ash_dac said:
It is _much_ easier to control the dimensions of vertical structures
by choosing diffusion time rather than controlling horizontal structures
with masks, exposure and etching. This was THE breaktrough of the
planar process.
FET channel length and width usually depend on horizontal structures.
This unlucky fate is shared by lateral PNPs that exist in some low cost
bipolar processes, where the extra diffusion steps needed to produce
nice PNPs are to be avoided.
Gerhard
What about the power amplifier compensation?
Ho Hum...
I have never personally seen any spectrum anaylisis of outputs to the VAS of jfets/mosfet (matched/unmatched - with all it's incipient problems) cascodes versus single bjts with and without current mirrors. Just another ten dimensions are added etc....
I think the multiple Vgs spreads compared with the single Vbe of BJTs means in a differential it is a sub optimal subtractor - much more incorrect from the audio signal. Monolithic op amps are a very different technology.a
There not the same.
What the hell has input noise to do with a power amplifer anyway?
Ho Hum...
I have never personally seen any spectrum anaylisis of outputs to the VAS of jfets/mosfet (matched/unmatched - with all it's incipient problems) cascodes versus single bjts with and without current mirrors. Just another ten dimensions are added etc....
I think the multiple Vgs spreads compared with the single Vbe of BJTs means in a differential it is a sub optimal subtractor - much more incorrect from the audio signal. Monolithic op amps are a very different technology.a
There not the same.
What the hell has input noise to do with a power amplifer anyway?
An even greater problem is with the reproduce noise of analog tape recorders. This is even more difficult than RIAA reproduction, because of the absolute output vs inductance and the lack of high frequency falloff after a 50us time constant or so. In those early days, we sometimes used selected fets, but it was marginal. By 1973, however, we found solutions to this problem and dropped the input noise with fets to less than 2nV/rt Hz. Today, we can do about 1nV/rt Hz differential, and 0.4nV/rt Hz for special function inputs like moving coil.
Back to the Butler input / OP275
Hi,
Butler design:-
http://www.freepatentsonline.com/5101126.pdf
The text is in patent speak but suggests that the BJT works for low input signals, and the FET works for higher input signals.
Is there any value in this approach ?
see also:-
http://www.diyaudio.com/forums/showthread.php?s=&threadid=68156&highlight=
Hi,
Butler design:-
http://www.freepatentsonline.com/5101126.pdf
The text is in patent speak but suggests that the BJT works for low input signals, and the FET works for higher input signals.
Is there any value in this approach ?
see also:-
http://www.diyaudio.com/forums/showthread.php?s=&threadid=68156&highlight=
Christer, the Holman circuit was just a cheap way of doing things. It is bad, because it does not properly cancel the even harmonics. The Butler design starves the bipolars, so that when they give up, the fet pair takes over. This causes FIM distortion i.e. Barrie Gilbert, just like the old fahioned 741, except that you don't actually get to slew rate limiting as fast. This is just what Matti Otala was concerned about in the slew rate vs TIM controversy. TIM measures the actual distortion at all measured levels, but slew rate 'predicts' distortion below slew rate limiting based on a class A model of a bipolar diff pair. The Butler circuit would measure high TIM, YET have a fairly high slew rate. Therefore, just a slew rate spec. is not enough, in all cases.
John, yes I find it hard to see a point in using different types of devices in a diff pair. That seems to defeat the whole idea and all the nice properties of it. As for the circuit I referred to, with a Rush cascode, I have the same concern. But to their excuse I can at least say that i think it migh have been intended as a clever way of avoiding level shifting. That is, if the quiescent Vgs of the JFET cancels the Vbe of the BJT, then the both inputs might have been at the same Q voltage.
I don't know if that was the case though. I only have a reverse engineered schematic from photos and never had a chance to measure it.
I don't know if that was the case though. I only have a reverse engineered schematic from photos and never had a chance to measure it.
Analog Devices, in their newer opamp circuits like AD8033, do use combination of JFETs and BJTs in the input stage. The reason is to get rail-to-rail input common mode performance and to prevent phase inversion. Of course input stage properties do change when it switches from JFETs to BJTs, but it is really better than input stage collapse. Especially for lower supply voltages like +/-5V one will appreciate this feature. The 8033 is my favourite for input stages for measurement purposes. The 8033 is 80MHz, 80V/us JFET input opamp.
john curl said:
Well a Rush cascode is not really a cascode, but rather very similar to a diff pair if you look at the equations although looking quite different in a schematic. I had never heard of it either until I once posted this schematic of the amp I was referring to, and Hugh Dean told me that this type of input stage is called a Rush cascode. Although it normally consists of two complementary BJTs, not a JFET and a BJT as in this case:
http://www.diyaudio.com/forums/attachment.php?s=&postid=130914&stamp=1045587290
This guy uses a "normal" Rush cascode in the third figure, although not giving any name for it.
http://www.4qdtec.com/opamp.html
Here is Hugh's educating comment on the topology
http://www.diyaudio.com/forums/showthread.php?postid=131231#post131231
And BTW, here is how NAD used it in their phono stage:
http://www.tcaas.btinternet.co.uk/nad3020pre.gif
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.