Discrete Opamp Open Design

Disabled Account
Joined 2012
SW discrete opamp BF862 (post 135).

I have put in bootstrapped cascodes on the front end and 5401/5551 as the cascode BJT's in the "VAS", so there is no problem using the circuit at higher voltages.

As you can see in the attached file I have stepped the source resistance from 50ohm to 2kohm and there is almost no change in THD.

It is possible to improve the circuit, but it is rather pointless to design a perfect front without using a relevant output stage.

So in order to move forward we must, among other things therefore agree on the load.
600 ohm or 30 ohm?

BTW: Maybe the circuit needs a better name than "post 135" or maybe not?

Stein

IC opamps can do 600 ohms. Make it do what a common IC cant do alone and be more flexible.... 30 Ohms. Sure, it's brutal and tough. And can drive anything you need for it to do. Otherwise, why do this if we can get same results with an existing 5532 IC? You want bragging rights? Then do 30 Ohms.
[But then it doesnt matter as the output devices can be replaced by beefier devices at any time. Just need the topology for extending power if needed.]
Name? DOA (Discrete Op Amp)
 
Last edited:
Disabled Account
Joined 2004
opamps can do 600 ohms. Make it do what a common IC cant do alone and be more flexible.... 30 Ohms.

and +24dBu, plus I already said there are plenty of IC's that drive 30 Ohms. Dick, I sent you a rerference on how those CFP outputs can go class B switching I consider it a problem that needs to be rectified.
 
Last edited:
Disabled Account
Joined 2008
And in a DIP8 footprint. Well, after all, no one has yet suggested a height restriction :eek: Or a requirement on airflow.

That was my next question. :D

Anyway 600 ohm was my target.
When I use my own favorite output stage it looks good down to about 150 ohms.
The picture shows THD with load stepped from 300 - 1200 ohms. (at 40V pp)

Stein
 

Attachments

  • DOA 300-1200.jpg
    DOA 300-1200.jpg
    59 KB · Views: 489
Disabled Account
Joined 2012
and +24dBu, plus I already said there are plenty of IC's that drive 30 Ohms. Dick, I sent you a rerference on how those CFP outputs can go class B switching I consider it a problem that needs to be rectified.

You did? I missed seeing the reference. I'll look into it if I find the reference... gimmie it again pls. The circuit (non-sim) in real life hasnt that issue that I have found... so I'll see what the conditions are and if they are reasonable for audio app/use. Then tweek it, if necessary. Partly because of the way i built the supplies perhaps? Let me see. I know you could fix it asap - wanna tell me or make me suffer thru it :) Or, is it something inherant in the topology?

IC's that drive 30 Ohms -- maybe even 8 ohms. I have seen high voltage opamps for decades as well. Are they sota in distortion and noise as well? Comparable cost to your discrete? Analog? Common opamp IC? Low cost? Exotic? which ones?

So we should at least do 30 ohms like the other IC's?

Or was the goal to fit in a dip space like the JE990 could? I dunno. But I wouldnt be interested in a discrete if for the same performance and approx cost could buy an IC.

I think I see the light at the end of the tunnel. I hope it isnt a train.
-RNM
 
Last edited:
Disabled Account
Joined 2012
Discrete vs IC performance

and +24dBu, plus I already said there are plenty of IC's that drive 30 Ohms.

Scott - I am refering to a 'common' IC like the popular 5532/34 type. Not a special IC opamp developed for high voltage or high current use. Like a OPA541 in a 125W TO3 package. A discrete design ought, IMO, do at least better than what can be done in a 'common' audio IC. Or as John Curl said, otherwise just buy the IC. -Dick

[I am not saying there is anything magical or special about doing 30 ohms -- it is a useful audio Z to design for. When done with development here, it will be something to brag about and might find a long and varied life in many products for high-end for years to come. Now i'm not going to say designing for more current output than a 600-100K load needs has any audible benefits,either... does it?]
 
Last edited:
if you can live with < ~ +22 dBu then DSL driver chips are fine output stages

the TPA6120/THS6012 has audio specs, will put out 400 mA so the +22 dBu could be into 30 Ohms too

smaller than the 2 pairs of TO-126 output Q (not counting necessary added bias, current limit parts) you'd probably use in discrete to get the same current in 2 channels

I know you're going to dismiss the very idea but maybe you should have an idea what the monolithic competition performance level really is today
 
Last edited:
Disabled Account
Joined 2012
Not a chance that the JE990 fit in a DIP space.

Erik's paper using a current feedback pair (adding a darlington) for high current output.

http://www-f9.ijs.si/~margan/Articles/Class_B_Dist.pdf

I see what you mean -- Class B switching --- or, crossover distortion. Ok. Got it. Probably just throw another bias diode at it. But, the topology is pretty nice in this article. Thermally stable. I'll look at error signal with a scope and see. The level should be pretty low -- the circuit I applied it to had modest gnfb and the THD + N is extreamly low. I wont be able to measure improvements until the test gear gets better. Which I am working on. BUT, because you say its an issue, I'll look at it further.
Which is better IYO - Erick's or the Diamond buffer for an output stage? Thx - RNM
 
Last edited:
AX tech editor
Joined 2002
Paid Member
Not a chance that the JE990 fit in a DIP space.

Erik's paper using a current feedback pair (adding a darlington) for high current output.

http://www-f9.ijs.si/~margan/Articles/Class_B_Dist.pdf

Thanks for the link Scott - a very interesting article.
I instantly recognised the graphs from time way back when I fist read it but then forgot about it.
The mind shudders to think how much more I have forgotten..:mad:

I'll ask Erik if I can post it online.

jan
 
I agree with Scott that one circuit doesn't fit all apps. In one of Walt Jung's articles he expounds the virtues of virtually noisless mosfet constant current sources. John Curl has posted some all fet/mosfet topologies. I certainly think that this effort should be, in it's final form, something that can't be accomplished with a chip op amp. I believe the output stage should operate Class A with no chance of ever getting out of Class A. I think the bandwidth should extend to 100Khz. I think the completed circuit should be able to produce a 40 volt peak to peak square wave at 100Khz with no ringing or overshoot and sit there and cook for an hour even if you have to use heat sinks. The 2n5551/2n5401 transistors have been used successfully in several other modern attempts. It should be made with readily avalable parts not obsolete or scarce or so new it is unobtainium. Ray
 
Disabled Account
Joined 2004
Pete Basel had a thread 5 years ago on the JE990 with some interesting research. Maybe one good idea would be to do a simple FET version since it seems the FET version is a trade secret.

In the thread there is a discussion about output devices and bias. If you look at the schematic the MJE devices are selected. In Pete's sim he uses stock 1N914 and Motorola models for the outputs. There is little hope the values for Is in the models will give exactly the right bias current. I would always provide for trim of the output stage bias, or just to get a sim to work, adjust the 1N914 Is until the output devices have the right current.

http://www.diyaudio.com/forums/solid-state/107404-simulation-je-990-op-amp-deane-jensen.html
 
Disabled Account
Joined 2004
Here is the first output buffer that was simple enough without compromising performance. I'm still looking at G = 10 and up performance because there is a some fine scale work needed to insure stability at low gains. Also as bcarso points out there might be an opportunity for power on latch-up so there might be a simple clamp needed somewhere. I don't think this is a big problem.

The output stage is biased all the way back from the input (via Q15 and Q16). If you follow the signal path this gives a feed forward slew boost directly to the output, unfortunately this only works in both directions if one of the fully differential input stages is used.This idea appears in many current feedback amps and this single gain stage approach provides the same opportunity. I will eventually repost all three ideas for input stages including this output stage. The other two versions were the Rush cascode and the complimentary version with hard to get FETs.

All versions rely on the input biased at 5mA per side. R8 provides trim of output bias which should be at 8mA. This gives a total supply current of 25mA. R19 should be chosen to give ~1mA for the supply voltage used. The values of R7 and C1are open to experimentation to vary GBW product and slew rate.

Nothing is chiseled in stone I'm sure there are numerous choices of devices that will work. At this waystation things look pretty good for G = 10 and up, 600 Ohm and 1000pF drive. Considering all the diode connected transistors could just be diodes there are only 16 or 18 active devices needed.

Attached are 20k 20V p-p THD and 100k square wave response (600 Ohm load)
 

Attachments

  • cir1.jpg
    cir1.jpg
    96.1 KB · Views: 707
  • cir2.JPG
    cir2.JPG
    59.5 KB · Views: 635
  • cir3.JPG
    cir3.JPG
    45.3 KB · Views: 577