Hey PB2, carry on like tha wind. Personally I lookin forward to your work. wahabby is bringing u down like a buzz kill. Don't let neh-sayers in yer way. Full speed, Full strength. Love everybody and we get to ZEN. More Zen Please...
I propose we understand what is actually built before we start changing things willy-nilly ie try to get the model as close to actual HW. Verify the model with lab testing. After that we could add small improvements with possibly different levels of mods.
level 1) a few component changes (no PCB changes)
level 2) add a few sub-circuits plus part changes (minor PCB mods)
level 3) brand new circuits including new PCB + part kit (using same PS and Heatsinks) maybe for level 3 use a National chipamp front end here?
level 1) a few component changes (no PCB changes)
level 2) add a few sub-circuits plus part changes (minor PCB mods)
level 3) brand new circuits including new PCB + part kit (using same PS and Heatsinks) maybe for level 3 use a National chipamp front end here?
I agree, in fact if I post the model with the mirror etc. it will be in a new thread with mod in the title. Let's keep this thread for analysis of the stock amp. Perhaps with just suggestions for component value or transistor changes - if needed.
Last edited:
I had some comments concerning Andy's models here and I never got around to try the fixes that I mentioned:
http://www.diyaudio.com/forums/software-tools/101810-spice-simulation-45.html#post1363190
The data sheet that I referenced was preliminary, and they took out the data points that I referred to in the final version. I do still have the old one.
I have a data sheet here for the 2SB554 that lists ft as 6 MHz and Cob as 460 pF.
These data could be used with this to adjust some known good models:
From: http://www.diyaudio.com/wiki/Creating_Spice_Models
The BJT default in ISSPICE4 gives first order, Ebers-Moll, DC
parameters; but it does not provide parameters for Transient or
AC analysis. The following default can be used when minimal
data sheet specifications are available by using the PARAM
program to evaluate the equations in curly braces. This model
will create a good transistor model from virtually any data sheet.
The parameters you must specify are:
IMAX Maximum collector current
COB Collector-base capacitance
FT Gain bandwidth product in Hz
TS Storage time
The equation-based model is shown below:
.MODEL NBJT NPN (CJC={2.2*COB} TF={.16/FT} TR={1.7*TS}
+CJE={7*COB} RC={.5/IMAX} VAF=100 IKF={.7*IMAX}
+ IS={2E-15*MAX} )
It would be excellent to have a simple way to quickly come up
with new models.
http://www.diyaudio.com/forums/software-tools/101810-spice-simulation-45.html#post1363190
The data sheet that I referenced was preliminary, and they took out the data points that I referred to in the final version. I do still have the old one.
I have a data sheet here for the 2SB554 that lists ft as 6 MHz and Cob as 460 pF.
These data could be used with this to adjust some known good models:
From: http://www.diyaudio.com/wiki/Creating_Spice_Models
The BJT default in ISSPICE4 gives first order, Ebers-Moll, DC
parameters; but it does not provide parameters for Transient or
AC analysis. The following default can be used when minimal
data sheet specifications are available by using the PARAM
program to evaluate the equations in curly braces. This model
will create a good transistor model from virtually any data sheet.
The parameters you must specify are:
IMAX Maximum collector current
COB Collector-base capacitance
FT Gain bandwidth product in Hz
TS Storage time
The equation-based model is shown below:
.MODEL NBJT NPN (CJC={2.2*COB} TF={.16/FT} TR={1.7*TS}
+CJE={7*COB} RC={.5/IMAX} VAF=100 IKF={.7*IMAX}
+ IS={2E-15*MAX} )
It would be excellent to have a simple way to quickly come up
with new models.
Last edited:
Too bad there isn't a hybrid PI model which would be at a level somewhere between this one and the Gummel-Poon. This 1st order BJT model without AC analysis is not-too useful me thinks? I'm crippled using the Gummel-Poon model ie way too detailed and from a physical process point of view that is very painful to extract from a normal data sheet.
Hi infinia,
I agree with your other points, but I'd like to add that prototype testing (lab tests) must include the severe conditions common to what the market will subject these amplifiers to. Then there are our "friends" at CSA Laboratories who devise various unnecessary tests that seem to have destruction of equipment as their only goal. The line cord drop test was one of my favorite examples of needless destruction.
What bothers me about an unbalanced diff pair is the extra distortion this creates. Many designers use a DC servo that achieves low DC offset by forcing the diff pair to be unbalanced. This works against one of the reasons why a differential pair is used in the first place. This is probably why designers and arm chair critics express a dislike for a DC servo circuit (Luxman used to call it "Duo-beta"). A better way to achieve the same result is used in the BGW 750C amplifier (and others).
Anyway, I still consider a simulation to produce starting values for the circuit design, and as a canary to alert us to serious problems. When you have a prototype working as designed, that's when you know the design does really work.
-Chris
I'm trying to figure out what you are getting at. I am primarily a service technician, but I also do design work. From my experience over the years, it's pretty common to run into a technician who has designed "the best amp out there" using some "secret" tricks and optimizations. These normally turn out to be okay at the best, and extremely unreliable at the worst. Insta-smoke.
Completely agree, from an experience perspective with this line and these amplifier models from new to now.
Simulations using less expensive programs and low computing power give the designer a guide only. The simulation is only as good as the information you feed into it, and often the basic parameters have some errors. If simulation did work really well, I don't think we would see many ECOs except for component changes due to supply changes. My experience in the field shows that this is not the case. Could simulation ever be close enough to the real world to put these designs into production without any prototypes? Maybe, with enough time and effort directed in that direction, but the money simply isn't there. Also, we often see problems at frequencies well above the system cut-off frequency. Individual parts respond to very high frequencies and may unexpectedly operate in their active regions if they aren't frequency limited beyond what compensation exists in the overall feedback loop.
I agree with your other points, but I'd like to add that prototype testing (lab tests) must include the severe conditions common to what the market will subject these amplifiers to. Then there are our "friends" at CSA Laboratories who devise various unnecessary tests that seem to have destruction of equipment as their only goal. The line cord drop test was one of my favorite examples of needless destruction.
This is a trait that many amplifier designs share. It shouldn't be too bad since there is a current source in the tail. What is your simulation showing? Another point to consider is that this amplifier takes a long while to reach thermal equilibrium.
What bothers me about an unbalanced diff pair is the extra distortion this creates. Many designers use a DC servo that achieves low DC offset by forcing the diff pair to be unbalanced. This works against one of the reasons why a differential pair is used in the first place. This is probably why designers and arm chair critics express a dislike for a DC servo circuit (Luxman used to call it "Duo-beta"). A better way to achieve the same result is used in the BGW 750C amplifier (and others).
Anyway, I still consider a simulation to produce starting values for the circuit design, and as a canary to alert us to serious problems. When you have a prototype working as designed, that's when you know the design does really work.
-Chris
Hi Pete,
I'll look for those posts in case another moderator did the deed already.
-Chris
Edit: Hi Pete,
Can you please check my work on those posts you wanted moved? Jobs done.
I'll look for those posts in case another moderator did the deed already.
Please let us know what your findings are either way then, but it should be possible to carry on a discussion on spice here as well. Moderators have already spoken on the recent issues in this thread.
I am interested. You are probably correct. When the output is shorted, the feedback loop is broken and no longer operates normally. In this situation, the designers probably called it a "don't care" situation and relied on the fuses to control the problem.
I'm thinking that a new thread that covers your alterations to this amplifier might be the way to go.
Yup. Thank you.
Now that will be helpful to anyone using a spice program! Is this what Andy was using?
-Chris
Edit: Hi Pete,
Can you please check my work on those posts you wanted moved? Jobs done.
Last edited:
Hi wahab,
I think your comments are inaccurate at best, just plain wrong.
Consider the age of these amplifiers, and the numbers that were sold. I have been directly involved with servicing these from the time they were brand spanking new until now and continuing forward. The only two problems I can accept for these amplifiers are not due to their design. They have the leaking capacitors, something that many products suffered from. The other major problem we see are the masses of individuals who perform work on these amplifiers improperly. Those are often the posts you may see on web sites. There are many amplifiers around due to the low cost of the original product, coupled with the high performance. The best amplifier out there? No, they were not intended to be the pinnacle of design. They were intended to offer a very high level of performance at reasonable dollars. They did succeed in that. That means there are many around that become projects for improvements.
-Chris
All amplifiers are device dependent to some degree.
I think your comments are inaccurate at best, just plain wrong.
Consider the age of these amplifiers, and the numbers that were sold. I have been directly involved with servicing these from the time they were brand spanking new until now and continuing forward. The only two problems I can accept for these amplifiers are not due to their design. They have the leaking capacitors, something that many products suffered from. The other major problem we see are the masses of individuals who perform work on these amplifiers improperly. Those are often the posts you may see on web sites. There are many amplifiers around due to the low cost of the original product, coupled with the high performance. The best amplifier out there? No, they were not intended to be the pinnacle of design. They were intended to offer a very high level of performance at reasonable dollars. They did succeed in that. That means there are many around that become projects for improvements.
-Chris
It's just a observation that I have seen over the years and was happening on this and the repair thread. Engineers that automatically go into design mode changing circuits ie improvements when doing a repair. And some techs (not you Chris) doing design work that may be a little beyond their foundation.
I don't think I would EVER propose simulation without any lab work at least not in my lifetime. Simulation is just another tool and I would suggest a few hand calculations done as a preliminary check. Use the software in iteration process to optimize the design in a few areas. Stability analysis is an area in which software is ideal. In fact LTspice was developed mainly as a tool for simulating control loops for LTC switch mode chips ie SwitcherCad. LOL
I agree but I think this was NOT seen as a product for abusive PRO environments. It was a value design (put the bucks where it matters) for Home HiFI. I'm sure some of those audiophile folks would object to protection circuits getting in the way of the sound. LOL
FWIW I've put a product thru UL Listing before. I got stories too.
almost a tenth of a volt, That Temco analysis is difficult stuff indeed, esp on this one with a thermistor only on one drivers side.
I'm not a big fan of AC integrators used for DC correction in audio amps either. I believe in the KISS principal. Adding more circuits to solve one thing usually creates more problems in other places.
I like matched pairs built on the same substrate no hand matching.
So far, i did remade the sims using pb2 schematic , and the circuit
is instable only when there s a capacitive load added, from 20nF to
200nf in paralel with a 4R load..;
Instability occur around the zero crossing point with a sine wave
of 2V pp at the input.
I hadn t the time to go further, i ll recheck all this as soon as
ive got a better understanding of LTspice..
is instable only when there s a capacitive load added, from 20nF to
200nf in paralel with a 4R load..;
Instability occur around the zero crossing point with a sine wave
of 2V pp at the input.
I hadn t the time to go further, i ll recheck all this as soon as
ive got a better understanding of LTspice..
wahab you only seem to be posting your results. I think until you get a model which reflects reality, it would be a useful first step to propose your methods of determining stability. The universal method for control systems engineering using negative feedback is showing loop gain and testing w/ step function as an input or output stimulus . Remember there are no RFI output filters on this Adcom so capacitive loads right at the output terminal are going to be problematic and IMO not a realistic load. FWIW output loads that are highly capacitive are not representing any speakers that I've ever come across. Maybe silly exotic speaker cables non withstanding. Even electrostatic panels have some series loss element at their inputs.
Agree, usualy, i use a step function, but since it was related
that the amp has perhaps current issues , i did sim it with a sine
to check if there was cross conduction, but it s not the case.
When using a half step, that is, an input step that goes from 0
to 1V and then return to 0V, oscillaton occur only when the
step travel at the 0V point, that s why i checked with a sine,
to see if there s a problem around the 0V crossing point.
No current spikes so far but oscillation occuring when
the sine is around 0V...
I will try with other models as output devices , with
Fairchild devices which are reputated to have honnest
pspice modellings..
Spice simulation? It is not mature...
I tried LTSpice in the past day. I tried it in the simple model of the discrete op-amp that i am building. I had no problem to learn the LTSpice in one day, because my previous experience with EDSpice simulator of EDWinXP. I had the opportunity of comparison between the two Spice simulation engines. The only difference it is that LTSpice offers possibility of directly representation of FFT analysis in plot, instead EDSpice offers FFT analysis only in text mode which must be exported in Microsoft Excel to be converted in plot. I have as well, the circuit assembled in a veroboard and i can do comparisons of measurements with real instruments like the DSO and the Virtins MI FFT analyzer. The results between Spice analysis and real instruments are in disagreement of 30% at least.
This is due to inaccurate models given from constructors, or from that the Spice engine it supposes the given model of semiconductor as is, instead in real world the component has enough inclination from the model.
For example, Zetex gives the same model for BC547 and BC550.
The LTSpice software it is very sophisticated, but the models no. Except those models of Linear Technology devices.
Fotios
I tried LTSpice in the past day. I tried it in the simple model of the discrete op-amp that i am building. I had no problem to learn the LTSpice in one day, because my previous experience with EDSpice simulator of EDWinXP. I had the opportunity of comparison between the two Spice simulation engines. The only difference it is that LTSpice offers possibility of directly representation of FFT analysis in plot, instead EDSpice offers FFT analysis only in text mode which must be exported in Microsoft Excel to be converted in plot. I have as well, the circuit assembled in a veroboard and i can do comparisons of measurements with real instruments like the DSO and the Virtins MI FFT analyzer. The results between Spice analysis and real instruments are in disagreement of 30% at least.
This is due to inaccurate models given from constructors, or from that the Spice engine it supposes the given model of semiconductor as is, instead in real world the component has enough inclination from the model.
For example, Zetex gives the same model for BC547 and BC550.
The LTSpice software it is very sophisticated, but the models no. Except those models of Linear Technology devices.
Fotios
Thank you for the numbers, Fotios..
Since the models have not extremely good accuracy,
30% is still fairly accurate..
BC550 and 547 are the same devices, the difference
lies in cherry picking at the output of the chain production..
Since the models have not extremely good accuracy,
30% is still fairly accurate..
BC550 and 547 are the same devices, the difference
lies in cherry picking at the output of the chain production..
hmmm... Zero crossings ( dV/dt is highest) and caps at the output.
Your tests for stability are in the time domain I'm guessing, at what frequency does it oscillate? It's OK to use a square waves for testing
but the test stimulus should not exceed the amps slew rate for linear testing!
Fotios agree about LTspice and transistor device models. LTspice (switcherCAD) is mainly provided as a free tool for designing control loops for LTC products. The value for pro versions of spice tools is in accurate device models and since the cost here is low...
infinia,
Just a bit of clairification. LTspice is a high performance, general purpose, circuit simulator. Of course, it is offered free to help LTC sell their products, and it excells at SMPS simulation, but it can be used like any other SPICE package.
I agree with the comments on models, but this is not a SPICE problem.
Rick
Just a bit of clairification. LTspice is a high performance, general purpose, circuit simulator. Of course, it is offered free to help LTC sell their products, and it excells at SMPS simulation, but it can be used like any other SPICE package.
I agree with the comments on models, but this is not a SPICE problem.
Rick
agreed
Did you think I was implying a problem with LTspice😕
It's almost always an operator problem, then interpreting the results, and then the accuracy of models esp non LTC devices. LTC is not in the business of providing accurate transistor models is the only point I'm trying to make here.