A Discrete Op Amp Alternative - 990/2520 Adapter

[owel]

Just finished prototyping this little guy tonight. So far, so good with the basic oscilloscope/signal gen/audio analyzer testing.

It's basically a DIP8/SOIC8-Adapter for use as a 990/2520 DOA substitute.
To help the chip you want to use drive more difficult loads, an output line driver stage has been added.

Some comparison readings. Condition: +/-18Volts, 600R load
IC chip only                  :  0.0038% THD @ 4dBu, f=1Khz; 30K BW
IC chip on DOA-Adpater :  0.0029% THD @ 4dBu, f=1Khz; 30K BW

Look at that, THD got even lower!
The output driver helped achieve this.
SNR and SINAD ratio also increased by a few dB.

Some photos:
With a DIP8 IC socket, and using a National LME chip.

You can also use a SMD SOIC8 chip... (haven't tested this configuration yet, I need to buy/salvage some SOIC chips to use in testing.)

This weekend I'll try driving a real transformer with it (and not just a 600R resistor), plug it into a 512C pre... see how it works out.

You can use the trimmer to adjust how much current you want to feed your line driver. (The transistors will get hotter though as you increase it.)
For example: @ +/-18Volts.
PDIP8 chip only : 6mA
on DOA-Adapter: 11mA @ lowest setting  (you can increase this up to 60mA and above... I don't know yet the max. upper limit, don't want to push this 1st prototype too far this early.

** Notice the BYPASS mode? You can solder a chip and a 0R resistor (or jumper) if you don't want to use the line driver output transistors and extra components.
So it will work as a simple straight DIP/SOIC to 990/2520 Adapter.

If this works out, anyone interested in a KIT?

 

[mikefatom]

This is brilliant.  I have adaptor PCBs on the way myself.  No driver stage though.

Cheers,

Mike

 

[user 37518]

very nice!

 

[JohnRoberts]

Just finished prototyping this little guy tonight. So far, so good with the basic oscilloscope/signal gen/audio analyzer testing.

It's basically a DIP8/SOIC8-Adapter for use as a 990/2520 DOA substitute.
To help the chip you want to use drive more difficult loads, an output line driver stage has been added.

Some comparison readings. Condition: +/-18Volts, 600R load
IC chip only                  :  0.0038% THD @ 4dBu, f=1Khz; 30K BW
IC chip on DOA-Adpater :  0.0029% THD @ 4dBu, f=1Khz; 30K BW

Look at that, THD got even lower!
The output driver helped achieve this.
SNR and SINAD ratio also increased by a few dB.

It is not unexpected to see THD improve when driving a low impedance load with a good output buffer stage, it is not logical for the noise to change even a little, since that should be dominated by the input stage. The only hypothetical way for an added output stage to reduce measured noise is if that noise is measured wide band and the added stage reduces HF bandwidth. 

You can use the trimmer to adjust how much current you want to feed your line driver. (The transistors will get hotter though as you increase it.)
For example: @ +/-18Volts.
PDIP8 chip only : 6mA
on DOA-Adapter: 11mA @ lowest setting  (you can increase this up to 60mA and above... I don't know yet the max. upper limit, don't want to push this 1st prototype too far this early.

You might want to look at the 600 ohm THD while trimming the amount of class A current, to see if more current makes any significant improvement, or just heats up the room (while you are already a couple zeros to the right of the decimal point).


JR

 

[MikeClev]

Brilliant work! I've made a couple with vero board as devices for testing projects without risking damage to my precious APP2050s and 2520s. I'll be watching this closely!

 

[owel]

It is not unexpected to see THD improve when driving a low impedance load with a good output buffer stage, it is not logical for the noise to change even a little, since that should be dominated by the input stage. The only hypothetical way for an added output stage to reduce measured noise is if that noise is measured wide band and the added stage reduces HF bandwidth. 

Hey John,

Yes you're right (of course   At 80Khz BW measurement, the SINAD and SNR increased by 1 to 2 dB.
But at 30Khz BW, it's less than a dB difference, about equal for practical purposes. 

You might want to look at the 600 ohm THD while trimming the amount of class A current, to see if more current makes any significant improvement, or just heats up the room (while you are already a couple zeros to the right of the decimal point).

Yeah, I tried that. No change in THD regardless of higher or lower Class A current.
Also no change in THD, regardless whether it's unloaded or loaded with 600R.

Thanks!

 

[owel]

Ok, did more testing...  set the Class A current to +/- 20mA.

Plugged in the DOA (with the National LME chip) into a 512C and it works, no surprise.
The THD of the pre was a little higher running with the LME chip, .0068% vs .0056% with a real 2520 DOA.

But here's the kicker... tried the often ignored, lowly NE5534 chip into the DOA-Adapter, plugged into a 512C (photo below) and repeated the same tests.

Wow!... THD/SINAD/SNR of the NE5534-DOA-Adapter vs. 2520-DOA is for all practical intent and purposes is the same and equal! **

The DOA-Adapter's output driver transistors barely get warm, and absolutely no problem driving the big output transformers.

Moral of the story: Don't ignore the NE5534 chip.

** of course, we only looked at f=1Khz. I have no easy way to graph the whole THD vs. 20-20Khz range automatically and easily.

 

[gemini86]

Okay, perked my interest. 

 

[owel]

AD797 - this chip is famous for being expensive ($10.50 to $12.50 each) and it's "ultralow distortion, ultralow noise" performance. So let's try this in the DOA-Adapter.

And meeehhhh.... I'm not impressed. In fact, it's kinda disappointing and a big let down.

At buffer configuration, it was unstable... THD readings jumping all over the place, the best it can do is about 0.015%.

I'm looking at the datasheet and there's no mention whether it's unity stable or not. Probably not??? It only give design notes for inverting and non-inverting mode.

So I switched to Inverting Configuration and still not impressed.  (I didn't bother writing it down...)

Then tried Non-Inverting configuration, with a gain of 6dB, only then it started giving me decent numbers. 0.066% THD

But compared to the National LME chip and the TI NE5534 chip, I would say the other two chips beat the AD797 easily.

Maybe I'm just not using the AD797 correctly... I think the AD797 needs more care if you want to use it's full potential. It's not an easy drop-in replacement.

Takeaway here: Save your money and look for something else.

PS: The Burr-Brown OPA134 beats the AD797 too.

 

[JohnRoberts]

Not to rain on your parade, but perhaps the noise and distortion is dominated by components in that audio path other than the opamp... Using better and better opamps will stop making a difference when several dB below the other sources of noise or nonlinearity in the components around it.

wrt AD797 the application note shows unity gain configurations so it is clearly unity gain stable. They also suggest the high performance opamp needs substantial PS decoupling. 

I doubt the spec sheet is a total fiction. http://www.analog.com/static/imported-files/data_sheets/AD797.pdf

JR

 

[owel]

Not to rain on your parade,

No not at all.

but perhaps the noise and distortion is dominated by components in that audio path other than the opamp... Using better and better opamps will stop making a difference when several dB below the other sources of noise or nonlinearity in the components around it.

True, BUT I'm not expecting to see the theoretical lowest level possible with the 797 with my equipment... but at least, I was expecting it to be on par with the other lesser expensive opamps.  But clearly, it was performing worse than the others, much worse.

I swap back and forth several times between the OPA, 5534, and LME chip and the 797 chip. Just in case, something went wrong with the equipment or a jack got loose, etc... But the other chips consistently had nice numbers, and only the 797's reading was also consistently the worst of the bunch.

Of course, there is ONE other possibility, that the 797 I have is defective... unfortunately, this is the only 797 chip I have and don't have another one to try out. Or the 797 doesn't like the transistor line drivers following it? I maybe should buy another 797 and repeat the tests.

wrt AD797 the application note shows unity gain configurations so it is clearly unity gain stable.

They do show it unity gain for a Non-Inverting config. But as a voltage follower (unity gain buffer), the datasheet didn't say if this is allowed for this opamp. This is when it shows a .015%THD.

They also suggest the high performance opamp needs substantial PS decoupling. 

They show a 0.1u and 4.7u in parallel. I only have 0.1uf.  This may prevent me from seeing the full capabilities of the 797, but assuming it is more capable than the others, at least I was expecting it to show readings on par with the others.

 

[Samuel Groner]

Maybe I'm just not using the AD797 correctly...

The AD797 requires considerable expertise for stable implementation, and they don't tell you all in the datasheet. With the added phase shift from the discrete buffer, stable operation at high unity loop gain frequencies will be nearly impossible.

The THD of the pre was a little higher running with the LME chip, .0068% vs .0056% with a real 2520 DOA.

They have now about a dozend of these with this prefix, so you can't omitt the part number.

Wow!... THD/SINAD/SNR of the NE5534-DOA-Adapter vs. 2520-DOA is for all practical intent and purposes is the same and equal!

No why's that? Isn't the (whichever) LME chip much better than both the NE5534 and 2520-DOA in every respect?

"Low noise" is a qualification which makes absolutely no sense without considering the source impedance of the intended application. OpAmps have voltage and current noise; at low source impedances, the former is important. At high source impedances, the latter. The use of an input transformer allows scaling of the source impedance as seen by the opamp. The transformer which is inside your preamp is optimised to fit the noise properties (high voltage noise, low current noise) of the 2520-DOA. If you drop in something else with much higher current noise, total noise will be higher, even if the new opamp has much lower voltage noise. No magic here, just Ohm's law applied to noise calculations.

Moral of the story: Don't ignore the NE5534 chip.

Moral of the story: Don't ignore current noise.

Samuel

 

[owel]

Maybe I'm just not using the AD797 correctly...

The AD797 requires considerable expertise for stable implementation, and they don't tell you all in the datasheet. With the added phase shift from the discrete buffer, stable operation at high unity loop gain frequencies will be nearly impossible.

Yeah, I didn't know what I'm doing with the 797. I haven't used this part, just had it on my shelf and dropped it in the adapter. I suspect it doesn't like the transistor buffers following it.

They have now about a dozend of these with this prefix, so you can't omitt the part number.

I'm talking about the LME49710, the single opamp version.

Don't ignore current noise.

Got it!  ;D

 

[Balijon]

I realize this is an old post, but this exactly what I am looking for as a buildingblock.

Owel, will this work with BD139/BD140 & BC550 as well? Or did you specifically select certain transistors for this application?
Did you have some preferences after testing, to what circuit applications work better for you comparing a 5543 to a LME49710?
Are you planning to make pcb's available?
Is this the schematic you used?:

Theo