According to Fairchild datasheet from October 2009: "This device is designed for application as a video output to drive color CRT and other high voltage applications" but IIRC older datasheets (not necessarily from Fairchild) discloses them as Nixie-drivers, hence switching devices.
Last edited:
I wholly agree, but it is my publishers that make that decision, not me.
Can you tell me where the "Thompson" name comes from? Perhaps it's buried deep in the last 170 pages, but I haven't managed to find it.
I was not aware of that reference; many thanks for pointing it out.
Many thanks for the reference.
However if I understand the authors correctly, they are saying that to the best of their knowledge Thompson was the first to use a current-mirror in the input pair collectors to perform differential-to-single-ended conversion. (sometimes called phase summing)
This may very well be so, but the basic 3-stage configuration of transconductance stage - transimpedance (VAS) stage - unity gain output stage was of course well established long before that. (eg Tobey & Dinsdale in 1961)
I therefore don't think it helpful to call it the Thompson configuration, as it just leads to people saying "Who??" just as I and Bob Cordell did.
However if I understand the authors correctly, they are saying that to the best of their knowledge Thompson was the first to use a current-mirror in the input pair collectors to perform differential-to-single-ended conversion. (sometimes called phase summing)
This may very well be so, but the basic 3-stage configuration of transconductance stage - transimpedance (VAS) stage - unity gain output stage was of course well established long before that. (eg Tobey & Dinsdale in 1961)
I therefore don't think it helpful to call it the Thompson configuration, as it just leads to people saying "Who??" just as I and Bob Cordell did.
An excellent point. Do you happen to know in what year these models were introduced?
I enjoyed very much reading your book, it is full of very interesting experiments.
A comment/question about your simulation of a feedback system with distortion on page 64-67
This simple system can be analyzed analytically for second and third harmonic distortion. It will then give exact values for the distortion.
Using your values and levels of signals I end up ( input 20.02 and beta = 1) with close but different values : H2 = 0.068689% instead of 0.07724 and H3 = 0.0094364% instead of 0.01162%.
The analytic expression for 2° and 3° harmonic distortion shows also that they are exactly proportional to second and third power of the input. Therefore I believe that Baxandall statment is correct.
I suspect ( if I am correct) that the error is coming from the numerical precision of the Spice algorithm which per se is interesting.
A comment/question about your simulation of a feedback system with distortion on page 64-67
This simple system can be analyzed analytically for second and third harmonic distortion. It will then give exact values for the distortion.
Using your values and levels of signals I end up ( input 20.02 and beta = 1) with close but different values : H2 = 0.068689% instead of 0.07724 and H3 = 0.0094364% instead of 0.01162%.
The analytic expression for 2° and 3° harmonic distortion shows also that they are exactly proportional to second and third power of the input. Therefore I believe that Baxandall statment is correct.
I suspect ( if I am correct) that the error is coming from the numerical precision of the Spice algorithm which per se is interesting.
A small correction/mistake: in your test the input to the system is 20.2 and not 20.02 as I used wrongly.
I recalculated with 20.2 as input and the distortion is:
H2 0.0693% and H3 0.00960% still not the same as your simulated values.
But the real problem is that the exact formulae are showing H2 proportional to Vin and H3 to Vin².
I recalculated with 20.2 as input and the distortion is:
H2 0.0693% and H3 0.00960% still not the same as your simulated values.
But the real problem is that the exact formulae are showing H2 proportional to Vin and H3 to Vin².
I have being searching the year of introduction but did not find precise information. It should be around 1975.
A few years later, in Wireless World, B.J. Codd proposed this amp with a cascoded VAS (and a differential input stage with 3 transistors Wilson current mirror) which was on the way to the blameless philosophy.
An externally hosted image should be here but it was not working when we last tested it.
Same year, the Tomlinson Holman's APT 1 power amp appeared.
It had a differential input and a push-pull cascoded VAS.
The VAS push-pull was obtained by reverting the signal of one collector of the input stage by a current mirror, in a scheme such as the one depicted Figure 7.14, page 144, in Bob Cordell's book.
Last edited:
I suspect the differences here are not so much due to errors in the SPICE simulation as such, but in the subsequent Fourier analysis. I used PSPICE, and the Fourier analyser (is that part of the SPICE core? - I have no idea) does need a bit of coaxing to get it to work properly.
Glad to know you enjoyed the book.
Last edited:
I have delved into my archives and found a user manual for the Radford HD250 dated Feb 1974, which I think must be the date of introduction.
I remember it well; I was much impressed at the time, though I did think it a bit over-complex. Coming out of nowhere, Mr Codd got a lot of things right. However the design seemed to make very little impact at the time, for reasons unknown.
The only APT-1 schematic i have is dated 1979. Do you have an earlier version?
I like a bit of history. If anyone can contribute to the history of the cascode power amp I would be very grateful.
This shows that the accuracy of harmonic distortion estimated via Spice is not good and even very bad for high order. It is not a prove but any conclusion about distortion based on Spice is to be taken with care.
This is absolutely correct, especially the part "taken with care". Distortion simulation analysis, done properly and with good sanity checks, is very valuable in design and also understanding of how changes in a circuit can affect distortion. The directions of these changes are often correct.
However, comparing even carefully-done SPICE distortion simulations with real-world measurements will usually result in differences, and they can go either way. We all know, for examples, that even good transistor models are way from perfect representations of the transistors, especially for something like distortion nonlinearities. Also, the real things in a real amplifier that are not modeled can contribute a lot to distortion readings, especially things like nonlinear class AB currents that are induced into the signal path of the amplifier.
Cheers,
Bob
Of course model accuracy and pcb layout have a strong influence on the discrepancy between spice reults and measurement results.
But here the thing is that the simple second order non linearity followed by a gain of 100 in a loop with 100% feedback is an exact model to be simulated and for which an exact calculation of the harmonic distortion can be obtained in closed form.
The deviation with Spice result is a big question mark.
Reading this I wonder how many have built valve amplifiers ? Zero loop feedback seems a good option . My experience is that the transistor amps attempting zero loop feedback sound totally different . As said very unpredictable . One thing the valve amp can have is what some call pre-distortion . That is to cancel traits in devices when the phase is inverted between stages . Some talk of dramatic difference with grid stoppers when doing that , I am yet to see that myself . BTW . I find 1 % THD with harmonics in exponential series ( very hard to do ) sound as undistorted as zero distortion amps ( - 80 db or better ) . Anything else I have doubts about . Be certain I am saying the 1% THD exp is undistorted and not just to my ear . All in all the - 80 dB is far easier to build . Typical lower distortion valve amps sound wrong me . Processed and slow . Marantz 9 not so and Dynaco moderately so . Quad 2/22 is not my cup of tea .
I once asked Michael Gerzon what he thought about negative feedback . His answer was he never had , however if he had he might have said this " Some amplifiers need a lot of feedback and some very little " . Then after a pause and some uncontrollable laughter " You can be sure of one thing . Most have the wrong amount " .
I remember the PMA Denon 737 . It was an oddball . PMA 250 outclasses it . Light on it's feet if slightly yellow in tone colour ( best I can do as to how it sounds , not magnolia which would be better , this suggests a dynamic tonal quality ) . Yamaha A300 whiter and slightly bland , perhaps correct with undersized PSU ( cure that I think and who knows how good it might be ) . Rotel RA 930 pinkish with more punch . By a gnats whisker I preferred the Rotel . NAD 3020 a bit dark ,but equally good . I have very few examples of amplifiers costing 5 times the price I liked as much . Naim Nait is possible ( Inca-Tech Claymore ) . these were PMA 373 period . The NAD makes a superb pre amp or power amp stand in . Use two , one pre amp the other power . Or build a preamp psu . If I go to a house and they have a NAD I am happy to listen to music . The only regret is if they do not have a good source of sound . A Linn LP12 would be a good minimum standard . It is not the best turntable I ever heard but works well enough . AR turntable is a good choice . Digital front ends . Like the food we eat it will have to do . I eat at KFC last night so neither a digital nor food snob . KFC not take away , runaway .
I once asked Michael Gerzon what he thought about negative feedback . His answer was he never had , however if he had he might have said this " Some amplifiers need a lot of feedback and some very little " . Then after a pause and some uncontrollable laughter " You can be sure of one thing . Most have the wrong amount " .
I remember the PMA Denon 737 . It was an oddball . PMA 250 outclasses it . Light on it's feet if slightly yellow in tone colour ( best I can do as to how it sounds , not magnolia which would be better , this suggests a dynamic tonal quality ) . Yamaha A300 whiter and slightly bland , perhaps correct with undersized PSU ( cure that I think and who knows how good it might be ) . Rotel RA 930 pinkish with more punch . By a gnats whisker I preferred the Rotel . NAD 3020 a bit dark ,but equally good . I have very few examples of amplifiers costing 5 times the price I liked as much . Naim Nait is possible ( Inca-Tech Claymore ) . these were PMA 373 period . The NAD makes a superb pre amp or power amp stand in . Use two , one pre amp the other power . Or build a preamp psu . If I go to a house and they have a NAD I am happy to listen to music . The only regret is if they do not have a good source of sound . A Linn LP12 would be a good minimum standard . It is not the best turntable I ever heard but works well enough . AR turntable is a good choice . Digital front ends . Like the food we eat it will have to do . I eat at KFC last night so neither a digital nor food snob . KFC not take away , runaway .
I agree that there is USUALLY a discrepancy between a SPICE model and the real thing, but it's interesting to note one of the reasons why this is so.
Over the years, I noticed that the vast majority of simulator users simply use them "as is", and unfortunately, that means with 0% tolerance by default. Kid of ridiculous. This also applies to power supply tolerances, where most never even allow for the manufacturer's tolerance specs.
Changing say capacitor type (electrolytic, polar on bipolar, polyethyline or polycarbonate, etc) is hardly ever done. As if it's all one and the same thing.
Well, it's not. It does metter, and any simulator worth its salt will have different definitions for each type of component. Their tolerances are usually written on them, and are sometimes not small at all, for capacitors going up to +20, -50 %, although in recent years, most get it to +/- 20%.
Paying attention does make the model much more like the real thing, not ideal but way more realistic. I find this brings the model much closer to reality. My own sim hes never once let me down, if it says the circuit will work, it will work. If anything, it's a little condervative, so I find that if it says say 100 kHz, actual bandwidth will be 120 kHz or a bit better. If it says THD is 0.03%, I measure 0.02% or so.
But I do pay attention to my parts actual, real world specs, and my lowest tolerance is for metal film resistors at 1%, all my electrolytic caps are defined as 20%, etc. All it takes is just a little care.
Over the years, I noticed that the vast majority of simulator users simply use them "as is", and unfortunately, that means with 0% tolerance by default. Kid of ridiculous. This also applies to power supply tolerances, where most never even allow for the manufacturer's tolerance specs.
Changing say capacitor type (electrolytic, polar on bipolar, polyethyline or polycarbonate, etc) is hardly ever done. As if it's all one and the same thing.
Well, it's not. It does metter, and any simulator worth its salt will have different definitions for each type of component. Their tolerances are usually written on them, and are sometimes not small at all, for capacitors going up to +20, -50 %, although in recent years, most get it to +/- 20%.
Paying attention does make the model much more like the real thing, not ideal but way more realistic. I find this brings the model much closer to reality. My own sim hes never once let me down, if it says the circuit will work, it will work. If anything, it's a little condervative, so I find that if it says say 100 kHz, actual bandwidth will be 120 kHz or a bit better. If it says THD is 0.03%, I measure 0.02% or so.
But I do pay attention to my parts actual, real world specs, and my lowest tolerance is for metal film resistors at 1%, all my electrolytic caps are defined as 20%, etc. All it takes is just a little care.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.