It depends. For BJTs a simulated 'THD' might not be too far from the truth. For FETs and valves simulated 'THD' can be ignored. 'THD' itself is only of limited value, of course.
The thing which annoys me is where someone says "measured" when they mean 'result of simulation'.
If it is any consolation, professionals and academics do this too. I remember being at an international conference (on RF electronics) where delegates had to ask some speakers whether their 'measurements' were genuine or simulated. It seems that at some of the lesser 'universities' very little real work takes place, but a lot of virtual work.
The thing which annoys me is where someone says "measured" when they mean 'result of simulation'.
If it is any consolation, professionals and academics do this too. I remember being at an international conference (on RF electronics) where delegates had to ask some speakers whether their 'measurements' were genuine or simulated. It seems that at some of the lesser 'universities' very little real work takes place, but a lot of virtual work.
I made a career (41 years at Motorola) out of taking the output from those people who "calculate and simulate" and turning in into real (prototype) hardware. It seems that most of the simulator geniuses did not want to build functioning hardware, or didn't see the value in it.
This was all RF in the 100 MHz to 2.5 GHz range. For some designs the simulator does OK, the later versions of HP/Agilent/Keysight ADS is pretty good at predicting the noise figure of an LNA, but try simulating the phase noise of a 1 GHZ VCO that runs on 3 volts and 5 mA......random number generator.....got to build one, or a dozen.
Same deal with using LTspice on tube amps. Some designs and tubes simulate well, as long as you drive G1. Screen drive simulations are rarely close. If LTspice says something won't work, it usually won't. Success, can not often be accurately predicted, let alone an accurate distortion profile.
category error?
or another run at the strawman "conventional engineer", with the common lie that "they" think "THD only" is the end all be all of audio circuits?
I don't really think there are that many threads where the participants don't understand the limitations and proper uses of sim
what I see is mostly others not really interested in the EE theory that is being illustrated break in with their "just listen" testimony, particularly when the circuit theory, sim indications and often the Psychoacoutics go against their pronouncements of what they find xxx circuit/topology/part type "sounds like"
claiming they don't need to understand the issues because their ears tell them "truths" that the despised "sim jockeys" can't understand
of course they mostly disappear when you ask about properly controlled listening tests - replicated, "blueprinted" pairs of circuits with just the change under discussion different between them switched with a blinding protocol
so they aren't even proper "subjectivists"
or another run at the strawman "conventional engineer", with the common lie that "they" think "THD only" is the end all be all of audio circuits?
I don't really think there are that many threads where the participants don't understand the limitations and proper uses of sim
what I see is mostly others not really interested in the EE theory that is being illustrated break in with their "just listen" testimony, particularly when the circuit theory, sim indications and often the Psychoacoutics go against their pronouncements of what they find xxx circuit/topology/part type "sounds like"
claiming they don't need to understand the issues because their ears tell them "truths" that the despised "sim jockeys" can't understand
of course they mostly disappear when you ask about properly controlled listening tests - replicated, "blueprinted" pairs of circuits with just the change under discussion different between them switched with a blinding protocol
so they aren't even proper "subjectivists"
Last edited:
Tubelab - I'm sure you can relate here... How many times have you sit through a design review where the designer has put EXACTLY what the simulator spit out as the expected typical performance in the compliance matrix? And pretty much guarantees to marketing/management that is is EXACTLY what they are going to get? With a straight face....
I think that depends on the individual. I spent 10 years working for National Semiconductor and Texas Instruments combined working as a circuit design engineer, so I fit pretty squarely in the "calculate and simulate" category. As did the 15-or-so other engineers I worked with. Of those 15-or-so engineers, there were maybe two who tended to shy away from the lab whereas the rest of us enjoyed the break from the simulator that the lab time provided.
Really? I've simulated quite a few VCOs. One of National's VCO experts sat a few cubicles away from me and simulated VCOs all day. High-end VCOs and other RF circuits used in the cellphone base stations. Our simulations were a tad optimistic. The sim showed results that were about 2 dB better than reality, so we took that into account in the design phase. If we needed to hit a -160 dBc/Hz noise floor, we'd design for -162 or -163 dBc/Hz (which is actually very hard to achieve at the frequencies we were operating). The final product, measured on an HP/Agilent/Keysight/whatever-their-name-is-this-week E5052B phase noise analyzer came in on target.
You can see the results of this methodology in the LMK04000, LMK04800, LMX2581, LMX2541, LMX2592, LMK03806, etc.
That said, National's semiconductor modelling group was pretty awesome. The models weren't a perfect match for reality, but they were pretty darn close and there was a strong push to make them fit reality even better.
Sure. If your models are a poor match for reality, you'll get results of your simulation that poorly matches simulation. The tube models that are out there are generally fine for getting the DC operating point figured out. They may be OK for getting a general idea of the frequency response of the circuit. I would certainly not rely on them for simulating distortion, noise, or any other parameter that pushes the models.
If the management bought that, they're pretty gullible. For new designs I would always say, "the simulator says... but reality may differ". That's the risk of a new design. For tweaks of an existing design, it's much easier to correlate simulation and reality. Even in the cases where I was making a small tweak to an existing circuit and had both simulation and lab data, I was getting quite a bit of push-back from management, basically assessing my confidence in the simulated values. That's their job. They're the ones who have to take on the risk.
Depends. For THD sims, it's common that the time step used in the simulation is different from the sampling time used in the FFT used to calculate the THD. This causes artificially inflated THD numbers.
Yeah... There appears to be some overlap with religion.
The US vs THEM is pretty easy to spot as well. It's unfortunate as it degrades the quality of the dialogue. It can be interesting to watch from a cognitive psychology perspective, though.Tom
I read only the first half of this thread, but when it comes to loudspeakers, it may be impossible for a DIYer to measure plots of general value. Low frequencies, a small room with few damping, a bigger or complicated (noncorrelated multiway) loudspeaker -- it makes no sense to post the "random" results. People, who have access to anaechoic rooms, heated barns and big yards show good new loudspeaker ideas once in a while, but not nearly often enuf.
when I use sim here it is to illustrate EE, feedback, Signal Theory in a way still hopefully somewhat accessible to hobbyists, many of who do read schematics, know basic circuit probing, debugging from schematics
far more of those than the few willing to parse equations
and I do have decades of pro EE experience in product development and design in industries with objective performance standards, requiring 3rd party product performance certifications, to give me confidence that the "theory" actually does usefully work "in the real world"
but we do have a "naïve subjectivist" contingent that want to tell stories, claim validation by builds that we don't have any way of inspecting, that they assure us that if we only used our ears...
while they don't seem to acknowldge large parts of perceptual psychology, plasticity of memory, unreliability of unverified "experience"
and the dominating role of highly variable source, speakers and rooms in what reaches our ears compared to merely competent electronics
obviously there is a need or threads supporting real builds with complete schematics, parts lists, construction tips - and it is a big undertaking that should be encouraged
but the fannish behavior, denial of Psychoacoustics, irrational EE design decisions that apparently can't be questioned seem to be a problem
far more of those than the few willing to parse equations
and I do have decades of pro EE experience in product development and design in industries with objective performance standards, requiring 3rd party product performance certifications, to give me confidence that the "theory" actually does usefully work "in the real world"
but we do have a "naïve subjectivist" contingent that want to tell stories, claim validation by builds that we don't have any way of inspecting, that they assure us that if we only used our ears...
while they don't seem to acknowldge large parts of perceptual psychology, plasticity of memory, unreliability of unverified "experience"
and the dominating role of highly variable source, speakers and rooms in what reaches our ears compared to merely competent electronics
obviously there is a need or threads supporting real builds with complete schematics, parts lists, construction tips - and it is a big undertaking that should be encouraged
but the fannish behavior, denial of Psychoacoustics, irrational EE design decisions that apparently can't be questioned seem to be a problem
Last edited:
Far too many. For about 20 years I worked in one of the product groups, designing either high spec hand held two way radios for public safety (police walkie talkies) or cell phones (Nextel, Clearnet, Telus etc. walkie talkie phones). There would be several design reviews along the way from concept to product. Simulation only data is acceptable if not required in the early going, but at some point there has to be real measured data on a prototype of each individual stage or circuit, along with a block diagram and some high level system simulations. Then someone like me would glue it all together, and lay out a PCB for the first of several iterations of hardware. These were used not only for system level data, but software development, and reliability testing. The product development cycle for a "mission critical" product where someone's life depends on it working correctly in ALL possible circumstances could range from one to three YEARS! The hardware development team could be as big as 100+ people, with even more on software.
Worse were the design reviews where all the players brought all the required data, including real measured hardware on enough units to have statistical data on key performance goals, but failed to see the critical flaw in the design concept that would prevent the product from working correctly in the real world.......
There were quite a few "old timers" in the early days (70's through mid 90's) but little by little company policy (and benefits) changed to discourage this. One of the high level bosses remarked that "technology changes so quickly that long term experience is of no value, and is often wasteful." He also stated that "engineers are cheap, don't like somebody, fire them and get another." This was really prevalent in the cell phone group where the average age of a hardware engineer was 27. The software people were even younger.
I was 20 years old when I started working as a technician in the Motorola factory testing and tuning radios. I had previously worked running the service department in an Olson's Electronics (stereo) store and as a TV shop tech. In that first year there was an issue that put a $1M contract in jeopardy because the first batch of radios often heard cartoons instead of the police dispatcher. After about a month of "panic level engineering" during which time I was borrowed for radio testing, it was determined that a TV station was broadcasting on the radios "image" frequency. The 470 to 520 MHz radio spectrum was (and still is) shared between TV and two way radio.
....There would be 4 more times in my 41 year career where this same mistake would be made, and each time I was the only "old timer" who could remember the original blunder. The second time I was ignored, and Motorola lost a zillion dollar contract that led to the early demise of the paging division. The 3rd and 4th time, I managed to effect a change (or an actual product cancellation) before it was too late. There is some use for us "old farts" even if it's just to remember past mistakes, before repeating them.
Each group was different. In the phone group there were few (maybe 20%) of the people who actually had test equipment on their bench other than a cell phone test box.
In the research group, it was maybe 50/50, but the lines were much stronger. There were only two of us who could do both.
The last twelve years of my career were spent in a research and development group where we developed the new technology that went into the next generation of mission critical radio products. Our group had about 250 people scattered across 6 or 8 different facilities in 5 countries. We were exploring all aspects of next generation radio technology. There were guys that worked on one type of circuit for their entire career, say VCO's. I was a transmitter guy, and general radio system integrator. Over time the group shrank. Every year there was another layoff, or two or three. There about 30 of us left in Florida, and a few more in two other countries. We had morphed into an IC design center. I went to the boss and asked what needed to be done to continue my employment, and poof, I was a EVB, and whatever else they needed to test and evaluate IC chips, designer.
The two way radios that we designed work in one of 5 frequency bands from 30 MHz to 941 MHz. Each radio channel is 12.5 or 25 KHz wide, and channels are adjacent. Some frequency bands are shared with TV stations with 1 MEGAWATT transmitters, and cell towers with say, fifty 30 watt transmitters.
We had the -160 dBc/Hz noise floor requirement and a noise level requirement mask that was hard to meet, yet needed to ensure the radio products requirements for adjacent channel rejection. -150 dbc/Hz at 1 MHz and -125 dbc/Hz at 25 KHz were the hard points, in a Frequency Generation Unit that covered 30 Mhz to 1 Ghz. The radio had to run from battery power, and last a police officers 10 to 12 hour shift, and function from -40C to +85C. The power budget for the VCO was 15 milliwatts, and the entire FGU got 25 milliwatts.
Together these requirements are seriously difficult. The simulator was reasonable at getting the noise floor, but close in noise simulation is a crap shoot (+/- 10 db), since board layout, parts choice, and even module assembly quality affected the actual parts. There was only one varactor that could function in those VCO's and they were screened and selected by the manufacturer.
I can tell you that I have evaluated some of those chips, and ignoring the power requirements, they are not quite there on close in phase noise. In the last 5 years there were 3 guys that designed an integrated solution that replaced a bank of VCO's and two PLL's with a single IC chip and a single "top secret" VCO. Even though I was one of the testers and designers of the test boards, I or most of the department, was not allowed to go to any of the meetings, or know how it all works.
I'm guessing the 25 kHz offset is what you're referring to as "close in". Yeah... That gets you into the noise of the PLL itself (the charge pump and phase detector noise mainly). That's a tough one to hit. The 800 kHz offset phase noise number is another tough one.
Either way. Our customers were happy with it.
Did you ever evaluate the crystal oscillator in those chips?
Tom
PS: I like your shiny new user name, by the way.
Anything below 1 MHz is considered close in, but the requirements get tough inside 100 KHz, with 25 KHz being the hard one, and the most difficult to simulate correctly, since the discrete device models don't accurately reflect device parameters in that region. As I stated most of the FGU has been successfully integrated with the exception of two high spec VCO's. That system is now in a shipping product, but most of that work was done after I left. My 100 MHz to 1GHz RF power amp using a GaN device and 28 volt boost converter to power it, did make it into the radio.
Motorola APX? 8000 Portable Radio - Motorola Solutions
At the time Motorola had over 90% of the US public safety radio market, and about 60% world wide. There are several ways that radio reception can be interfered with, but they come down to either an external interfering source, an internal interferer, or two or more external sources combining in a non linear element inside the radio to create an internal source.
External sources can be a lot of things, with TV and cell towers being the obvious sources, but another two way radio with marginal performance can be a big problem. It was noticed back in the 80's when synthesized radios became popular that two cops in the same cop car couldn't talk on their radios at the same time. Ditto the early Nextel Pocket Phone. There was so much broadband noise coming from the transmitter that it blanketed a nearby receiver. Hence the ever tightening noise floor spec. Fractional "N" spurs are another biggie. A typical big city multiple transmitter system has its channels spaced 1 MHz apart, hence the 1 MHz requirement.
The "close in" specs have to do with the receiver. It is not uncommon in metropolitan areas for a radio used to be receiving a transmission from his dispatcher 20 miles away, while someone a mile away is blasting him with a 100 watt base station on the adjacent channel 25 KHz away. This is the classis "near / far" problem and had been an issue ever since radio was invented. Assuming the interferer's transmitter is "perfect" and emits nothing on the adjacent channels, then it is up to the receiver's selectivity to reject the unwanted strong signal, and receive the weak signal. If the receiver's local oscillator has significant noise power in the adjacent channel, then that unwanted signal will be converted to an on channel signal in the receivers mixer.
Most public safety radio systems operate on radio channels spaced 25 KHz apart, hence the 25KHz requirements, there are some 12.5 KHz spacings, but they are not supposed to be used for "mission critical" applications. We still have a noise spec for 12.5 KHz offset, and 50 KHz because the APCO (Association of Police Communications Officers) spec uses 50 KHz for IMD testing.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.