DOA - Discrete Op Amps

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Composite Tube-Opamp dual Dip8

well it’s official, i have the doa bug.
i’m doing it with a classic opamp mounted/combined with sub miniature triode.

Edit 10-2018 : I've learned a great deal during this project and now taking that knowledge to the next level.
So there will be no more updates from me on this fantastic concept. Thanks to Brent Butler for publishing it in 1998.
-Bruce
 
Last edited:
I'd call that a composite amplifier, which I think is actually a wiser idea than to insist on 'discrete only', a true DOA. There are some things, like current sources and mirrors, thermal coupling, and close device matching, that are much easier to do with monolithic ICs than with a purely discrete amplifier. That's generally how 'design purity mandates' often work with engineering - religious purity often compromises the design and performance.

I too always think of 'dead on arrival' when I see DOA. Must be a generational thing!

Have fun!
 
this is a pic of working ckt. its not yet stacked for dip plugin.
The output is classA from tube but the opa445 is servo.
 

Attachments

  • 42D8A73F-58CA-467C-900A-5CBEB4640ED0.jpeg
    42D8A73F-58CA-467C-900A-5CBEB4640ED0.jpeg
    630.4 KB · Views: 1,771
I must be doing something wrong, have tested numerous fet input parts and they all are very difficult to stabilize. distortion always increases with frequency but the ne5532 requires no compensation , verrry low dist across audio spectrum too at least in topologies i’ve tested so far. the fets need inner and outer loop compensation and would i’m sure smear upper register music.
say it ain’t so please
 
sorry here is some data from earlier testing with ne5532

Nice curves, but is there a way to do an average of those waterfall curves, and not a waterfall? There's no time or decay component, so the extra time dimension isn't as useful as is taking the average of all of those curves. That will help you to resolve low level details better and provide less error on what you measure. Not certain if that software is up to it, but a friendly suggestion nonetheless!
 
I must be doing something wrong, have tested numerous fet input parts and they all are very difficult to stabilize. distortion always increases with frequency but the ne5532 requires no compensation , verrry low dist across audio spectrum too at least in topologies i’ve tested so far. the fets need inner and outer loop compensation and would i’m sure smear upper register music.
say it ain’t so please

If you're using a non-inverting amplifier topology, there is an effect found with junction isolated JFET input stages that is worth considering. Basically, the input common mode signal voltage modulates a parasitic drain to substrate capacitor on each input FET, adding distortion that rises at 6dB/octave. These papers explain it in more detail:

http://www.analog.com/media/en/technical-documentation/application-notes/742022599AN232.pdf

http://www.ti.com/lit/an/slyt595/slyt595.pdf

These effects won't happen with an inverter, since the inverter topology results in no common mode signal voltage applied to the input stage, and thus no modulation of this parasitic nonlinear capacitance, so it doesn't tickle this form of distortion. However, most people who use JFETs do so for high impedance non-inverting stages, and thus encounter this problem.

There are other reasons for an amplifier to misbehave with common mode input signals, so in general, if you can use an inverter, your circuit will be much cleaner.
 
I had no clue. Thanks , going to re-test as inverting. That limitation should be in the data sheets.

Here is a pic on one i tested this morning. Its a basic CF with feedback from cathode out to ne5532 non-inv input. 1/2 of 12au7 is used (as output). Distortion is almost at noise floor of my setup. Really only one resistor is required, the cathode load. ferrite bead is on gate. +-15 to +-24 works. The 180 ohm resistor is a protect for sig gen. I’ve destroyed two of the subminiature tubes while testing opa445, tl071 and Op275.

I been doing discrete hybrids a long time. This is new for me using opamp servos.
 

Attachments

  • 6A81EC3B-0193-4844-AAC3-812FB109061D.jpeg
    6A81EC3B-0193-4844-AAC3-812FB109061D.jpeg
    603.3 KB · Views: 1,061
Glad that things are working more predictably now!

BTW, I didn't look closely before, but it's quite the sight seeing a subminiature tube plugged into a proto board! Interesting stuff - cheers for getting that to work!

Proto boards are very nice, but beware that each "strip of 5 holes" has a couple of pF of capacitance to its neighbor. This isn't always a problem, and it can often be compensated for, but that capacitance, when applied between the op amp inverting and noninverting inputs can sometimes destabilize a fast op amp. I suspect that you've probably compensated for this with capacitance around the feedback resistor, and that will work well most of the time. With a current feedback amp, you won't have any way to compensate for that capacitance, but with a voltage feedback amp, you can get away with compensation and usually not exceed the amp's overall output capacitive load limit, since the proto board's strays aren't super large.

Again, glad things are working out well, and have fun!
 
UR22 based audio analyzer

Nice curves, but is there a way to do an average of those waterfall curves, and not a waterfall? There's no time or decay component, so the extra time dimension isn't as useful as is taking the average of all of those curves. That will help you to resolve low level details better and provide less error on what you measure. Not certain if that software is up to it, but a friendly suggestion nonetheless!

I did a set of external hardware loopback runs, the file names in zip file file detail settings. The waterfall previous was 250ms avg. I did a 125ms and 250ms waterfall here. Also thd, and freq response. The system is OSX High Sierra, Yamaha Steinberg UR22 USB sound interface, external cables line 1/4 jack to BNC. The clip leads are all connected together at cable ends. 96khz 24 bit word. Let me know if questions on settings or problem with zip file.
Thanks

-
 

Attachments

  • UR22-24bit-96khz-THD-loopback-floors.zip
    249.3 KB · Views: 75
Well did more testing today with the goal of replacing all op-amps in my Hafler DH100 pre-amp with DOA/Composite type 8Dip packages each utilizing a 12au7 'riding high'.
Here are some pics of the various op-amps tested. A 12ax7 works fine but less output swing unless cathode resistor is matched. I get a minimum of 15Vpp out with the 12au7 running at +-15V. I tested gains of 1,5 and 10.
File names indicate which op-amp is used. Blue line is system baseline and magenta is DUT.
 

Attachments

  • Higher-Input-w-2nd-3rd.png
    Higher-Input-w-2nd-3rd.png
    57.1 KB · Views: 113
  • Normal-Input.png
    Normal-Input.png
    58 KB · Views: 125
  • FreqResp.png
    FreqResp.png
    39 KB · Views: 123
  • 12au7-12ax7-opamp-mix.png
    12au7-12ax7-opamp-mix.png
    290.9 KB · Views: 134
  • 12au7-g5-op275.png
    12au7-g5-op275.png
    17 KB · Views: 131
  • 12au7-g5-opa445-jade-adapter.png
    12au7-g5-opa445-jade-adapter.png
    22.2 KB · Views: 752
  • 12au7-g5-ne5532.png
    12au7-g5-ne5532.png
    17 KB · Views: 783
  • 12ax7-g5-opa2134.png
    12ax7-g5-opa2134.png
    16.7 KB · Views: 803
Last edited:
in the pic on previous post containing four plots, the OPA2134 has been found to be fake. Lacquer thinner used to remove fake laser marks.
Odd they turned out to sound great and are installed on some of my hi end eqpt. I've been using them for some time. No fails yet.
 

Attachments

  • 810E8BE4-2B55-418F-81AE-395B1E8C74D5.jpeg
    810E8BE4-2B55-418F-81AE-395B1E8C74D5.jpeg
    375.7 KB · Views: 366
Been running a few hours now, must be in heaven. I have some select tracks for live auditions. Midnight in the Garden of Evil soundtrack is one listening now. Incredible definition now , sax , snare ,vocals. I'm hooked really on DOA ( composite op-amps?). I'll put up schematics and notes after its been tested on a few more units. Too much fun.

Shown in DH100 using stock board, no changes, same ps but you can see wires for triode heaters. The original DH100 op-amp is TL072 on rev 1 units and ne5532 on later models. The prototype is in IC4 active 'tone' circuit of dh100.
 

Attachments

  • dip8-triodes.jpg
    dip8-triodes.jpg
    34.4 KB · Views: 446
I like the stuff you are doing here. It is cool, and is what DIY is all about. Valves ain't my thing, but "roll your own" certainly is.

However, sorry to be an "engineer" at you...

All the distortion plots you showed a couple of posts back are fine - but do show distortion levels from your composite valve / opamp that are an order of magnitude higher than the op amps would present by themselves.

I think it is cool you are playing with this. But in many ways, your concern neednt be about distortion, more than a concept or philosophy.
- If it were all about distortion, go the op amps.
- As a concept / philosophy, bugger the numbers, make it cool / sound nice / make you smile.

OR, if it were me, I would take a discrete (discrete) op amp and use a valve as the output class A emitter follower.... thought of that?

There are plenty of op amps out there, look at my BLOG if you want to try some, if you want I could even send you some of the multitude of DOA's sitting on my shelf in which you could turf those pesky transistor output devices.
 
Dont kno. It sure sounds great.
The ur22 has a thd floor of .005 - .003, so i would need other test equement for a real dist evaluation.
But you are very correct , its not only distortion that defines good sound.

So this example uses a nullator (op-amp) of your choice and the norator ( a triode in class A).

The nullor block is 8 pin dip housing (twin sections).

This is the more demanding challange compared to a single opamp or one of the pro audio module footprints.

A single stage triode cannot achieve thd levels ( in previous plots) without the op-amp.
An op-amp alone cannot sound as musical as a triode.

Next I'll do a novel 9 pin type prototype (12au7 ) etc but wanted to do the twin minature first.


-
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.