By request, here’s where I would start on a discrete op-amp using my floating current mirror error correction. I tried to keep the device count minimum so someone could actually fit it on a small PC board/adaptor. The basic signal path is folded cascode, floating mirror, and triple EF output. Bias is a single FET current maker, the reuse of as much of this current as possible makes virtually the entire bias current noise common mode. Almost all resistors are simply scaled from 500 Ohms and series parallel connections of 500 Ohms so a strip would work very well for matching. Offset and gain can be trimmed with input source and drain resistors, so I would not obsessively match devices.
This is a single gain stage amplifier compensated with a single pole at node VA. Using generic LTSPICE 2N4401/2N4403 models and 2SK140’s the Aol was 300K and Vos 800uV untrimmed. The floating current mirror acts as a unity gain current conveyor to the voltage error of the OS across whatever impedance is between VB and the Output, and as stated before this gives a “positive feedback” of one with first order cancellation of the OS Vbe voltage modulation.
Experimenters could play with “VAS killing” by placing a resistor and compensation capacitor to ground at VA and the same network from VB to Output.
With the same generic models I saw ppm distortions at 10kHz 20V p-p into 100 Ohms. This is a work in progress and I hope to take a spare differential VAS board that I made and put this together over the holiday break. There are lots of possibilities for component selection, input stage gm and GBW for different applications.
As usual nothing proprietary here (patent is expired), free to use/not use/abuse. Probably a <5$ BOM here on an all SOT version. All constructive comments welcome.
This is a single gain stage amplifier compensated with a single pole at node VA. Using generic LTSPICE 2N4401/2N4403 models and 2SK140’s the Aol was 300K and Vos 800uV untrimmed. The floating current mirror acts as a unity gain current conveyor to the voltage error of the OS across whatever impedance is between VB and the Output, and as stated before this gives a “positive feedback” of one with first order cancellation of the OS Vbe voltage modulation.
Experimenters could play with “VAS killing” by placing a resistor and compensation capacitor to ground at VA and the same network from VB to Output.
With the same generic models I saw ppm distortions at 10kHz 20V p-p into 100 Ohms. This is a work in progress and I hope to take a spare differential VAS board that I made and put this together over the holiday break. There are lots of possibilities for component selection, input stage gm and GBW for different applications.
As usual nothing proprietary here (patent is expired), free to use/not use/abuse. Probably a <5$ BOM here on an all SOT version. All constructive comments welcome.
Attachments
I would have an impossible task to fit this in a 10x10mm footprint (DIP8 compatible). 🙂
Thanks for sharing,
Patrick
Thanks for sharing,
Patrick
Trout, salmon, seabass... all delicious! I have salmon in the oven as I type this 😉
Another interesting design Scott. Thanks for sharing it! I look forward to simulating this one 🙂
Another interesting design Scott. Thanks for sharing it! I look forward to simulating this one 🙂
It even lends itself to low/non-GFB. Might even try MAT-02's with the inductor bypassed emitter resistors ala the JE990.
I have to pay for those 🙂 But yes they are a fine choice. Not always mentioned the RHOM trannies are some of the best ever in and near saturation so they are superlative in those Leach type MC head amps running on a couple of cells.
Yes I know... Would AD come up with new types at a point in future, like those, possibly in duals?
OK MAT are superior hfe, but lower Cob and hfe would fit better non GFB CCTs, wouldn't it?
OK MAT are superior hfe, but lower Cob and hfe would fit better non GFB CCTs, wouldn't it?
Frankly I'd tweak the THAT folks. They are commited to this stuff and their business plan more suits the pro-summer audio market.
Those THAT folks do make fine transistors.😉
I use THAT 300 series matched arrays quite often. The beta isn't as high, ~90, but it doesn't matter much since they are small current devices. They are ripe on those other good properties.🙂
I use THAT 300 series matched arrays quite often. The beta isn't as high, ~90, but it doesn't matter much since they are small current devices. They are ripe on those other good properties.🙂
Scott,I have a query about the choice of FET.... 2SK140 ? according to my data it's an X band GaAs Fet from NEC and made from purest unobtainium? Was it a typo?
Sorry, 2SK170 standard Toshiba Low noise FET. The LSK389's would do for SOT, even BF862's with "casual" match might work. The first breadboard will try the Toshibas though the high Cgd is not optimal (I want to try and avoid simply cascoding everything). For high level applications there are lots of choices that would still keep noise at a reasonable level.
Thanks Scott, that makes much more sense!
Yes cgd can be an issue with these unless driving from a few ohms source impedance, have had to cascode them in some of my own designs to keep the evil intentions of Mr. Miller at bay!
Can now proceed with sim... should be interesting 🙂
Yes cgd can be an issue with these unless driving from a few ohms source impedance, have had to cascode them in some of my own designs to keep the evil intentions of Mr. Miller at bay!
Can now proceed with sim... should be interesting 🙂
Scott, here's a stupid question from an inexperienced guy. I'm wondering, besides the educational value, which is great motivation for me to do something like this, what other advantages would you say building the discrete opamp would offer? I mean, what could I expect to be excited about rolling my own opamp as opposed to getting the $8 wonders that are available today everywhere?
Some arguments there to discuss about going discrete, in a pro sound sector known standard's literature.
- Status
- Not open for further replies.