Discrete Opamp Open Design

Been quiet too long.... What's the status?

FET990 derivative, in DIP8 15mm x 15mm (~0.6" x 0.6"), double-sided SMD (components on both sides, bottom (pin) side not shown). This is the first hand-assembled prototype, built without the help of a illuminated magnifier and the board holding-clamp, so many of the parts are misaligned. However, it still works fine, thanks to the large track clearances (which wastes space, but makes it easier to assemble). This was a validation run, and there may be another re-spin that may include the offset/balance pins that were not included in this prototype to save board space.

A through-hole version of the FET990 will be ~4x the board area (through-holes occupy a lot of space that could be used for routing and components, and components can only be mounted on one side - each of these contributes a ~2x multipler to the board area). So we're talking about 30mm x 30mm (~1.2" x 1.2"). IMHO, that's too big for most DIP8 sockets. The alternative is a mezzanine construction with the input stage on one one board, and the VAS/output stages on a second board mounted above it. It's still going to be about 0.9" x 0.9" - still too large for DIP8. SWOPA has about 2x the component count of FET990, so add another factor of at least 2x to the board area.

Please consider all these factors before committing to a through-hole design, especially SWOPA. IMHO, the more practical DIP8 implementation is the SMD FET990 as shown, with SOT23 actives, SOT89 outputs and 0805 passives. Some space can be squeezed out by going with SMD duals where possible, 0603 passives, tighter clearances, and SOT23 outputs - say 5-10% area reduction.
 

Attachments

  • lf06s_v1.2_bare_9x9.jpg
    lf06s_v1.2_bare_9x9.jpg
    294 KB · Views: 1,027
  • lf06s_v1.2_top_9x9.jpg
    lf06s_v1.2_top_9x9.jpg
    306.9 KB · Views: 998
Hi linuxguru,
Your PCB looks nice as it is.
What is it's schematic?
Will PCBs be available?

The bare PCBs are nice - manufactured by the best vendor in India, with electro-less nickel immersion gold, tight tolerances, etc. It is my hand-assembly that is sloppy.

Schematic is as shown below - some values/components may change.

Bare boards are available right away, kits and completed assemblies will take some time (matching of the input JFETs is mandatory to get the input offset voltages under control, and I haven't done that yet on the SMD JFETs).
 

Attachments

  • lf06s_v12.jpg
    lf06s_v12.jpg
    73.2 KB · Views: 963
Last edited:
Hi linuxguru,
Thanks.
I'm trying to figure out the basics of your circuit. It seems that J1 (cascoded by J4) and J2 (cascoded by J3) are LTP, while the gate of J2 is connected directly to the output. That seems like a huge amount of GNFB, or am I wrong?
Also, there is only (+) input, no (-) one, so it wouldn't be a substitute for any OPAMP in any circuit.
 
diyAudio Member RIP
Joined 2005
Hi linuxguru,
Thanks.
I'm trying to figure out the basics of your circuit. It seems that J1 (cascoded by J4) and J2 (cascoded by J3) are LTP, while the gate of J2 is connected directly to the output. That seems like a huge amount of GNFB, or am I wrong?
Also, there is only (+) input, no (-) one, so it wouldn't be a substitute for any OPAMP in any circuit.
It's hooked up as shown as a voltage follower, with maximum loop gain, but unless that J2 gate is hard-wired it will otherwise serve as a negative input, and you can introduce a feedback divider, or ground J1's gate and configure for inverting mode, etc.
 
In he above schematic, the gate of J2 is had wired, internally, to the output.

To clarify, the schematic is an LTSpice simulation schematic of a voltage follower using the FET990/LF06s. Just cut the net between the gate of J2 (inverting input) and the output node to obtain the LF06s opamp schematic, as Brad noted above. The output isn't hard-wired to the inverting input in the implemented module, and it does have all the opamp interfaces (IN+, IN-, OUT, COMP, +Vs, -Vs) found in a typical opamp. Iit has been plugged in and tested in place of an LM318 in a MyRef Rev E (where it works much like an OPA627 or LT1028 - it triggers SPiKe protection in the LM3886TF chipamp at moderate output levels). In other words, it should be a plug-in replacement for any bipolar or JFET-input DIP8 single opamp if you take care of compensation, rail voltages, supply current requirements, bypass, etc. It doesn't have output short-circuit protection - the hefty 2A Toshiba output pair, plus the over-engineered thick traces from the rails to output pair and output node can smoke things if you aren't careful about overloads/shorts - it can comfortably fry a pair of low-Z headphones if you try to drive them rail-to-rail.

The parts designations between the LTSpice schematic and the Eagle schematic/layout don't necessarily match - I tend to relabel the Eagle schematic and layout during the layout process and I often forget to go back and make the LTSpice schematic consistent, so that's the reason for the discrepancies. However, exactly the same values and parts shown in the schematic, have been used in the first prototype.

I can supply a full kit of all the parts except the matched input JFETs (any of 2sk209, BF862, MMBF4393, etc. should work there),. The matched JFETs will have to wait until I can rig up a measurement fixture for the SMDs.
 
Last edited:
SWOPA Thru-Hole Build Package Rev B

The attached files are a revised version of the build package posted at Post #2885 in this thread ( http://www.diyaudio.com/forums/anal...screte-opamp-open-design-289.html#post3470900 ). Except as noted here, all the comments in Post #2885

Significant changes:
- On the PWB, default trace width was increased to 25 mils from 15 mils. Selected traces in the output section were further increased.
- Locations of some components and traces were adjusted to maintain required clearances.
- With the exception of the TO92 and SOT23 mounting patterns, and traces passing between the pads in these patterns, this PWB complies with 15 mil trace width/15 mil trace spacing design requirements.
- Reference designator prefix for the three wire jumpers was changed to "JP" from "JU". ("JP" seems to be the more common industry usage.)
- The wire jumpers were added as line-items in the Parts List/BOM.
- LTSpice simulation files were added to the compressed archive.

Dale
 

Attachments

  • Post2885_Docs_Edited_Part2_B.zip
    587.6 KB · Views: 132
  • Post2885_Docs_Edited_Part1_B.zip
    868.5 KB · Views: 146
  • SWOPA2885PWB_B_ALL.pdf
    603.2 KB · Views: 188
  • SWOPA_Post2885_Parts_B.pdf
    202.3 KB · Views: 137
  • SWOPA_Post2885_Schem_Parts_B.pdf
    366.5 KB · Views: 273
  • Post2885_Docs_Edited_Part3_B.zip
    858.7 KB · Views: 119
To clarify, the schematic is an LTSpice simulation schematic of a voltage follower using the FET990/LF06s.

The parts designations between the LTSpice schematic and the Eagle schematic/layout don't necessarily match

Thanks.
Can you please post the Eagle schematic?

I can supply a full kit of all the parts except the matched input JFETs (any of 2sk209, BF862, MMBF4393, etc. should work there),. The matched JFETs will have to wait until I can rig up a measurement fixture for the SMDs.

How much is the kit and what does it include?
 
Can you please post the Eagle schematic?
How much is the kit and what does it include?

Schematic attached - please PM me for the details of the kit (which will include the ENIG-finish board as shown in post #2901, all actives except the matched input JFETs (I can provide unmatched JFETs right now), all passives, and the gold-plated pin set).

Edit: The schematic shows a dual PNP Q4 for the current mirror, but the final layout that was fabricated actually uses separate SOT23 PNPs.
 

Attachments

  • lf06sv12_sch.png
    lf06sv12_sch.png
    7.5 KB · Views: 1,050
Last edited:
Linuxguru: nice! Do you think +-23v operation would be feasible? If so I would love some kits :)

Thanks. As shown, +/-23v operation is outside the Vds max of the cascode FETs, the VAS current source, and possibly the VAS buffer (bc549c). You'll need 50V-rated JFETs (there are some Sanyo/OnSemi SMD parts that may fit the bill) for the cascode and VAS CCS, and bc560c or any PNP rated at 50V or higher for the VAS buffer. The output BJTs, the VAS and the current mirror are fine for +/- 23V operation.
 
I'm totally late, so can someone please summarize what design(s) have been chosen, and why?

There are several parallel prototyping efforts:

1. Dale (dchisholm) is focussed on Scott's SWOPA with API2520 or larger footprint.
2. I have built working FET990 prototypes as well as the folded-Kaneda/Constellation,
with focus on compact DIP8 plug-in replacements in SMD.
3. Richard (kgrlee) is also focussed on FET990.
4. Several other designs remain in simulation.

(Any others in various stages of prototyping/simulation may please be added to the list).

IMHO, SWOPA will be practical in larger formats including professional API2520 or similar form-factors, while DIP8 form-factors may require a simpler topology like FET990 or something else. SWOPA in a stacked DIP8 is possible in principle, but I don't know of any effort in that direction yet (except maybe Scott's initial prototypes).
 
Member
Joined 2011
Paid Member
Is the OP still around?

Here's how I obtained the answer to that question: First, I went back to post #1 in this thread. Next, I clicked on the username of the member who made post #1 a/k/a, the Original Poster ("EUVL"). Finally, I clicked on "Find more posts by EUVL". The diyAudio software returned them chronologically, most recent post first.

I screen-grabbed the most recent post; after reading it, I think the answer to your question is: Yes.
 

Attachments

  • OP_click_name.png
    OP_click_name.png
    138.3 KB · Views: 764
3. Richard (kgrlee) is also focussed on FET990.
I don't think I'm so much 'focussed on FET990' as attempting to achieve Guru Wurcer's SWOPA performance with a much simpler circuit.

In SPICE world at least, the circuits in #2060, #2321 & #2323 achieve this and in most cases, the THD residual is less spiky; even with less Iq .. ie less xover. Guru Wurcer points out it is possible to Muntz these even further without losing performance.

#2062, #2083, #2183, #2306 have more and include comparisons with SWOPA.

The thread from #2324 - 2350 has some interesting discussions.

BTW, my main illustrative performance figures are at 10Vp @ 600R with 40dB gain ... and voltage follower 3.162Vp @ 15R.

In #2801, linuxguru demonstrates enhanced voltage follower performance up to near clipping. To match this, the i/p cascodes become u309.

keantoken also demonstrates elsewhere that bc550/560 are best at low Vce.

Hence, my present favourite circuit is as attached though I've not touched it for 3 mths and need to apply some of the tools I've learnt to use in the intervening period .. eg Tian probes & the THD Analyzer.
_______________

2 factors give this simple circuit performance rivalling even Guru Wurcer's SWOPA.
  • pure Cherry compensation
  • keeping the Holy HiZ VAS output clear of evil stuff like clamping diodes. The remaining distortion in the JE990 is almost all due to evil modulated C of the diode.
There was some talk about bootstrapping the evil clamp diode but no one has posted any results.
 

Attachments

  • 990-9.gif
    990-9.gif
    48.1 KB · Views: 573