Wavebourn said:
Because the internal base resistance Rbb is the biggest contributor to voltage noise.
Having more transistors in parallel (or a wider transistor from the start) reduces Rbb in proportion.
Rbb behaves like a resistor in the base lead, and 10 Ohms generates noise of
0.4nV/sqrt Hz in series to the signal source.
1K would generate 4 nV/sqrt Hz (grows/shrinks with the sqrt of the resistance)
Gerhard
Wavebourn,
Kind regards...
I`ve been pondering all night long still have minor difficulty finding that obvious... the emission seems to be caused by the applied base to emitter voltage...and there`s a problem with leaving out that parameter from the equation... the base current is just a coinciding byproduct...
Kind regards...
gerhard said:
The question was, why they did not allow more than 30 MA if devices are big enough?
Actually, I am searching for double transistors for current mirrors that will work on about 1A currents. Sanken devices have a transistor and a diode; it is an option. But 2 transistors as close together thermarly as possible would be better.
This is where different educational systems show differences.
For example, in the early days of solid state electronics, BETA was king! It was true in all the books, complicated or popular. Heck, I didn't even know how to calculate the Gm of a bipolar transistor, being educated in engineering through my junior year. This was at the time when I regularly tested Gm of tubes, in a transconductance tube tester. I knew the concept, it just didn't seem very important.
The emphasis on Vbe came when betas increased in value, AND Gummel-Poon, I'm pretty sure, a couple of guys from Bell Labs, came up with a new model of the bipolar transistor, about 1970. Often, even today, thinking of BETA as the driving force will give you accurate and elegant solutions to circuit problems.
Wavebourn's question is fair enough. Why so many paralleled devices, yet almost no rated current? EVEN single transistors, with high beta, in the 1960's usually had a peak rated current of 50 ma at least, and they were only a single device, optimized for low currents, such as 100ua.
As far as paralleling devices, we have been doing it to lower Rbb' since the middle '60's due to an article by E.A. Faulkner (sp?) an Englishman, who showed us in an article in 'Electronics Letters' in 1966-67. He was the first, as far I can tell. The rest is history, or and the usual progress of levels 1 to 3.
For example, in the early days of solid state electronics, BETA was king! It was true in all the books, complicated or popular. Heck, I didn't even know how to calculate the Gm of a bipolar transistor, being educated in engineering through my junior year. This was at the time when I regularly tested Gm of tubes, in a transconductance tube tester. I knew the concept, it just didn't seem very important.
The emphasis on Vbe came when betas increased in value, AND Gummel-Poon, I'm pretty sure, a couple of guys from Bell Labs, came up with a new model of the bipolar transistor, about 1970. Often, even today, thinking of BETA as the driving force will give you accurate and elegant solutions to circuit problems.
Wavebourn's question is fair enough. Why so many paralleled devices, yet almost no rated current? EVEN single transistors, with high beta, in the 1960's usually had a peak rated current of 50 ma at least, and they were only a single device, optimized for low currents, such as 100ua.
As far as paralleling devices, we have been doing it to lower Rbb' since the middle '60's due to an article by E.A. Faulkner (sp?) an Englishman, who showed us in an article in 'Electronics Letters' in 1966-67. He was the first, as far I can tell. The rest is history, or and the usual progress of levels 1 to 3.
The limit is only 20 mA for Mat02.
As Scott has already written, they were intended for low parasitic
resistance so they would conform to the log law over many decades of
current. They were technology enablers for log converters, exp functions
and precise analog multipliers.
This is a completely different world than 2*1A*18V = 36W current mirrors.
I don't think that they have extra thick metalisation or bond wires.
At the 1A level one would probably try to get away with less
reference current and try to use a scaling current mirror.
Perhaps one could use those ON Semi ThermalTrak transistors.
Gerhard
Once more: I'm not questioning what for emitters are so large. I myself used medium power transistors about 30 years ago in dynamic microphone preamplifiers, when found that medium power RF transistors KT626 developed in our institute worked fine on low currents.
The question is an opposite: can MAT02 be used on higher currents?
Datasheet shows dependence of beta on current up to 1 mA only.
The question is an opposite: can MAT02 be used on higher currents?
Datasheet shows dependence of beta on current up to 1 mA only.
Wavebourn said:
Once more: IC abs.max. = 20 mA, it's under Absolute Maximum Ratings, could they be clearer?
For beta @ Ic=1-20 mA:
If it's not in the data sheet, measure it. It's easy enough
and more reliable than the wild guesstimates and speculations
on a hobbyist and self-adulation platform.
Gerhard
john curl said:
Please don't be surprised when IC designers ignore you. Beta has virtually nothing to do with modern design, it is a nuisance only. The logarithmic relationship between Vbe and Ie is all that matters, log based circuits even work at beta ~= 1. Please get a copy of the current mode design book that Barrie Gilbert co-wrote (contributed heavily too?) and get up to date!
BTW even in 1969 I never learned that bi-polar transistors were current driven devices. This is very wrong thinking.
A new servo
I was giving the front end servo some more thought; certainly, it's an interesting challenge. In the previous schematic, the servo output current noise contributes to the input referred noise as In/Gm where Gm is the first stage (JFET) transconductance. The V/I stage in the servo, required to convert the output voltage to an current error imposes some challenges; it needs a decent headroom (10-15mA) to compensate for gross JFET mismatching, hence the 100kohm resistors. These makes the V/I stage pretty sensitive to the opamp input noise current and therefore a JFET input, high voltage opamp (such as OPA445) is required. Certainly, not a cheap part and not in everybody's drawer. Such opamps are usually not optimized for noise voltage and precision. Even if the noise voltage doesn't contribute much to the total noise, it would certainly useful to use a cheaper low noise opamp. Also, the whole servo arrangement looks pretty complicated...
Here's a modified servo. It uses only one opamp, JFET input type. This one doesn't even need to be low noise, any fair precision model will do. It also doesn't need to be high voltage! I've tried so far OPA134, AD8065, AD8097 and OPA827 with very similar results, overall still around 0.31nV/rtHz.
Now, the question. Small differences (in terms of noise) between these opamps still exist. I was not able to find any correlation between the opamp datasheet specs and the noise results (OPA827 is doing the best, the worst being OPA134). Is there any well established relation between an opamp input voltage/current noise performance, the load, and the resulting noise current in the supply lines? This noise contributes to the total noise as much as the output current noise in the previous V/I convertor...
I will of course measure this independently, just curious if there is any well established theory about.
I was giving the front end servo some more thought; certainly, it's an interesting challenge. In the previous schematic, the servo output current noise contributes to the input referred noise as In/Gm where Gm is the first stage (JFET) transconductance. The V/I stage in the servo, required to convert the output voltage to an current error imposes some challenges; it needs a decent headroom (10-15mA) to compensate for gross JFET mismatching, hence the 100kohm resistors. These makes the V/I stage pretty sensitive to the opamp input noise current and therefore a JFET input, high voltage opamp (such as OPA445) is required. Certainly, not a cheap part and not in everybody's drawer. Such opamps are usually not optimized for noise voltage and precision. Even if the noise voltage doesn't contribute much to the total noise, it would certainly useful to use a cheaper low noise opamp. Also, the whole servo arrangement looks pretty complicated...
Here's a modified servo. It uses only one opamp, JFET input type. This one doesn't even need to be low noise, any fair precision model will do. It also doesn't need to be high voltage! I've tried so far OPA134, AD8065, AD8097 and OPA827 with very similar results, overall still around 0.31nV/rtHz.
Now, the question. Small differences (in terms of noise) between these opamps still exist. I was not able to find any correlation between the opamp datasheet specs and the noise results (OPA827 is doing the best, the worst being OPA134). Is there any well established relation between an opamp input voltage/current noise performance, the load, and the resulting noise current in the supply lines? This noise contributes to the total noise as much as the output current noise in the previous V/I convertor...
I will of course measure this independently, just curious if there is any well established theory about.
scott wurcer said:
When I was educated in the 70's, I have started with the stupid H parameters and the observation that H21e=beta. One year later we were taught the Giacoletto model that finally logically ties the AC analysis results to the bias point. Never had a chance to use the quadripole parameters again, except in a couple of occasions when I had to deal with the Sij (for a BFR91 that I cloned and then for a proprietary 10GHz silicon transistor that I designed the process for).
gerhard said:
I respect datasheets, but when devices are made for particular usage specs are given according to potential applications. Bright example: once I found that 6P15P tube is the same as 6P14P; except the 3'd grid has own pin. It was positioned as more linear tube that 6P14P, and maximal screen grid voltage was rated on much lower level. Investigations revealed that tubes are identical, the 2'nd grid is not fragile like the majority believe; it was rated according to applications (video amplifiers) they were specified to be used, while 6P14P were made for audio output stages.
The same happens with transistors made for certain applications.
I used long time ago 4-transistor devices in musical synthesizers: 2 transistors in a converter, one transistor as a heater, and the last one as a temperature sensor. They match and track.
The problem with my current need is, I need 2 transistors that have to much and track, while one of them will generate a heat enough to screw down precision of mirroring if another on will not track it.
And I want to use only 2 transistors, no additional feedback devices are desirable.
Something like TL014, but for higher currents.
javascript
penreq('http://www.datasheetcatalog.org/datasheets/270/502479_DS.pdf')- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.