That s not true , a given NPN is not modeled with the same
parameters values than its supposed PNP symetrical.
Same thing with fet/mosfets devices.
This blatlanly appears in simulators.
MODEL FORMAT: PSpice
.MODEL Qmjl1302aOS pnp
+IS=7.56722e-11 BF=148.663 NF=1.0861 VAF=33.2962
+IKF=6.03056 ISE=7.00157e-12 NE=3.37477 BR=3.57438
+NR=1.38503 VAR=2.45693 IKR=7.65272 ISC=4.75e-13
+NC=3.96875 RB=3.55494 IRB=0.217628 RBM=0.1
+RE=0.000905557 RC=0.0997087 XTB=1.30215 XTI=1
+EG=1.206 CJE=1.62625e-08 VJE=0.99 MJE=0.496165
+TF=2.98166e-09 XTF=1000 VTF=3.58167 ITF=546.327
+CJC=1.72466e-09 VJC=0.814009 MJC=0.457122 XCJC=1
+FC=0.207567 CJS=0 VJS=0.75 MJS=0.5
+TR=1e-07 PTF=0 KF=0 AF=1
MODEL FORMAT: PSpice
.MODEL Qmjl3281aOS npn
+IS=8.26882e-10 BF=97.5769 NF=1.27505 VAF=63.6913
+IKF=7.75466 ISE=7.44075e-12 NE=3.39947 BR=0.1
+NR=1.6501 VAR=7.8831 IKR=15.9415 ISC=4.16531e-13
+NC=3.98207 RB=1.0642 IRB=2.65716 RBM=1.0642
+RE=0.0001 RC=0.051212 XTB=0.82132 XTI=1
+EG=1.206 CJE=1.10527e-08 VJE=0.677092 MJE=0.394732
+TF=1.93853e-09 XTF=1000 VTF=4.32084 ITF=149.92
+CJC=1.00256e-09 VJC=0.924803 MJC=0.524843 XCJC=0.100008
+FC=0.8 CJS=0 VJS=0.75 MJS=0.5
+TR=1e-07 PTF=0 KF=0 AF=1
It is often used for learning purposes. Like, after students learn to dive school promises to fill the pool with water.
Hi,
Read differentials of same polarity devices, current mirrors and so on...
Ciao T
Read differentials of same polarity devices, current mirrors and so on...
Ciao T
I resisted sims for a long time and designed very complex systems without their aid, with plenty of "hand" calculation and prototyping. But when a technician, having breadboarded some of them, told me that a respected engineer's filter designs all oscillated, and that this showed up in transient response simulations, I realized the time had come. The engineer had made a sign error, and this did not show up on frequency response plots. He was also in a hurry, having resigned to go to another company, and had spent most of his time on optimization of certain aspects of his filters, little on testing them in simulation. And these were tricky Friend topologies realizing generalized biquads --- without a lot of experience it's difficult to assess them at sight.
As has been discussed extensively, simulations are no substitute for insight and experience. But come to them with the latter and they can be extremely helpful.
One of the features lacking from most simulators is the explicit thermal modeling of semiconductors, including signal-induced self-heating shifts. Oscilloscope vertical amplifier designers learned ways of dealing with these back in tube days, and needed to make the fastest possible circuits while avoiding strange shifts in offset and gain. In many audio applications the chore is much easier if local/global feedback is allowed.
For some parts the thermal time constants are fairly long. For smaller bipolars they can be quite short. And when you take into account the temperature dependence of base-emitter voltages and beta, about -2mV/degree K and about 0.5%/degree K around room temperature, you can see how quickly this will disrupt things if not taken into account.
People often speak of bipolars with "linear" beta (by which they really mean constant beta as a function of collector current). But if this constant beta is a function of chip temperature and the latter is changing significantly with signal, there goes your simulation accuracy if this is not modeled explicitly.
Brad
Brad,
I can agree to many of your points. It is very difficult if not impossile to completely sim a circuit to the last detail, but often that is not needed.
I use Spice as a 'proof of concept' to check out my (sometimes crazy) ideas. Is the idea I have halfway sensible, or is there a basic flaw in my reasoning that I missed?
And the easy with which you can do 'measurements' on your circuit to really find out what's going on, measurements that are just not possible in practise or would require a multi-million dollar lab.
I would hate to have to build each and everyone of them to check them out.
With Spice, I can quickly weed out the nonsense ideas and converge on the promising ones.
THEN, of course, it is time to prototype and sometimes you find out that your idea was fine but that practicalities make it unattractive.
But, Spice for me is a great time saver and insight generator and I wouldn't want to miss it.
jan
I can agree to many of your points. It is very difficult if not impossile to completely sim a circuit to the last detail, but often that is not needed.
I use Spice as a 'proof of concept' to check out my (sometimes crazy) ideas. Is the idea I have halfway sensible, or is there a basic flaw in my reasoning that I missed?
And the easy with which you can do 'measurements' on your circuit to really find out what's going on, measurements that are just not possible in practise or would require a multi-million dollar lab.
I would hate to have to build each and everyone of them to check them out.
With Spice, I can quickly weed out the nonsense ideas and converge on the promising ones.
THEN, of course, it is time to prototype and sometimes you find out that your idea was fine but that practicalities make it unattractive.
But, Spice for me is a great time saver and insight generator and I wouldn't want to miss it.
jan
Absence of SPICE trained well my ability to build something in my imagination. Punishment was quite strong: death of scarce and expensive components when something was wrong or stupid; loss of fruits of many hours of work for nothing, boredom if solution was too complex and suboptimal. That's why I don't start prototyping until I can imagine everything in imagination, except details that I want to check and prove by experiment.
For a good deal of recent work I often go to layout directly from design and simulations. But since I don't do layout myself I need to coach the folks and proof things a lot, especially if there is no netlist link. A while back a challenging hypercompact amplifier bristling with features (the proverbial ten pounds of x in a five pound package) required four boards, and the far east manufacturer was switched at the last moment. There was the capability of linking CadInt schematics to layout by the intermediary drafter and the layout person, but it was bypassed (!) in the heat of battle with a show deadline looming. I got the first article a day before the show and it was already obvious that there were hideous errors as well as assembly defects, poor soldering, etc. Needless to say I didn't try to turn it on, particularly as I noticed among other things that the polarity of the d.c. power connecter was backwards
But the layout guy said Why should I fix the layout if we don't know that the circuit works?
So at that point I wished that I had indeed breadboarded the thing. And much as I like work, I'm not regretting that I don't do any for that client anymore.Brad
The drawback of doing so much work at the computer and less at the bench is I sit on my behind too much, which has recently been shown to shorten one's life appreciably, with even significant exercise at other times not adequate compensation. So when a sim is running I try to get up and move around
45 years ago, I was simulating electronic circuits. It was all new, and my company was even written up in engineering magazines as being a pioneer in the field. Guess what? On one occasion, a SLIDE RULE proved that the results were wrong. Did my boss and I get into trouble! However, we had done nothing wrong, it was the IBM program that had the weakness in worst case computing.
I had to create the electronic parts models myself from my own measurements with a curve tracer. This is a great way to learn about how the 'ideal' models drift from reality.
I then changed jobs and moved to Ampex, where we did not depend on computer simulations, but did what we needed to with mechanical calculators, and hand calculation. It worked OK.
A few years later, I was there when SPICE was being developed by Dr. Don Pederson and others for LBL, on the hill above us. I saw some fantastic emulations of the uA741, etc, but nothing really new. A couple of years after that (1973) I tried to get a distortion analysis of one of my complementary differential jfet input designs. The answers were all off, so why bother? My trusty HP-35 gave me the calculations that I needed, anyway. Everything else was cut and try. Different manufacturer's jfets with the same part number gave different measurements, sometimes, go figure!
Finally, home computers came of age (15 years later) and I bought one, however, it did NOT have double precision, AND no really good engineering programs would work on an Apple II for example, without custom programing.
The real weak spot in analog design, at the time, was analog filters. They could be done by hand calculation, but what a BEAR it was, and maybe one might make a mistake along the way, losing everything. I was saved a lot of problems by someone giving me a modified program (on cassette tape) that could do a pretty good AC analysis on the Apple. What a life saver!
Over the years, I got a MAC, and early emulation programs like Microcap. Spice was also available, and potentially very good, but it was difficult to use, compared to Microcap, so I wallowed in MC for years.
Come the mid-90's and finally Spice, etc had come of age, and relatively easy to use, BUT the development companies then ignored MAC, seeing that it became a minor market. So even though Spice had became of age, I just did not need it to develop the Parasound amps and preamps. If I had a problem, I could use the very old Microcap program that I had. Today, I still do not use Spice as I would like, sometimes, but it is relatively unnecessary anyway, because I have so much experience working around it. That may not go for many of you here. If you can, learn and use Spice. It will save you time and energy in getting 'up to speed' and that is sorely needed here. '-)
I had to create the electronic parts models myself from my own measurements with a curve tracer. This is a great way to learn about how the 'ideal' models drift from reality.
I then changed jobs and moved to Ampex, where we did not depend on computer simulations, but did what we needed to with mechanical calculators, and hand calculation. It worked OK.
A few years later, I was there when SPICE was being developed by Dr. Don Pederson and others for LBL, on the hill above us. I saw some fantastic emulations of the uA741, etc, but nothing really new. A couple of years after that (1973) I tried to get a distortion analysis of one of my complementary differential jfet input designs. The answers were all off, so why bother? My trusty HP-35 gave me the calculations that I needed, anyway. Everything else was cut and try. Different manufacturer's jfets with the same part number gave different measurements, sometimes, go figure!
Finally, home computers came of age (15 years later) and I bought one, however, it did NOT have double precision, AND no really good engineering programs would work on an Apple II for example, without custom programing.
The real weak spot in analog design, at the time, was analog filters. They could be done by hand calculation, but what a BEAR it was, and maybe one might make a mistake along the way, losing everything. I was saved a lot of problems by someone giving me a modified program (on cassette tape) that could do a pretty good AC analysis on the Apple. What a life saver!
Over the years, I got a MAC, and early emulation programs like Microcap. Spice was also available, and potentially very good, but it was difficult to use, compared to Microcap, so I wallowed in MC for years.
Come the mid-90's and finally Spice, etc had come of age, and relatively easy to use, BUT the development companies then ignored MAC, seeing that it became a minor market. So even though Spice had became of age, I just did not need it to develop the Parasound amps and preamps. If I had a problem, I could use the very old Microcap program that I had. Today, I still do not use Spice as I would like, sometimes, but it is relatively unnecessary anyway, because I have so much experience working around it. That may not go for many of you here. If you can, learn and use Spice. It will save you time and energy in getting 'up to speed' and that is sorely needed here. '-)
Last edited:
I recall Bob Pease railing against simulation many times, but it's a tool like any other, and as recent posts allude to, to use it effectively one needs to know its limitations as well as its capabilities. Too many people don't learn about the limitations, or worse don't even think it has limitations.
Even where it gives good and accurate answers (I'm thinking of passive speaker crossovers, especially if you have a better model for the speaker than a resistor), you still want to hear the result, and likely compare several designs before deciding on one.
Even where it gives good and accurate answers (I'm thinking of passive speaker crossovers, especially if you have a better model for the speaker than a resistor), you still want to hear the result, and likely compare several designs before deciding on one.
There's a tool for that too, those newfangled treadmill desk things. Then you can surf over here while some long task is running.The drawback of doing so much work at the computer and less at the bench is I sit on my behind too much, which has recently been shown to shorten one's life appreciably, with even significant exercise at other times not adequate compensation. So when a sim is running I try to get up and move around
Getting there. Trying to add everything up. Some of it does not make sense like the seemingly very low f3 of R11/C3 but R14/C7 zero is 100 times higher. Here is where looking at this specific design may be catching up to me.
Lets see..
Would the Q of the pole C3/R11 be the same as seen by Q10 and Q11, or does Q8 make that a wash.
As the output stage has less than unity gain, it needs no compensation. It's pole is set by the gate resistors as per the application notes I mentioned a bit back. I noticed the later versions of his amps were right in line with those numbers.
I sure wish I could afford a good 1M current probe to see how close reality in circuit is compared to the Bode plots. I have no doubt the transistors and parastatics do not behave as nicely as the textbook suggests. I think I can measure up to 40K or so with SoundEasy as I have a 96K sound card.
The frequencies and phase is starting to make sense, the Q's don't. I figure I'll have it when I can visualize it. I am a very visual thinker. Guess some bench square wave testing is in order. Replacement transistors came today, so matching time.
Lets see..
Would the Q of the pole C3/R11 be the same as seen by Q10 and Q11, or does Q8 make that a wash.
As the output stage has less than unity gain, it needs no compensation. It's pole is set by the gate resistors as per the application notes I mentioned a bit back. I noticed the later versions of his amps were right in line with those numbers.
I sure wish I could afford a good 1M current probe to see how close reality in circuit is compared to the Bode plots. I have no doubt the transistors and parastatics do not behave as nicely as the textbook suggests. I think I can measure up to 40K or so with SoundEasy as I have a 96K sound card.
The frequencies and phase is starting to make sense, the Q's don't. I figure I'll have it when I can visualize it. I am a very visual thinker. Guess some bench square wave testing is in order. Replacement transistors came today, so matching time.
John, I love using my sim software. The problem is, I rarely seem to actually build many circuits anymore because simulating them is so quick and easy. Too often, though, I think I move on to simulating the next creative design variation which strikes me (and there always seems to be one) without following through on building and actually LISTENING to previous circuit which have appeared worthy in simulation.
Last edited:
Exactly right, I wasn't referring to complementary symmetry (NPN/PNP) topologies, but rather to uni-polarity topologies.
Although, I have modeled complementary symmetry topologies using the Toshiba 2240/970 pair and obtained what at first appeared to be outstanding open-loop THD results. Only to modestly alter a few parameters of the device models and disappointedly watched the resulting THD climb substantially in the sim.
Last edited:
JC, interesting what you wrote; sims (use LTspice to get rough ideas) still don't work for analog filters, at least not at low frequencies. Calculations for me are faster and just as precise. The rest comes up to building the sh*t, to measure and see what comes out, mod etc.
On an other topic, I see there was a short spat in this thread on capacitors again. I fell in love with NGO. Not so much because they sound better than polystyreen, to me they sound the same, but you can put them in a frying pan without degradation. In other words, soldering will do them no harm. Plus they are current. And small. Anyone thoughts/experience on this?
On an other topic, I see there was a short spat in this thread on capacitors again. I fell in love with NGO. Not so much because they sound better than polystyreen, to me they sound the same, but you can put them in a frying pan without degradation. In other words, soldering will do them no harm. Plus they are current. And small. Anyone thoughts/experience on this?
SY,
I can remember the chip design verification engineers - back when I was employed for an major IC design and mfg. house - running Monte Carlo analysis on their Sun workstations. They would begin an analysis during the day and leave it run all night long, checking the results the next morning. I always found the "luck" aspect of this aptly named analysis fascinating.
I've not seen any frequency-dependent difficulties with Spice filter analysis --- perhaps you could provide an example. Although I agree that with relatively simple topologies (Sallen-Key for example) they are easy to design without simulation.
As for CG0/NP0, other than poor volumetric efficiency, and high cost for large values, most have seemed to be quite adequate. I remember reading about one made for the International Space Station. IIRC, 270uF (yes microfarads!), very low inductance, one big mother, about the size of a small loaf of bread. Don't ask the price. It was for a key switching regulator.
I specified an NP0 recently for a non-audio product because of the requirement for a low temperature coefficient of capacitance. The largest one I could readily find in surface mount was 22nF, which worked out to be about right for the application. Unfortunately the tempco of the resonant inductor's loss turned out to be quite peculiar, but I will resist further off-topic excursions
Brad
- Status
- Not open for further replies.