Trafo pre for high fidelity, low noise and high gain

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

dbelousov

Well-known member
Joined
Jul 1, 2019
Messages
113
Location
Moscow, Russia
I had some thoughts yesterday an drew this. Ignore the component values, obviously.

I'm not a big fan of transformers, but sometimes there is no other way to achieve some goals. I believe a good high ratio trafo and a low noise FET op-amp will get me there, so I'm trying to get maximum out of this approach.

This is an InAmp variation with servos and an input transformer. With precise op-amps the offset should be low enough for direct coupling to any equipment. A trafo, a double op-amp for the input stage and a quad op-amp for the servos and the output.

So, should I give it a try?
 

Attachments

  • Trafo Preamp.pdf
    131.4 KB · Views: 74
It's always good to build stuff IRL, always teaches me something real.

What is it with the 1K/10pF filters - do you need this 16MHz filtering for some particular reason? I'd make it orders of magnitude lower..

The 1K's at output - why?

If you need transformer balanced, but clean-clean, perhaps look at zero-field input configurations?

Jakob E.
 
gyraf said:
What is it with the 1K/10pF filters - do you need this 16MHz filtering for some particular reason? I'd make it orders of magnitude lower..

The 1K's at output - why?

dbelousov said:
Ignore the component values, obviously.

I just copy-pasted components according to assigned footprints, gonna think about the actual values later, according to the chosen op-amps capabilities. There is why all ceramic caps are 10 pF, etc. It's just a concept now.



gyraf said:
If you need transformer balanced, but clean-clean, perhaps look at zero-field input configurations?

Oh, thank you, I didn't know about such thing. Googled it, looks promising.

What I'm trying to achieve is to make a very low noise mic pre for "weak" sources. I've built dramatically overengineered, SSM2212 based ultra-low noise, lightening fast and precision transformerless preamplifier already :) So I believe I pushed this approach close to its limits. It's time to try transformers.

 
dbelousov said:
I had some thoughts yesterday an drew this. Ignore the component values, obviously.

I'm not a big fan of transformers, but sometimes there is no other way to achieve some goals. I believe a good high ratio trafo and a low noise FET op-amp will get me there, so I'm trying to get maximum out of this approach.

This is an InAmp variation with servos and an input transformer. With precise op-amps the offset should be low enough for direct coupling to any equipment. A trafo, a double op-amp for the input stage and a quad op-amp for the servos and the output.

So, should I give it a try?
Why not, it's always good learning experience, however I would question two things:
First, the input transformer. It's extremely difficult and costly to find transformers with a secondary impedance higher than about 20 k.
With that ratio, most bipolar opamps will give a noise performance comparable to FET opamps.
What LN FET opamps do you intend to use. AFAIK the quietest FET opamps are about 4nV/sqrtHz. The major noise contributor will be the 200 ohm source resistance.
Using 5534's results in a very similar situation, where noise from the electronics is about 1/3 of the source noise.
Actually, a lower ratio xfmr coupled with ULN bipolar opamps will result in slightly less noise, yet I doubt this difference is audible.

Now, why the servos, when the transformer cannot pass DC, and anyway audio excludes DC?
What's wrong with a good, well dimensioned capacitor?
 
I think about something like Jensen JT-115K-EPC, 1:10, PCB-mounted. Yes, it has 15k secondary. And I live in the country where finding transformers is extremely difficult and costly. Any audio transformers.

The quietest FET op-amp I could find is the ADA4625, 3.3nV/√Hz at 1kHz, and 3.6nV/√Hz at 100 Hz, it's quite impressive. The price is impressive as well.

The quietest Bipolar op-amp is the ADA4898 I suppose: 0.9nV/√Hz at 1 kHz, 2.4pA/√Hz at 1 kHz, astonishing 1.2nV/√Hz at 10 Hz.

I have a stash of 2sk170bl, but I don't want to rely on the obsolete part.

Thank you very much for the noise considerations. It's tricky for me to calculate things like this, what dominates at what level, etc. I have to learn about it more, definitely.

The servos are the most arguable thing. Now I am at "just because I can" and "let's try to do it this way" stage with them.
 
dbelousov said:
I think about something like Jensen JT-115K-EPC, 1:10, PCB-mounted. Yes, it has 15k secondary. And I live in the country where finding transformers is extremely difficult and costly. Any audio transformers.
Transformers have losses, that do not allow them to provide the whole benefit one can expect by perfectly matching the actual source Z to the OSI (Optimum Source Impedance) of the circuit.
The level of quietness of dedicated circuits by THAT or SSM makes it hard to beat.

I have a stash of 2sk170bl, but I don't want to rely on the obsolete part.
The LSK170 equivalent is current.
 
abbey road d enfer said:
Transformers have losses, that do not allow them to provide the whole benefit one can expect by perfectly matching the actual source Z to the OSI (Optimum Source Impedance) of the circuit.
The level of quietness of dedicated circuits by THAT or SSM makes it hard to beat.

I have a THAT 151x premp with the OPA1656 on its output which is a DC servo for itself. Works fine, sounds good, very straightforward schematic. The best bang for the buck, really. I use it as a "portable" device.

My primary recording preamps are Soundraft Delta's. Oh, they are tremendous on the drums because they are virtually unity gain at the minimum, so no pads are needed even for close micing with condensers. I sold every other of my studio preamps after I finished the "Delta Project", it just beat the sh*t out of everything, some famous brands included. It needs a lot of work, though. The second op-amp (line in/pre's second stage) oscillates even in the stock version, which is modulated into the audible noise. I was shocked when I found this. As I know it was designed with the 5532 in mind for this position, and it behaves there, I checked, the stock TL072 doesn't.

And I finished a "Cohen" InAmp with the SSM2212 (SOIC replacement for MAT-02) x 2 front-end without caps and servos at all, only phantom blockers. It's close to the Samuel Groner's Monte Generoso design, but there are not so many ways to achieve this goal anyway. It's powered by semidiscrete regs based on Jung's design. Heavy decoupling for every op-amp. One piece of overengineering at my kidney's price. Just sensing the output with maximum gain put a smile on my scope's screen.

So what I hope to achieve with transformers is to outperform that preamp in the "low noise with high gain" game, CMRR and RFI protection included in the equation. But may be I am delusional about their capabilities, I don't know abut the transformers that much.

abbey road d enfer said:
The LSK170 equivalent is current.
Looks even better, than the original, if I remember its datasheet correctly.
 
dbelousov said:
The second op-amp (line in/pre's second stage) oscillates even in the stock version, which is modulated into the audible noise. I was shocked when I found this.
I don't remember that. I was Soundcraft's distributor at the time. I had a team whose only work was to open each box and check operation, every pot and switch, prior to shipping to customer. If there had been such a widespread issue, I would have been informed.
You must be one of the unlucky bastards that got a made on Friday unit.  ;)

And I finished a "Cohen" InAmp with the SSM2212 (SOIC replacement for MAT-02) x 2 front-end without caps and servos at all, only phantom blockers. It's close to the Samuel Groner's Monte Generoso design, but there are not so many ways to achieve this goal anyway. It's powered by semidiscrete regs based on Jung's design. Heavy decoupling for every op-amp. One piece of overengineering at my kidney's price. Just sensing the output with maximum gain put a smile on my scope's screen.
I don't doubt it.

So what I hope to achieve with transformers is to outperform that preamp in the "low noise with high gain" game, CMRR and RFI protection included in the equation.
Good luck with that!

Re: LSK170. Looks even better, than the original, if I remember its datasheet correctly.
Linear Systems know what specs are important for their customers, audio boffins. Toshiba could not concentrate on one segment of their customer base.
 
Thank you for the right direction, I know how to google the things I have to know for this task now. Really, I understand now that a "brute force" approach "let's take the biggest ratio trafo and a FET opamp" won't bring the best performance possible.

I have found this topic https://groupdiy.com/index.php?topic=17885.0 and a couple of basic articles.

P.S. The oscillation happened at a very low gain only, somewhere between 10 and 20%. I found it checking LM4562 in this position, but I wasn't surprised, it's a fast beast, I tamed it later, though. The real surprise was the "original" TL072, that oscillated in the very same gain range. It was significantly lower in amplitude and barely hearable at the maximum gain of my monitor controller, but it was there, very visible on the scope far beyond the audio range. I started to put there every op-amp what I had, just out of curiosity, and the 5532 was stable, the OP275 too, every other op-amp oscillated in the same gain range.
 
I am not sure of the value of pursuing the ultimate in low noise. Once your noise factor is around 3dB there is very little to be gained. Bottom line is a 150 ohm source has a noise level of close to -131dBu. Even a totally noiseless preamp with 60dB gain will increase this to -71dBu. So if your nominal signal level is +4dBu you cannot achieve a signal to noise ratio better than 75dB. And a regular preamp with a 3dB noise figure will achieve 72dB. If you are not happy eith a 72dB S/N I don't expect you would be happy with 75dB.

Cheers

Ian
 
Yes, I see. The source dominates the equation anyway. I'm reading this right now https://www.ti.com/lit/an/slyt470/slyt470.pdf and have found some staff on transformer-coupled inputs as well.
 
dbelousov said:
Yes, I see. The source dominates the equation anyway. I'm reading this right now https://www.ti.com/lit/an/slyt470/slyt470.pdf and have found some staff on transformer-coupled inputs as well.
That paper is an excellent introduction and covers most things you need to consider.

Cheers

Ian
 

Latest posts

Back
Top