Heres my lame attempt to transplant models from TINA to LTSpice.
I highly doubt I got this right, but who knows... Maybe I did???
*IRF9540 parameters from TINA-TI
.model IRF9540 PMOS(Level=3 Gamma=0 Delta=0 Eta=0 Theta=0 Kappa=0.2 Vmax=0 Xj=0
+ Tox=100n Uo=300 Phi=.6 Rs=64.15m Kp=10.14u W=1.5 L=2u Vto=-3.643
+ Rd=62.54m Rds=444.4K Cbd=2.021n Pb=.8 Mj=.5 Fc=.5 Cgso=1.024n
+ Cgdo=470.8p Rg=0.3371 Is=5.382E-17 N=1 Tt=140n)
*IRF540 parameters from TINA-TI
.model IRF540 NMOS(Level=3 Gamma=0 Delta=0 Eta=0 Theta=0 Kappa=0.2 Vmax=0 Xj=0
+ Tox=100n Uo=600 Phi=.6 Rs=21.34m Kp=20.59u W=.94 L=2u Vto=3.134
+ Rd=22.52m Rds=444.4K Cbd=2.422n Pb=.8 Mj=.5 Fc=.5 Cgso=1.144n
+ Cgdo=443.8p Rg=5.557 Is=2.848p N=1 Tt=142n)
Anyways, any power devices what don't find use as audio finals,
You can always put to good abuse in your linear supply...
Pretend its like junkyard wars... And this is the very best trash
that dude on the three wheeler could drag back...
I highly doubt I got this right, but who knows... Maybe I did???
*IRF9540 parameters from TINA-TI
.model IRF9540 PMOS(Level=3 Gamma=0 Delta=0 Eta=0 Theta=0 Kappa=0.2 Vmax=0 Xj=0
+ Tox=100n Uo=300 Phi=.6 Rs=64.15m Kp=10.14u W=1.5 L=2u Vto=-3.643
+ Rd=62.54m Rds=444.4K Cbd=2.021n Pb=.8 Mj=.5 Fc=.5 Cgso=1.024n
+ Cgdo=470.8p Rg=0.3371 Is=5.382E-17 N=1 Tt=140n)
*IRF540 parameters from TINA-TI
.model IRF540 NMOS(Level=3 Gamma=0 Delta=0 Eta=0 Theta=0 Kappa=0.2 Vmax=0 Xj=0
+ Tox=100n Uo=600 Phi=.6 Rs=21.34m Kp=20.59u W=.94 L=2u Vto=3.134
+ Rd=22.52m Rds=444.4K Cbd=2.422n Pb=.8 Mj=.5 Fc=.5 Cgso=1.144n
+ Cgdo=443.8p Rg=5.557 Is=2.848p N=1 Tt=142n)
Anyways, any power devices what don't find use as audio finals,
You can always put to good abuse in your linear supply...
Pretend its like junkyard wars... And this is the very best trash
that dude on the three wheeler could drag back...
Whist lookin for 2N5564 .model, I find post by AuroraB suggesting
that Erno Borberly liked to abuse National Semiconductor AH5020
as the complimentary matched pair... I don't know truth or fiction,
I'm just repeating a rumor, and poorly...
Since its an older chip, might be worth another trip to Tanner to
see if that one was also on the shelf and I just overlooked it?
Didn't see matched pairs of P-JFETs, easily recognized as such.
Doesn't mean they weren't there, staring me square in the face...
that Erno Borberly liked to abuse National Semiconductor AH5020
as the complimentary matched pair... I don't know truth or fiction,
I'm just repeating a rumor, and poorly...
Since its an older chip, might be worth another trip to Tanner to
see if that one was also on the shelf and I just overlooked it?
Didn't see matched pairs of P-JFETs, easily recognized as such.
Doesn't mean they weren't there, staring me square in the face...
Whoa.. Those N-JFET pairs are apparently VERY special parts.
5mV threshold matching, and zero temp coefficient at 4mA...
For .model I came up empty... You might take a crack at lifting
these parameters from TINA if you feel up to the challenge.
I'm tempted to start using TINA with all the default models it seems
to know. If only I could figure how to plug in my own??? Frustrating
enough just learning LTSpice. To start all over again, I'm not ready...
5mV threshold matching, and zero temp coefficient at 4mA...
For .model I came up empty... You might take a crack at lifting
these parameters from TINA if you feel up to the challenge.
I'm tempted to start using TINA with all the default models it seems
to know. If only I could figure how to plug in my own??? Frustrating
enough just learning LTSpice. To start all over again, I'm not ready...
Attachments
keantoken,
Do you know why the circuit you posted in post 113 works so well? It's because it has an extremely high open loop gain (over 1000), and the 3 db bandwidth is around 500 kHz. A closed loop system with a very high open loop gain will reduce any inherent distortion by a factor of about 1/OLG.
I have attached a revised version of your circuit that can be used to observe the open loop response (I deleted your models). A closed loop system will be unstable of there is more than 180 degrees of phase lag at the frequency where the OLG crosses 0 db (OLG of 1). Notice that the phase lag is about 168 degrees at that point. This is why you have so much peaking in the closed loop response. I would suggest you change the compensation caps t 100 pF.
Rick
Do you know why the circuit you posted in post 113 works so well? It's because it has an extremely high open loop gain (over 1000), and the 3 db bandwidth is around 500 kHz. A closed loop system with a very high open loop gain will reduce any inherent distortion by a factor of about 1/OLG.
I have attached a revised version of your circuit that can be used to observe the open loop response (I deleted your models). A closed loop system will be unstable of there is more than 180 degrees of phase lag at the frequency where the OLG crosses 0 db (OLG of 1). Notice that the phase lag is about 168 degrees at that point. This is why you have so much peaking in the closed loop response. I would suggest you change the compensation caps t 100 pF.
Rick
kenpeter said:
Dunno. Feel like breaking a world record?
How about this:
http://www.rmcybernetics.com/projects/DIY_Devices/homemade_jet_engine.htm
sawreyrw said:
Thank you for your information. I found the easiest option was to put a 100 ohm resistor in series with the bases of Q5 and Q6, and 470p caps across the B-C of the same. It is important to make the series resistors small to avoid distortion. Although I'm not quite sure how to measure the OLG in your schematic...
Looking back at the circuit, I have decided to revise a little. CFP outputs is a less stable and unnecessary option. By switching to Darlington, we have more Vce on the Allison transistors, AND better bandwidth. Distortion is usually affected very little by this change.
The 100 ohm resistors help with the peaking a lot. With the extra Vce I got from changing to darlington outputs, I put another pair of them on the collectors of Q6 and Q5. I didn't really think this would matter much - but it seems it does! Without further stability compensation, there is a large reduction in gain peaking at HF. I am learning more and more how useful collector resistors can be.
It seems that any capacitance across the B-E of the output transistors causes peaking, which is why the 100 ohm resistors are so beneficial. I don't know if this solves the problem completely.
- keantoken
keantoken,
Regarding the latest circuit you posted:
When LTspice runs an AC analysis, it first finds the operation point with all caps being open circuits and all inductor being shorts. So the AC analysis will be stable.
If you run an open loop analysis using the technique I previously showed you, you will see that the phase lag is greater than 180 degrees when the OLG is 0 db. This means the closed loop circuit will be unstable. The OLG is the gain from the sine wave source to V(Vout).
You can make the circuit stable by adding 1000 pF caps across both current sources, and the distortion is pretty good too. Look at the AC analysis as well; the phase lag is less than 180 degrees when the OLG is 0 db. Using AC analysis is a good way to see what your gain and phase margins are.
Rick
Regarding the latest circuit you posted:
When LTspice runs an AC analysis, it first finds the operation point with all caps being open circuits and all inductor being shorts. So the AC analysis will be stable.
If you run an open loop analysis using the technique I previously showed you, you will see that the phase lag is greater than 180 degrees when the OLG is 0 db. This means the closed loop circuit will be unstable. The OLG is the gain from the sine wave source to V(Vout).
You can make the circuit stable by adding 1000 pF caps across both current sources, and the distortion is pretty good too. Look at the AC analysis as well; the phase lag is less than 180 degrees when the OLG is 0 db. Using AC analysis is a good way to see what your gain and phase margins are.
Rick
keantoken said:
sawreyrw said:
It is also best to run pulse analysis, ie; square wave. Getting it
stable is one thing but optimising the step response is the real
challenge. This is an area where Ltspice can differ significantly
from reality due to all the parasitics. However it will tell you places
that require attention. Ultimately the proto will need some tuning
up IME. I have also found real life circuits often to be more stable
than the sim would indicate.
It is also worth considering, with local FB unity gain buffers circuits
like this, there is often a minimum value of source impedance where
instability ensues, so a small value of R may be required at the IP.
cheers
Terry
Terry Demol said:
Yes, I have found real-life circuits are more stable than in the simulator, strangely.
The circuit does tend to oscillate with specific source impedances, but I am hoping to get a circuit that will be happy with any source impedance. It has very high input impedance, so a series 100 or even 1k resistor could be fine, but with added compensation the input impedance will rapidly roll off.
Thank you for your advice. I will go do it now.
- keantoken
How about this?
Sawrey's solution with the 1n caps worked pretty well and wasn't affected badly by source impedance. The main drawback of this approach however is that the input impedance suffers rapidly beginning at 1kHz. So I tried to find a way to lower the values of these caps.
It doesn't oscillate anymore with different load impedances.
- keantoken
Sawrey's solution with the 1n caps worked pretty well and wasn't affected badly by source impedance. The main drawback of this approach however is that the input impedance suffers rapidly beginning at 1kHz. So I tried to find a way to lower the values of these caps.
It doesn't oscillate anymore with different load impedances.
- keantoken
Attachments
I've put mine across the collector to base of what would be Q5
and Q6 in your drawing. Plus a small base resistor makes a pair
of low pass filters that dumbs out the error correction for higher
freqs where the oscillations seem to occour.
I'm guessing the loop feedback needs attenuated less than 1 for
all frequencies above where feedback phase inverts twice???
Of course, any such filter changes the phase too. I'm none too
clear how the math of all that phasery comes together. Just play
reactance tweakage with the sim till stability seems bulletproof.
and Q6 in your drawing. Plus a small base resistor makes a pair
of low pass filters that dumbs out the error correction for higher
freqs where the oscillations seem to occour.
I'm guessing the loop feedback needs attenuated less than 1 for
all frequencies above where feedback phase inverts twice???
Of course, any such filter changes the phase too. I'm none too
clear how the math of all that phasery comes together. Just play
reactance tweakage with the sim till stability seems bulletproof.
We've all been doing this the wrong way I think!
Look at this.
The 1n caps suggested by Sawrey are effective but put an unnecessary amount of strain on the Allison transistors. I knew there had to be some far more effective way of doing this and I think I found it - no B-C capacitors which don't seem to do the job well. In this solution, the compensation cap is only affective if against oscillation, and doesn't interfere with the signal path.
Looky here:
1: Compensation components are C4, R19 and R20.
2: I traded BC5xx transistors for the input devices - we now have an input impedance of about 21 megaohms! I knew of this before but was bleary of how it might affect the HF characteristics.
3: Yes, I know C1 is 680fF. Any tweaking of this will have to be in real life to take into account PCB stuff - we could easily make this capacitor directly out of the PCB board rather than trying to buy such a thing.
4: These CCSs are very special: output impedance is around 60 megaohms with good HF characteristics. I have had them oscillate in some weird situations though (I'm sure I was abusing them).
I have made a model of the input impedance - this scraps my Tektronix with its 2megaohm+47pF input impedance. If these figures can be attained in real life, then cable and probe parasitics will be much more important.
I would use this up to 1MHz, and gain rolls off around 10MHz.
Distortion is still low enough at 1MHz for an oscilloscope display.
- keantoken
Look at this.
The 1n caps suggested by Sawrey are effective but put an unnecessary amount of strain on the Allison transistors. I knew there had to be some far more effective way of doing this and I think I found it - no B-C capacitors which don't seem to do the job well. In this solution, the compensation cap is only affective if against oscillation, and doesn't interfere with the signal path.
Looky here:
1: Compensation components are C4, R19 and R20.
2: I traded BC5xx transistors for the input devices - we now have an input impedance of about 21 megaohms! I knew of this before but was bleary of how it might affect the HF characteristics.
3: Yes, I know C1 is 680fF. Any tweaking of this will have to be in real life to take into account PCB stuff - we could easily make this capacitor directly out of the PCB board rather than trying to buy such a thing.
4: These CCSs are very special: output impedance is around 60 megaohms with good HF characteristics. I have had them oscillate in some weird situations though (I'm sure I was abusing them).
I have made a model of the input impedance - this scraps my Tektronix with its 2megaohm+47pF input impedance. If these figures can be attained in real life, then cable and probe parasitics will be much more important.
I would use this up to 1MHz, and gain rolls off around 10MHz.
Distortion is still low enough at 1MHz for an oscilloscope display.
- keantoken
Attachments
Here is a higher voltage version.
It also has much greater input impedance due to a slight change in the CCSs. There is little chance you will EVER find a CCS as effective as this one, unless you use an opamp. The key to tweaking this CCS is in the 680 ohm resistors. Output rejection of below -150db can be attained by tweaking this resistor. (properly tweaked the output impedance will continually drop the lower the test frequency, phase will be 90 degrees across spectrum, and output will be 3rd harmonics with no fundamental)
The input impedance is now about 1.4Gohms at frequencies below 100Hz! The CCSs used set the limit on input impedance; these are the last things you can tweak assuming all other factors are taken care of.
These are simulated figures, of course.
- keantoken
It also has much greater input impedance due to a slight change in the CCSs. There is little chance you will EVER find a CCS as effective as this one, unless you use an opamp. The key to tweaking this CCS is in the 680 ohm resistors. Output rejection of below -150db can be attained by tweaking this resistor. (properly tweaked the output impedance will continually drop the lower the test frequency, phase will be 90 degrees across spectrum, and output will be 3rd harmonics with no fundamental)
The input impedance is now about 1.4Gohms at frequencies below 100Hz! The CCSs used set the limit on input impedance; these are the last things you can tweak assuming all other factors are taken care of.
These are simulated figures, of course.
- keantoken
Attachments
I've just read that reverse biasing VBE into reverse breakdown
(like a 5 or 6 volt Zenier) will quickly and seriously degrade HFE.
It might be worthwhile have reverse pass diodes to assure that
can never happen here, if input is driven futher than the output
(shorted, back EMF'd, or clipped?) can follow.
I also wonder if this abuse can be deliberately applied to a single
emitter junction transistor to tweak a match? I've never heard of
anyone doing it on purpose, other than as an el-cheapo Zenier
substitute that will never again see service as amplifying device...
(like a 5 or 6 volt Zenier) will quickly and seriously degrade HFE.
It might be worthwhile have reverse pass diodes to assure that
can never happen here, if input is driven futher than the output
(shorted, back EMF'd, or clipped?) can follow.
I also wonder if this abuse can be deliberately applied to a single
emitter junction transistor to tweak a match? I've never heard of
anyone doing it on purpose, other than as an el-cheapo Zenier
substitute that will never again see service as amplifying device...
kenpeter said:
Good point.
I suspect that even if we can tweak Hfe values this way, I don't think it will have a good effect on the device curves. If it could be done this way, that's probably how labs would do it (I don't think they do, but what do I know).
I will look into overinput protection.
- keantoken
It is true that beta is degraded, if the base emitter junction is subjected to reverse breakdown; however, the degradation is not significant so long as the reverse current (power) is not too large. In any case, reverse breakdown should be avoided. Having said that, I don't think reverse breakdown will occur in your Allison circuits.
Manufacturers of discrete devices do not tweak the device beta, after the devices have been fabricated. ICs can be "trimmed", but that involves changing the values of resistors lazing away some of the resistive element.
Manufacturers of discrete devices do not tweak the device beta, after the devices have been fabricated. ICs can be "trimmed", but that involves changing the values of resistors lazing away some of the resistive element.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.