@@sabamacx I should probably have mentioned those. The 5 mV 'max' offset voltage on the data sheet is a little concerning, although the 450 uV 'typical' value is quite attractive. The cheap non-monolithic pairs do about as well. A lot of them specify 2 mV max, 1 mV typical - and the last bunch of DMMT5551 I got had Vbe matching to about 300 uV. I suspect that the Cool Audio parts do better than it says on the tin and they just can't be troubled to test them, but I hate having to do incoming inspection on SMT parts. The 3046 is nice for exponential converters, where you can use one pair for the converter, one transistor for a temperature sensor, one as the tail current source, and the fifth as a heater. Running it at a constant 50 or 60 C means you don't have to have outboard temperature compensation. The super-matched pairs are scarce. LM394 is long gone. SSM2212 is still sort of in production, but the complementary SSM2220 is dead. MAT12 isn't terribly affordable. Nothing else came close to their specs!
@@KludgesFromKevinsCave Might be also fun exploring self-binning matched pairs for matching Vbe's. At least, it's still very useful when we are building small time stuff for ourselves. Sure, it's not applicable for mass manufacture, but I doubt that a great majority of the audience here are exclusively looking to use your presented material directly in that circumstance.
Hmm, maybe. But commercial matched pairs are dirt cheap nowadays. I can spend twenty bucks, get a hundred DMMT3904W's and not have the hassle of matching them (unless I need better than a millivolt in Vbe, which I seldom do.) If I need the pair on a breadboard, I can reflow solder it onto a carrier board pretty quickly. I also don't scour the Internet for tempco resistors or PTC thermistors to go with the pairs. A tiny platinum RTD bonded to the pair is a cheap and easy solution. I might still do what you suggest at some point, mostly for the nostalgia factor. I remember VCOs with pairs of PN2222's gooped to thermistors with heat-sink compound and zip-tied, after being carefully hand-matched with a rig like the one in musicfromouterspace.com/analogsynth_new/TRANSISTORMATCHER/TRANSISTORMATCHER.html The 'go/no-go tester', with its instrumentation amp built out of TL071's is mind-boggling. Nowadays, I'd just spend two bucks on something like an INA823, have that whole circuit (except for the window comparator) on a single IC, and never have to recalibrate it. Thinking aloud: Actually, a matcher for incoming inspection of cheap pairs might be an interesting idea. Pick up a SOT23-6 test socket off AliExpress, and add a little circuit with an instrumentation amp to measure them. That could work for a future video, so thanks for the idea!
Ouch! Something I'd rather forget! (Although if I remember right, they mostly used a solid blue, which was exactly the same colour as the Blue Screen of Death!) I'm so glad I allowed extra time to catch my flight, the day that all the departure boards in the New Orleans airport were showing flickr.com/photos/ke9tv/1507913971 !
@@KludgesFromKevinsCave Solid blue was used by early stages of Windows installer itself and many DOS programs. Windows software usually used the gradient. Some drivers, etc. used it as late as early 2000s. BTW. I think you may have mixed in something extra within your message.
@@KludgesFromKevinsCave Ahh, ok. I thought it was by mistake. I don't open any links I am not certain about. I thought it was a fragment of a message to someone else. d'oh
That's exactly what the REF200 does. It comes at a cost - the range of currents that you can handle drops quite badly. (How close could a 10 mA mirror pull its output to the supply rail?) IC designers generally avoid that topology unless they're going for really tight tolerances (REF200 guarantees 0.25%) for that reason. Also, aboard a chip, a resistor is more expensive than another transistor pair. If you look at the internal schematic of most analog IC's, you'll see a single resistor generating a reference current, which then gets mirrored all over the place.
I find current controlled systems very interesting. Most of the high end instruments of HP in the 70's and 80's use current control rather than voltage control. It's a pity, that components have not evolve since then. Example: The other day I was looking for a matched pair of low noise transistors for an audio application. This was very disappointing. First I couldn' t find a single BJT with better noise ratings than the old BC550/BC560. Moreover there is no matched pair of these. And when it comes to JFET's things get even worse.
I'm on the road at the moment. I'll check my notes on low noise matched pairs when I get home. Ping me (email in profile) if I don't get back in a couple of days.
Learning Again properly after 35 years.
The Behringer/Music Tribe owned semiconductor spin-off Cool Audio has the V3046M which is a contemporary monolithic matched transistor array.
@@sabamacx I should probably have mentioned those. The 5 mV 'max' offset voltage on the data sheet is a little concerning, although the 450 uV 'typical' value is quite attractive. The cheap non-monolithic pairs do about as well. A lot of them specify 2 mV max, 1 mV typical - and the last bunch of DMMT5551 I got had Vbe matching to about 300 uV. I suspect that the Cool Audio parts do better than it says on the tin and they just can't be troubled to test them, but I hate having to do incoming inspection on SMT parts.
The 3046 is nice for exponential converters, where you can use one pair for the converter, one transistor for a temperature sensor, one as the tail current source, and the fifth as a heater. Running it at a constant 50 or 60 C means you don't have to have outboard temperature compensation.
The super-matched pairs are scarce. LM394 is long gone. SSM2212 is still sort of in production, but the complementary SSM2220 is dead. MAT12 isn't terribly affordable. Nothing else came close to their specs!
@@KludgesFromKevinsCave Might be also fun exploring self-binning matched pairs for matching Vbe's. At least, it's still very useful when we are building small time stuff for ourselves. Sure, it's not applicable for mass manufacture, but I doubt that a great majority of the audience here are exclusively looking to use your presented material directly in that circumstance.
Hmm, maybe. But commercial matched pairs are dirt cheap nowadays. I can spend twenty bucks, get a hundred DMMT3904W's and not have the hassle of matching them (unless I need better than a millivolt in Vbe, which I seldom do.) If I need the pair on a breadboard, I can reflow solder it onto a carrier board pretty quickly.
I also don't scour the Internet for tempco resistors or PTC thermistors to go with the pairs. A tiny platinum RTD bonded to the pair is a cheap and easy solution.
I might still do what you suggest at some point, mostly for the nostalgia factor. I remember VCOs with pairs of PN2222's gooped to thermistors with heat-sink compound and zip-tied, after being carefully hand-matched with a rig like the one in musicfromouterspace.com/analogsynth_new/TRANSISTORMATCHER/TRANSISTORMATCHER.html The 'go/no-go tester', with its instrumentation amp built out of TL071's is mind-boggling. Nowadays, I'd just spend two bucks on something like an INA823, have that whole circuit (except for the window comparator) on a single IC, and never have to recalibrate it.
Thinking aloud: Actually, a matcher for incoming inspection of cheap pairs might be an interesting idea. Pick up a SOT23-6 test socket off AliExpress, and add a little circuit with an instrumentation amp to measure them. That could work for a future video, so thanks for the idea!
That blue gradient background always reminds me of Windows 3.11 era software installers.
Ouch! Something I'd rather forget! (Although if I remember right, they mostly used a solid blue, which was exactly the same colour as the Blue Screen of Death!)
I'm so glad I allowed extra time to catch my flight, the day that all the departure boards in the New Orleans airport were showing flickr.com/photos/ke9tv/1507913971 !
@@KludgesFromKevinsCave Solid blue was used by early stages of Windows installer itself and many DOS programs. Windows software usually used the gradient. Some drivers, etc. used it as late as early 2000s. BTW. I think you may have mixed in something extra within your message.
@@pvc988 Something extra? I meant to include the link to a picture of the Blue Screen of Death on the US Airways departure boards.
@@KludgesFromKevinsCave Ahh, ok. I thought it was by mistake. I don't open any links I am not certain about. I thought it was a fragment of a message to someone else. d'oh
I would like to see the effect of ambient temperature on current control. Very few real circuits run at a constant temperature.
@@klave8511 The beauty of current mirrors is that both transistors are at the same temperature. The variation of I_s with temperature cancels out.
A couple 1k's above the Wilson mirror can add some marginal improvement as well in the same way, I think.
That's exactly what the REF200 does. It comes at a cost - the range of currents that you can handle drops quite badly. (How close could a 10 mA mirror pull its output to the supply rail?)
IC designers generally avoid that topology unless they're going for really tight tolerances (REF200 guarantees 0.25%) for that reason. Also, aboard a chip, a resistor is more expensive than another transistor pair. If you look at the internal schematic of most analog IC's, you'll see a single resistor generating a reference current, which then gets mirrored all over the place.
You didn't specify whether the data in your colored curves are experimental or theoretically calculated ?
Experimental. The previous episode gave details of the test setup.
I find current controlled systems very interesting. Most of the high end instruments of HP in the 70's and 80's use current control rather than voltage control. It's a pity, that components have not evolve since then. Example: The other day I was looking for a matched pair of low noise transistors for an audio application. This was very disappointing. First I couldn' t find a single BJT with better noise ratings than the old BC550/BC560. Moreover there is no matched pair of these. And when it comes to JFET's things get even worse.
I'm on the road at the moment. I'll check my notes on low noise matched pairs when I get home. Ping me (email in profile) if I don't get back in a couple of days.
@@KludgesFromKevinsCave Have a nice trip. And I forgot to praise your channel - very neat mixture of theory and practice.