I might do a mini tutorial on bipolar transistors from my experience, just to update the beginners.
Back in the bad old days, beta was everything. AND beta is nonlinear, what this means is it changes with current, and even collector emitter voltage.
40 years ago, we (real engineers) got 'engineering data sheets' on each part that we designed with. These were published by the manufacturer, with several pages on normal sized 8.5 by 11 inch paper. Many of these data sheets were VERY DETAILED. It was important then, because transistors were expensive, we were going to the Moon, and we had a war or two to make equipment for. Designers wanted to wring out everything that they could from each device, so they were given more info.
About 1970, after we had gone to the Moon, etc, and leaving the war zone, plastic transistors became more popular, and they were cheaper. Most designers were given less info on these devices as years passed, and that is where we are today.
However, in a nutshell, NPN transistors tend to sag in their beta at lower current. It so happens that every type of transistor has a PEAK BETA and this is one of the more interesting points to operate at, especially if you don't have a really, really low impedance drive signal. It so happens that Ft (a parameter that I can't define here at the moment) but related to the basic frequency response of the device, closely parallels the PEAK BETA. I have found examples of this in both small signal and power transistors, from at least 3 different manufacturers, that I have in front of me.
In general, PEAK BETA is noted on the data sheet, more that the PEAK Ft, but as they tend to track in every case that I can find at the moment, if you have one, you generally have the other. Now, is this more clear for everyone?
Back in the bad old days, beta was everything. AND beta is nonlinear, what this means is it changes with current, and even collector emitter voltage.
40 years ago, we (real engineers) got 'engineering data sheets' on each part that we designed with. These were published by the manufacturer, with several pages on normal sized 8.5 by 11 inch paper. Many of these data sheets were VERY DETAILED. It was important then, because transistors were expensive, we were going to the Moon, and we had a war or two to make equipment for. Designers wanted to wring out everything that they could from each device, so they were given more info.
About 1970, after we had gone to the Moon, etc, and leaving the war zone, plastic transistors became more popular, and they were cheaper. Most designers were given less info on these devices as years passed, and that is where we are today.
However, in a nutshell, NPN transistors tend to sag in their beta at lower current. It so happens that every type of transistor has a PEAK BETA and this is one of the more interesting points to operate at, especially if you don't have a really, really low impedance drive signal. It so happens that Ft (a parameter that I can't define here at the moment) but related to the basic frequency response of the device, closely parallels the PEAK BETA. I have found examples of this in both small signal and power transistors, from at least 3 different manufacturers, that I have in front of me.
In general, PEAK BETA is noted on the data sheet, more that the PEAK Ft, but as they tend to track in every case that I can find at the moment, if you have one, you generally have the other. Now, is this more clear for everyone?
Anatech, your questions are outside the discussion zone that I am concentrating on.
Ideally, a current source is quiet and linear. Reality is slightly different. Current sources should not transfer noise from the power supply to the output, but they can inject their own noise.
Ideally, a current source is quiet and linear. Reality is slightly different. Current sources should not transfer noise from the power supply to the output, but they can inject their own noise.
Bob (Cordell),
have you had any thoughts about re-doing your original distortion analyzer but with modern components?
Might make an interesting collaborative project on this site. There are a lot of newer, better components now so I am wondering if th e whole thing could be simplifed somewhat - single PCB, maybe a ucontroller to manage the user interface and automate measurements (or, maybe just use a PC and do it via USB).
PC based sound cards are good, but I think there is space for a top notch DIY analyzer that does'nt cost $40k (we just got a quote here on an AP system)
What do you think? I reckon there will be many willing volunteers.
have you had any thoughts about re-doing your original distortion analyzer but with modern components?
Might make an interesting collaborative project on this site. There are a lot of newer, better components now so I am wondering if th e whole thing could be simplifed somewhat - single PCB, maybe a ucontroller to manage the user interface and automate measurements (or, maybe just use a PC and do it via USB).
PC based sound cards are good, but I think there is space for a top notch DIY analyzer that does'nt cost $40k (we just got a quote here on an AP system)
What do you think? I reckon there will be many willing volunteers.
Bob, I also have a HP 3580 on my test bench, and it is wired in parallel to the HP 3563. I use if for quick measurements, sometimes, but it is much slower for what I want to do, so it doesn't get used very often. Also, I loaned a HP 3581 to SY, so he can do much the same thing, up to a point.
posted by Bonsai:
Bob (Cordell)..............have you had any thoughts about re-doing your original distortion analyzer but with modern components?...............Might make an interesting collaborative project on this site..........What do you think? I reckon there will be many willing volunteers.
Bonsai -- I am rebuilding my Cordell THD analyzer right now. One upgrade I want to try is to use the AD797 in place of many of the 5534's Bob Cordell originally specified. I would certainly be one of those volunteers to try some upgrades. I am open to any suggestions for improvements and willing to post my results.
Phil
Bob (Cordell)..............have you had any thoughts about re-doing your original distortion analyzer but with modern components?...............Might make an interesting collaborative project on this site..........What do you think? I reckon there will be many willing volunteers.
Bonsai -- I am rebuilding my Cordell THD analyzer right now. One upgrade I want to try is to use the AD797 in place of many of the 5534's Bob Cordell originally specified. I would certainly be one of those volunteers to try some upgrades. I am open to any suggestions for improvements and willing to post my results.
Phil
I use a Wavetek 7530A FFT plug in in a Tek 7603 connected to the output of my distortion analyzer. It works quite well on the ST1700 (and ST1710 I just confirmed) The technique makes looking at harmonics pretty easy. I have also used a Tek 7L5 (slow but very extended and high res) and a sound card with software. I use Praxis because I have it and it has a very powerful collection of processing capabilities. Used with a good 192 KHz sampling card and its max fft (16,777,216 points) the resolution and effective noise floor is very low. But I don't really want a computer in the middle of my test bench. Another noise source polluting all the measurements and the whole process is slower than the stand alone instruments.
There is a hazard trying to look at harmonics below the residual of the test system. The phase relationship of the specific harmonic and the harmonic generated in the DUT can either add or subtract and you can easily tell what is happening. That is the reason for the old requirement that the test system be 3X (or some say 10X) better than want you are measuring. Storing the residual and trying to subtract it from a measurement won't necessarily give you a real result.
John explained the beta peak concept, making a lot of sense. I have always used bipolar transistors as current gain devices, they seem to be intrinsically more linear that way. I have heard that FM Acoustics evaluates every single transistor for bumps in the beta curve and rejects any that are not perfectly flat. At his prices he can afford to.
There is a hazard trying to look at harmonics below the residual of the test system. The phase relationship of the specific harmonic and the harmonic generated in the DUT can either add or subtract and you can easily tell what is happening. That is the reason for the old requirement that the test system be 3X (or some say 10X) better than want you are measuring. Storing the residual and trying to subtract it from a measurement won't necessarily give you a real result.
John explained the beta peak concept, making a lot of sense. I have always used bipolar transistors as current gain devices, they seem to be intrinsically more linear that way. I have heard that FM Acoustics evaluates every single transistor for bumps in the beta curve and rejects any that are not perfectly flat. At his prices he can afford to.
Demian, you have a 7L5? You LUCKY GUY! Used one for years, had to give it back to JBL. Still miss it. Great for phono cartridge frequency response.
Demian, ask Brian what happened to the HP 3563 measurement when we turned on a TV next to the equipment, this week.
Hi John,
Totally agree w.r.t. the data sheets you get these days. I just bought the 1970 Motorola transistor data book, fifth edition. Wonderful amount of information in it. It still doesn't have all the info that earlier individual data sheets had.
For years I have noticed that US transistors had a marked peak in the Ic vs beta curve while the JIS devices had a much flatter curve. This is more pronounced in power devices.
Yes, I also think in terms of current when using BJTs. Voltage parameters are highly variable with temperature and current flow. The variable beta is also easier to get my head around.
The reason I brought up the CCS issue was that you had mentioned that you were going to use that technique with an op amp. In combination with the low noise floors you want to achieve, I felt that the differences might be more important to you. Otherwise, I would not have mentioned it.
Sorry for the O.T. in that case John.
-Chris
Totally agree w.r.t. the data sheets you get these days. I just bought the 1970 Motorola transistor data book, fifth edition. Wonderful amount of information in it. It still doesn't have all the info that earlier individual data sheets had.
For years I have noticed that US transistors had a marked peak in the Ic vs beta curve while the JIS devices had a much flatter curve. This is more pronounced in power devices.
Yes, I also think in terms of current when using BJTs. Voltage parameters are highly variable with temperature and current flow. The variable beta is also easier to get my head around.
The reason I brought up the CCS issue was that you had mentioned that you were going to use that technique with an op amp. In combination with the low noise floors you want to achieve, I felt that the differences might be more important to you. Otherwise, I would not have mentioned it.
Sorry for the O.T. in that case John.
-Chris
Chris, we have to walk, before we can run. If anyone wants any detail on current sources and their trade-offs, you should as Walt Jung. He gave me an ear-full when I spoke to him the other day, but the data is too detailed for this thread. I am not worried about adding a current source on the OUTPUT of an IC, the input, perhaps, when it comes to noise. Besides I would use a quiet fet, in any case. Just for fun, Chris, look at the 2n2219 data sheet in your book.
Hi John,
Okay, I'll have another look at the 2N2219. What am I looking for, or is it self evident?
In order to keep this thread on track, we could carry on with this via email. Sound okay to you?
-Chris
Okay, I'll have another look at the 2N2219. What am I looking for, or is it self evident?
In order to keep this thread on track, we could carry on with this via email. Sound okay to you?
-Chris
John:
I have 3 7L5's. One with the tracking oscillator, one modified to work with the 7854 and one with a trashed front panel for parts. . .
I don't need to turn on a monitor or a TV, The fluorescent over my bench has a HF ballast. Constant light output and enough emi to really limit the measurements when its on. And the Metcal also lights up the scope when its on. but there are no good substitutes for them for real work.
I have 3 7L5's. One with the tracking oscillator, one modified to work with the 7854 and one with a trashed front panel for parts. . .
I don't need to turn on a monitor or a TV, The fluorescent over my bench has a HF ballast. Constant light output and enough emi to really limit the measurements when its on. And the Metcal also lights up the scope when its on. but there are no good substitutes for them for real work.
I have been looking at the HP3562 and I have a question. The dynamic range is specified as 80dB. Does that mean that for any usefull low-level measurements you need to have a distortion analyzer, or at least a notch filter, in front of it?
With 80dB dynamic range you couldn't measure anything below 0.01%, right?
So, basically, you would use the HP just as an FFT engine, but use another piece of hardware to extract the signal nonlinearities (or suppress the carrier) as noted by Demian, correct?
Jan Didden
With 80dB dynamic range you couldn't measure anything below 0.01%, right?
So, basically, you would use the HP just as an FFT engine, but use another piece of hardware to extract the signal nonlinearities (or suppress the carrier) as noted by Demian, correct?
Jan Didden
Most distortion setups have a distortion output, i.e., the signal with the fundamental notched out. It's that part that gets FFTed. The HP 339, 334, and 333 do, anyway.
John, I suspect CCS noise at the output isn't an issue. It gets divided down by the ratio of the CCS source Z to the amp's output Z.
John, I suspect CCS noise at the output isn't an issue. It gets divided down by the ratio of the CCS source Z to the amp's output Z.
SY said:
OK, I thought I was missing something.
I must say that a good quality external USB soundcard looks pretty hard to beat by a stack of vintage HP or Tek, on all counts you can think of.
Better dynamic range/resolution, more user friendly (with the right software of course), excellent documentability (sp?), small, light, cheap.
OTOH, there's something about that originally very exensive build-like-a-battleship stuff taking over your bench, I concede that. 😉
Last weekend I visited a large radio-amateur get-together with the firm plan not to buy anything. I entered the hall, and within 10 minutes had bought a Tek 571 curve tracer. Do I need it? Probably not. Do I like it? Yes! 😀
Jan Didden
I use both- the soundcard plus my homebrew analog interface seem to work flawlessly, but it's nice to have a reality check. The other virtue of using the distortion analyzer residual output is that the qualitative appearance of the time-domain residual can often provide a better clue to the source of the distortion than looking at the spectrum.
I am using for my measurements here Amber 3501 and HP3563, together with a good sound card (modified RME Digi 96/8 PAD, goes down to about -114dB THD at midrange). This setup is mostly sufficient for my requirements.
Alex
Alex
janneman said:
Jan,
Take a look at this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=137843&highlight=
Has some good clues about soundcards. Sound cards have their advantages (dynamic range) OTOH there are lots of things where one of those tanks are much, much better. A few extra goodies I can quickly think of:
- DC coupling
- Calibrated attenuators
- Multiple inputs
- Amplitude and phase references available
- Low noise
OTOH, good software for a sound card is not free (yes, I know, but I have to tell this 🙂 ). If one would add the cost of the sound card hardware and the cost of the software you could easily buy one of those second hand "tanks".
- Status
- Not open for further replies.