1V.
The final result is gm[casc]=16.5S, which is excellent.
Is that figure coming from measurement of real devices ?
If not, which Spice model are you using ?
Our measurements says it is more like 5.5S at 1.3A.
The manufacturer's spice model is for switching and is miles out for our operating conditions.
The one from Keko is much better, but also only valid under certain conditions.
If you have a real devices, maybe some measurements would give a quick crosscheck.
Regards,
Patrick
That figure is from the attached spice model due to kesko22. Do you have a spice model that agrees more closely with your measurements? Also attached is the ltspice file for extracting the gm and mu values.
The LU/LD1014 parameters are very variable. See here for measured data:
https://mbrennwa.github.io/curvetracedata/LD1014/parameters.html
Can SPICE work with parameter variations to consider the variability observed with real parts?
https://mbrennwa.github.io/curvetracedata/LD1014/parameters.html
Can SPICE work with parameter variations to consider the variability observed with real parts?
1mR small enough ?
https://www.buerklin.com/en/p/isabellenhuette/wirewound-resistors/pbv-r001-f1-1-0/40E600/
Patrick
https://www.buerklin.com/en/p/isabellenhuette/wirewound-resistors/pbv-r001-f1-1-0/40E600/
Patrick
I do not have an AC voltmeter than can measure the 10uV-50uV AC voltage across it? I suppose I could run the voltage through a suitable opamp circuit with a gain of 50x to 100x.
The highest transconductance I have seen in TO247 devices is ~7S at 1.3A.
SJEP120R100, and 2SK3497. 2SK3163 is maybe a touch higher.
So if someone can pack 27S into a DPAK size package, they must have done a brilliant job.
Patrick
SJEP120R100, and 2SK3497. 2SK3163 is maybe a touch higher.
So if someone can pack 27S into a DPAK size package, they must have done a brilliant job.
Patrick
I have an AD797 for that purpose :
https://www.diyaudio.com/community/...sfet-relay-for-loudspeaker-protection.370678/
Patrick
Interesting data. Do you know how those measurements were made?
They seem to suggest gm and mu values at about 1/2 what my spice model gives.
@lhquam That 5.1 Vds operating point in your measurement jig is not a good choice if you ask me. It may be fine for some parts, but for others it will be at the edge of the sane operating range. See attached example. I'd go with 3 or 4 Volts, which should be more universally applicable.
Attachments
https://www.passdiy.com/gallery/amplifiers/zen-variations-8
"In the case of JFET part LU1014, we note that with a gate voltage of –1 volt, the curve is concave below about 5 amps and 4 volts.
In this range it has that triode character, and this is the area of interest to us here."
.
"In the case of JFET part LU1014, we note that with a gate voltage of –1 volt, the curve is concave below about 5 amps and 4 volts.
In this range it has that triode character, and this is the area of interest to us here."
.
