Prev: NXP LPCXpresso demo board
Next: Guitar one octave Up
From: Joerg on 26 Mar 2010 14:04
Michael Robinson wrote:
> "Joerg" <invalid(a)invalid.invalid> wrote in message
> news:8147ltFmnU1(a)mid.individual.net...
>> Michael Robinson wrote:
>>> On page 4 of this datasheet
>>> http://www.nxp.com/documents/data_sheet/BCV62.pdf
>>> the manufacuterer gives 0.7 to 1.3 for Ic1/Ic2.
>>> Other manufacturers' datasheets use the same number.
>>> How many sigmas away from the mean the 30% accuracy statistic is makes a
>>> huge difference in the attractiveness of this current mirror as a part I
>>> want to buy. But the datasheets don't give that kind of information.
>>> Have you used these current mirrors and can give me an idea what kind of
>>> accuracy could I expect from the actual physical parts?
>> It's usually better but they do not guarantee it. Not sure if they'd mix
>> devices from different wafers but that can make for a huge difference.
>>
>> If you need closer tolerance straight out of the gate it'll be a lot more
>> expensive:
>>
>> http://www.analog.com/static/imported-files/Data_Sheets/SSM2220.pdf
>>
>> --
>> Regards, Joerg
>
> The lack of promised accuracy was annoying but I can deal with it.
> Temperature is another matter. In all the threads I've seen here about
> current mirrors and dual transistors, nobody was ever able to tell for sure
> whether the parts were built on a single die. It would be nice if somebody
> did some more testing of various parts, like that infra-red photo John
> Larkin just posted. God forbid they should tell you anything useful on the
> datasheet.
>

Datasheets are nowadays rather dumbed-down, don't expect much. What does
help (mostly ...) is contacting application engineering. I know this
will dash some of your hopes but the typical dual-transistor is not on a
single die. It's multi-chip. Else you'd have entries for voltage
differences or substrate breakdown in the abs max ratings.

There are very few true arrays left. In the CMOS area you still have the
cheap CD4007, in the upper class the SD5400 series but that has gone
boutique, meaning higher cost. AFAIR National took over the CA3046
monolithic array from RCA if you can redesign for NPN:

http://www.national.com/ds/LM/LM3046.pdf

.... and Newark has stock:

http://www.newark.com/jsp/search/productdetail.jsp?SKU=41K4644&CMP=AFC-QO1721829242

Sometimes "re-purposing" chips like the MC1496 works but that depends on
what exactly you want to do:

http://www.onsemi.com/pub_link/Collateral/MC1496-D.PDF

These guys at least assure you that they only use devices from the same
wafer:

http://www.diodes.com/products/catalog/list.php?parent-id=28

Hope this helps.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.
From: Michael Robinson on 26 Mar 2010 19:57

"Phil Hobbs" <pcdhSpamMeSenseless(a)electrooptical.net> wrote in message
news:4BACF579.9090208(a)electrooptical.net...
> On 3/26/2010 1:34 PM, Michael Robinson wrote:
>> "Joerg"<invalid(a)invalid.invalid> wrote in message
>> news:8147ltFmnU1(a)mid.individual.net...
>>> Michael Robinson wrote:
>>>> On page 4 of this datasheet
>>>> http://www.nxp.com/documents/data_sheet/BCV62.pdf
>>>> the manufacuterer gives 0.7 to 1.3 for Ic1/Ic2.
>>>> Other manufacturers' datasheets use the same number.
>>>> How many sigmas away from the mean the 30% accuracy statistic is makes
>>>> a
>>>> huge difference in the attractiveness of this current mirror as a part
>>>> I
>>>> want to buy. But the datasheets don't give that kind of information.
>>>> Have you used these current mirrors and can give me an idea what kind
>>>> of
>>>> accuracy could I expect from the actual physical parts?
>>>
>>> It's usually better but they do not guarantee it. Not sure if they'd mix
>>> devices from different wafers but that can make for a huge difference.
>>>
>>> If you need closer tolerance straight out of the gate it'll be a lot
>>> more
>>> expensive:
>>>
>>> http://www.analog.com/static/imported-files/Data_Sheets/SSM2220.pdf
>>>
>>> --
>>> Regards, Joerg
>>
>> The lack of promised accuracy was annoying but I can deal with it.
>> Temperature is another matter. In all the threads I've seen here about
>> current mirrors and dual transistors, nobody was ever able to tell for
>> sure
>> whether the parts were built on a single die. It would be nice if
>> somebody
>> did some more testing of various parts, like that infra-red photo John
>> Larkin just posted. God forbid they should tell you anything useful on
>> the
>> datasheet.
>>
>>
>
> It's not that hard to figure out. For instance, check out Fig 4 of the
> BCV61 datasheet--it shows that for zero emitter resistors, a 1 mA
> collector current gives you 30% error (about 7 mV Vos or 3.5 C) with
> Vce=9V (9 mW) and a 5 mA does the same at about 3V. Either way, it's
> ballpark 200-300 K/W, which is way too high for a monolithic--it's
> comparable to the theta_jc of the package.
>
> Cheers
>
> Phil Hobbs
>
So if I can run 1mW that gives ballpark 500uV offset, or 2% current error.
Thanks, that helps.


From: krw on 26 Mar 2010 23:14
On Fri, 26 Mar 2010 13:34:15 -0400, "Michael Robinson" <nospam(a)billburg.com>
wrote:

>
>"Joerg" <invalid(a)invalid.invalid> wrote in message
>news:8147ltFmnU1(a)mid.individual.net...
>> Michael Robinson wrote:
>>> On page 4 of this datasheet
>>> http://www.nxp.com/documents/data_sheet/BCV62.pdf
>>> the manufacuterer gives 0.7 to 1.3 for Ic1/Ic2.
>>> Other manufacturers' datasheets use the same number.
>>> How many sigmas away from the mean the 30% accuracy statistic is makes a
>>> huge difference in the attractiveness of this current mirror as a part I
>>> want to buy. But the datasheets don't give that kind of information.
>>> Have you used these current mirrors and can give me an idea what kind of
>>> accuracy could I expect from the actual physical parts?
>>
>> It's usually better but they do not guarantee it. Not sure if they'd mix
>> devices from different wafers but that can make for a huge difference.
>>
>> If you need closer tolerance straight out of the gate it'll be a lot more
>> expensive:
>>
>> http://www.analog.com/static/imported-files/Data_Sheets/SSM2220.pdf
>>
>> --
>> Regards, Joerg
>
>The lack of promised accuracy was annoying but I can deal with it.
>Temperature is another matter. In all the threads I've seen here about
>current mirrors and dual transistors, nobody was ever able to tell for sure
>whether the parts were built on a single die. It would be nice if somebody
>did some more testing of various parts, like that infra-red photo John
>Larkin just posted. God forbid they should tell you anything useful on the
>datasheet.

The information is usually somewhere in the datasheet. It may take some
interpretation, though. You can also call an FAE.
From: Gerhard Hoffmann on 4 Apr 2010 06:24
Michael Robinson wrote:

>
> The lack of promised accuracy was annoying but I can deal with it.
> Temperature is another matter. In all the threads I've seen here about
> current mirrors and dual transistors, nobody was ever able to tell for sure
> whether the parts were built on a single die. It would be nice if somebody
> did some more testing of various parts, like that infra-red photo John
> Larkin just posted. God forbid they should tell you anything useful on the
> datasheet.

I just killed an Infineon BCV62C just to watch it's die.

It is definitely 2 chips. Thermal coupling should be _much_ better than
two TO-x glued together, but the one with the wide leg enjoys much better cooling.

regards, Gerhard.
From: Jon Kirwan on 4 Apr 2010 06:40
On Sun, 04 Apr 2010 12:24:58 +0200, Gerhard Hoffmann
<usenet(a)hoffmann-hochfrequenz.de> wrote:

>Michael Robinson wrote:
>
>> The lack of promised accuracy was annoying but I can deal with it.
>> Temperature is another matter. In all the threads I've seen here about
>> current mirrors and dual transistors, nobody was ever able to tell for sure
>> whether the parts were built on a single die. It would be nice if somebody
>> did some more testing of various parts, like that infra-red photo John
>> Larkin just posted. God forbid they should tell you anything useful on the
>> datasheet.
>
>I just killed an Infineon BCV62C just to watch it's die.
>
>It is definitely 2 chips. Thermal coupling should be _much_ better than
>two TO-x glued together, but the one with the wide leg enjoys much better cooling.
>
>regards, Gerhard.

Thanks very much for the effect. I think Phil Hobbs' logic
nailed it for me already, though. It was exactly the right
way to look at the datasheet and immediately made total sense
when he said it. You went out and found what theory said
must be true. Reminds me of Dave and the resistor grid! ;)

Jon
First  |  Prev  |  Next  |  Last
Pages: 1 2 3 4 5
Prev: NXP LPCXpresso demo board
Next: Guitar one octave Up