Discrete Opamp Open Design

diyAudio Member RIP
Joined 2005
- The selected components are surface-mount devices in 0805 or SOT23 case sizes. This was judged to be a reasonable choice for home construction.
- The specified resistors are 1% thick film parts. Going to 5% tolerance could save roughly US$1.00 in parts cost.

One immediate comment: although 0805 are not as bad as still smaller ones, I've seen horrendous excess (or so-called "current") noise in thick film parts. Rohm ones were among the worst. Although their availability is limited to 0.1% tolerance which is not of any particular benefit here for the most part, thin film parts tend to be very much better. But they are a good deal more expensive.

MELF parts are also fine except for their tendency to roll away :)
 
OK, back on track. My HP triple supply decided to become a 200kHZ power oscillator on the minus side. At first I thought it took out my breadboard but it is OK. On the downside the input stage had several layout mistakes that I had to cut and patch to get it working (and I forgot the offset trim).

The input stage seems to work as expected powers right up at 15mA, I did take the time to match the current source FET's so untrimmed offset is only 5mV. I made it into a closed loop gain of 10 inverter with 50k and 5k and the distortion is close to sims (scary). Now I can put both input and output together.

I did use an LSK389 on the input and I was going to try one with BF862's soldered on the same pads but I will need to spin another board (the cutting and jumpering was a pain).
 
Last edited:
diyAudio Member RIP
Joined 2005
yeah I have to concur, using thick film in a design of this caliber is a waste and you would not expect to get anything like what its capable of.

There are a few circuit positions where thick film noise probably wouldn't be noticed, close to the output for example, but only a few. That noise source bit me badly on a front end of a subwoofer amp once and I did a fair amount of study resolving the issue, and "published" the results on the old prodigy-pro forum some years back. As I recall Panasonic parts were better than some. Susumu seems to be the go-to brand for thin film in readily-accessible distribution.

Although you need current in this case to produce the excess over thermal noise, it's also important to realize that can be signal current, not just "standing" bias current. So your results for dynamic range based on peak signal versus zero signal noise can get quite overly optimistic with excess noise a function of signal.
 
I thought one of the objectives was access inside?
Otherwise we could use an off the shelf op amp.
With the tight spacing of the parts as described, how will DIYer be able to access internal points to experiment?

Yes, already showed simply 50k to ground on the gain node and you get ~60db OLG out to over 20kHz vs ~120dB OLG at a few 10's of Hz. This being one common hobby horse around some parts. :)
 
There are a few circuit positions where thick film noise probably wouldn't be noticed, close to the output for example, but only a few. That noise source bit me badly on a front end of a subwoofer amp once and I did a fair amount of study resolving the issue, and "published" the results on the old prodigy-pro forum some years back. As I recall Panasonic parts were better than some. Susumu seems to be the go-to brand for thin film in readily-accessible distribution.

Although you need current in this case to produce the excess over thermal noise, it's also important to realize that can be signal current, not just "standing" bias current. So your results for dynamic range based on peak signal versus zero signal noise can get quite overly optimistic with excess noise a function of signal.

hey thats a nice little reminder, cheers! and yes susumu RG are excellent parts that we could push the price down on pretty dramatically with a GB

the parts spacing isnt THAT tight, plus a few via test points can make things pretty accessible
 
The attached files describe one possible implementation of the SW-OPA topology.
...
- The PWB's electrical interface uses seven swaged-in-place contact pins. With attention to details and a little practice, a home constructor can successfully install these pins. Videos and written documents showing the process are available on the net.
...
- The physical size is 1.325" by 1.325". This is larger than the "standard" module size (1.125" by 1.125") of the API 2520 et al mentioned above.

Bravo! I haven't looked at the layout yet, but this is a great start.

Comments:

1) Perhaps if dual-transistor BJTs in SOT23-6, SOT363 or SMD6 were used, the outline could be shrunk to the API-2520 form-factor.

2) Similarly, dual-sided SMD layout with the gain stages on one side, and the diamond buffer on the other side, could also help shrink the layout.

3) Apart from the Mill-max gold-plate swage pins suggested elsewhere, Vector K31C/M tin-plated pins of 0.042" diameter, may be a lower-cost and/or more easily available alternative.
 
... resolving the issue, and "published" the results on the old prodigy-pro forum .... As I recall Panasonic parts were better than some. Susumu seems to be the go-to brand for thin film in readily-accessible distribution.

Although you need current in this case to produce the excess over thermal noise, it's also important to realize that can be signal current, not just "standing" bias current. So your results for dynamic range based on peak signal versus zero signal noise can get quite overly optimistic with excess noise a function of signal.
Brad, this would be Noise Modulation and hence more evil than benign Gaussian random noise of fixed level. One could consider Quantization Noise/Distortion a form of Noise Modulation and even small levels were detectable in my 1990's Double Blind Tests even on pop.

Got a link or even a title to the thread on prodigy-pro?

BTW, prodigy-pro.com is now groupdiy.com
 
Proposed Implementation: Comments Requested

The attached files describe one possible implementation of the SW-OPA topology. ... I would appreciate helpful comments on the design, PWB layout, and documentation.
Thanks for this Dale. I hope Guru Wurcer will chip in but here are some of my thoughts ..

- The input stage includes a balance adjustment potentiometer (R27). In a traditional opamp architecture this would probably be labeled as "offset adjustment".
Your R27 source resistor will be VERY twitchy. Scott recommended tweaking R9 instead and sims confirm this is less critical.

- Both the primary and secondary compensation networks are implemented as pole-zero (i.e., series R-C) networks as shown in Post #1568.
Q15-17, 20 are just CCS so there is no need for R13 C3

- The primary compensation node is brought out to an external connection. This may or may not be a desirable feature.
DESIRABLE. You may want C1 off board so you can tweak performance and select where it's earthed. Gets rid of an evil earth pin.

- There is provision to connect the secondary compensation networks to either side of the input differential pair. See my comments at Post #2405 et seq.
Scott actually uses that to add a zero to OL. But this is a tricky tweak. Your method is more conventional and safer.

- There is provision for on-board power supply decoupling (FB1/C4, and FB2/C5).
This is truly EVIL. For something with supersonic response like this, you MUST use VERY careful earthing. It's probably the biggest fault in the original 990.

See AD797 datasheet for some of what needs doing. You have to differentiate very carefully between clean (feedback/compensation) and dirty (decoupling/other crud) earths.

That means get rid of FB1/2 C4/5.

- Most of the PWB bottom side is devoted to a ground plane.
I don't like ground planes for LN audio. But at least make sure there is none under the +ve & -ve i/p tracks to minimize evil capacitance to earth.
 

Attachments

  • res noise.png
    res noise.png
    217.1 KB · Views: 870