On Tue, 20 Aug 2019 19:32:37 +0100, "Kevin Aylward"
<kevinRemovAT@kevinaylward.co.uk> wrote:
>"John Larkin" wrote in message
>news:somllehqm0ggs82tmrf1e20smdjt4lq0jk@4ax.com...
>
>>
>>Well, yeah; usually with other components around 'em. LM13700 (and OTA
>>clones
>>with several other part numbers) are basically current mirrors with
>>differential-input
>>polarity switching. Put half a volt delta-V on the inputs, and the output
>>current
>>is slave to the current-program input pin.
>
>>That's cute, but the the Gm spec of the LM13700 covers about a 2:1
>>range. The bias mirror alone may be better.
>
>>Once you are willing to buy an opamp, you can make an arbitrarily
>>accurate current mirror.
>
>Sometimes used in CMOS design, where one has a not bad npn, but no decent
>pnp but you need a low noise top side current source. The opamp can loop
>around a pmos with the noise of the npn.
>
>-- Kevin Aylward
>http://www.anasoft.co.uk - SuperSpice
>http://www.kevinaylward.co.uk/ee/index.html
>
One fun circuit uses the V+ supply pin of the opamp as the current
output, so it has one amp and two resistors. There is a little Iq
error. That circuit can be cascoded too.
Two of those can make the perfect class B signal splitter.
Reply by Kevin Aylward●August 20, 20192019-08-20
"John Larkin" wrote in message
news:somllehqm0ggs82tmrf1e20smdjt4lq0jk@4ax.com...
>
>Well, yeah; usually with other components around 'em. LM13700 (and OTA
>clones
>with several other part numbers) are basically current mirrors with
>differential-input
>polarity switching. Put half a volt delta-V on the inputs, and the output
>current
>is slave to the current-program input pin.
>That's cute, but the the Gm spec of the LM13700 covers about a 2:1
>range. The bias mirror alone may be better.
>Once you are willing to buy an opamp, you can make an arbitrarily
>accurate current mirror.
Sometimes used in CMOS design, where one has a not bad npn, but no decent
pnp but you need a low noise top side current source. The opamp can loop
around a pmos with the noise of the npn.
-- Kevin Aylward
http://www.anasoft.co.uk - SuperSpice
http://www.kevinaylward.co.uk/ee/index.html
Reply by ●August 20, 20192019-08-20
>When you're done add resonance & pfb for lots more gain.
Sure, put a high Q tank on it and bias it class C and god will knock on the door trying to buy your sine wave.
I thought we were talking about a stage that can amplify pretty much any old frequency.
Reply by ●August 20, 20192019-08-20
>Can you post the circuit? In LTspice?
OK, I have done zero math on this and this is all off it is off the top of my head. It will probably work but with certain transistors and maybe some tweaking... If you want math send a sixpack. But this should give you a bit of an idea of what kind of circuit values you might need.
Version 4
SHEET 1 880 680
WIRE 224 128 144 128
WIRE 384 128 224 128
WIRE 64 144 64 128
WIRE 64 144 0 144
WIRE 64 176 64 144
WIRE 160 176 144 176
WIRE 224 272 224 224
FLAG 224 272 0
FLAG 224 32 0
SYMBOL npn 160 128 R0
SYMATTR InstName Q1
SYMBOL res 208 32 R0
SYMATTR InstName R1
SYMATTR Value 470 meg
SYMBOL res 160 112 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R2
SYMATTR Value 4,700 meg
SYMBOL res 160 160 R90
WINDOW 0 0 56 VBottom 2
WINDOW 3 32 56 VTop 2
SYMATTR InstName R3
SYMATTR Value 820 meg
TEXT -24 144 Left 2 !IN
TEXT 384 128 Left 2 !OUT
TEXT 208 8 Left 2 ;Vcc
Should work, maybe. It should have a ton of voltage gainbut due to the current loss its power gain is not high, in fact might be less than unity.
And you know you need a really low impedance to feed the thing, and a well buffed high impedance thing to drive with it or it just shorts out the output, it really should feed an FET stage.
Bottom line is that voltage gin does not equal gain. No actually if the input and output Zs are the same then it is. Comes to power gain, which is real gain, if the output Z is half the input Z then power gain is twice that of the voltage gain. If the output Z is twice the input Z then power gain is half the voltage gain.
You cannot ignore Z in a circuit. Tubes came close, they gave you a hell of alot of leeway. In solid state, impedance is almost more important than voltage.
Then if you get into frequencies, impedance is now the vector sum of resistance and reactance. It gets fun, but the math isn't all that hard, really.
Reply by John Larkin●August 19, 20192019-08-19
On Sat, 17 Aug 2019 03:18:20 -0700 (PDT), whit3rd <whit3rd@gmail.com>
wrote:
>On Friday, August 16, 2019 at 8:07:40 PM UTC-7, jla...@highlandsniptechnology.com wrote:
>
>Does anyone still sell monolithic current mirrors?
>
>
>Well, yeah; usually with other components around 'em. LM13700 (and OTA clones
>with several other part numbers) are basically current mirrors with differential-input
>polarity switching. Put half a volt delta-V on the inputs, and the output current
>is slave to the current-program input pin.
That's cute, but the the Gm spec of the LM13700 covers about a 2:1
range. The bias mirror alone may be better.
Once you are willing to buy an opamp, you can make an arbitrarily
accurate current mirror.
Reply by whit3rd●August 17, 20192019-08-17
On Friday, August 16, 2019 at 8:07:40 PM UTC-7, jla...@highlandsniptechnology.com wrote:
Does anyone still sell monolithic current mirrors?
Well, yeah; usually with other components around 'em. LM13700 (and OTA clones
with several other part numbers) are basically current mirrors with differential-input
polarity switching. Put half a volt delta-V on the inputs, and the output current
is slave to the current-program input pin.
Reply by Allan Herriman●August 17, 20192019-08-17
On Fri, 16 Aug 2019 20:07:32 -0700, jlarkin wrote:
> On Fri, 16 Aug 2019 15:11:08 -0700 (PDT), whit3rd <whit3rd@gmail.com>
> wrote:
>
>>On Thursday, August 15, 2019 at 12:18:47 PM UTC-7, John Larkin wrote:
>>
>>> Some engineers (and a lot of scientists!) have an undeserved affection
>>> for current mirrors, which seldom work well made from discretes, even
>>> "dual" transistors.
>>
>>Anyone who designs an op amp uses a current mirror or three. For an
>>LM13700, it's eight to twelve. For a CD4046, it's only three or four,
>>but redesigning without 'em would be a nightmare.
>>
>>Speaking dismissively of "even dual transistors" seems to indicate you
>>don't care about integrated circuits?
>>
>>You needn't share the affection, but it is never undeserved!
>
> I don't design ICs. Does anyone still sell monolithic current mirrors?
>
> Dual transistors aren't much better thermally than two separate
> transistors. It doesn't take much Vce*Ic on the output transistor to
> warm it up enough that it hogs all the current from the other one, so
> instead of being a current mirror, it becomes a current*beta mirror.
> That can run away.
>
> I posted a thermal image.
>
> Does anyone still sell monolithic pair transistors, with good thermal
> behavior?
http://www.thatcorp.com/300-series_Matched_Transistor_Array_ICs.shtml
They have an eight transistor die with four NPN and four PNP. It comes
in a variety of packages pinouts that allow you to access four of the
transistors (4 x NPN, 4 x PNP, 2 x NPN + 2 x PNP) but unfortunately not
all eight at once.
Allan
Reply by Bill Sloman●August 17, 20192019-08-17
On Saturday, August 17, 2019 at 1:07:40 PM UTC+10, jla...@highlandsniptechnology.com wrote:
> On Fri, 16 Aug 2019 15:11:08 -0700 (PDT), whit3rd <whit3rd@gmail.com>
> wrote:
>
> >On Thursday, August 15, 2019 at 12:18:47 PM UTC-7, John Larkin wrote:
> >
> >> Some engineers (and a lot of scientists!) have an undeserved affection
> >> for current mirrors, which seldom work well made from discretes, even
> >> "dual" transistors.
> >
> >Anyone who designs an op amp uses a current mirror or three. For an
> >LM13700, it's eight to twelve. For a CD4046, it's only three or four, but
> >redesigning without 'em would be a nightmare.
> >
> >Speaking dismissively of "even dual transistors" seems to
> >indicate you don't care about integrated circuits?
> >
> >You needn't share the affection, but it is never undeserved!
>
> I don't design ICs. Does anyone still sell monolithic current mirrors?
>
> Dual transistors aren't much better thermally than two separate
> transistors. It doesn't take much Vce*Ic on the output transistor to
> warm it up enough that it hogs all the current from the other one, so
> instead of being a current mirror, it becomes a current*beta mirror.
> That can run away.
>
> I posted a thermal image.
>
> Does anyone still sell monolithic pair transistors, with good thermal
> behavior?
On Fri, 16 Aug 2019 10:54:40 -0700 (PDT), jurb6006@gmail.com wrote:
>>Can you do that with one transistor?
>
>Why not if the hfe can handle it. Low current, small signal, possibly a transistor meant for RF amplification at the front end. High hfe and low Icmax.
>
>So you got say a 33 meg collector resistor and say a 100 meg from collector to base to bias it. No Re at all. The 100 meg is going to clip your top end you know, but it can work. And don't come crying to me when it is thermally unstable. You either find another transistor or heat sink it. Now imagine putting a heat sink on one that has a maximum Pd of 30mW. But then there are instruments that actually do put certain components in an oven, I shit you not. Like the crystal that controls the frequency of a TV or radio station. They need that by law and it is not cheap.
>
>Yes, you can have a million voltage gain with one transistor and two well chosen resistors.
I don't think so.
Reply by ●August 17, 20192019-08-17
On Fri, 16 Aug 2019 15:11:08 -0700 (PDT), whit3rd <whit3rd@gmail.com>
wrote:
>On Thursday, August 15, 2019 at 12:18:47 PM UTC-7, John Larkin wrote:
>
>> Some engineers (and a lot of scientists!) have an undeserved affection
>> for current mirrors, which seldom work well made from discretes, even
>> "dual" transistors.
>
>Anyone who designs an op amp uses a current mirror or three. For an
>LM13700, it's eight to twelve. For a CD4046, it's only three or four, but
>redesigning without 'em would be a nightmare.
>
>Speaking dismissively of "even dual transistors" seems to
>indicate you don't care about integrated circuits?
>
>You needn't share the affection, but it is never undeserved!
I don't design ICs. Does anyone still sell monolithic current mirrors?
Dual transistors aren't much better thermally than two separate
transistors. It doesn't take much Vce*Ic on the output transistor to
warm it up enough that it hogs all the current from the other one, so
instead of being a current mirror, it becomes a current*beta mirror.
That can run away.
I posted a thermal image.
Does anyone still sell monolithic pair transistors, with good thermal
behavior?