Output terminology 1:1 or 1:2 question

I’m in alignment, generally, with Matt. At same time, I definitely don’t want to discourage any conversation on DIY circuits or disparage anyone on their way to an improved understanding (especially on RGO).

One way to build a “gut feeling” for noise tradeoffs, including impact of the transformer, is to simulate with LTspice. You can do .noise simulations and model your mic’s impedance. To get deeper (and this is the only “professional” way to approach design) you can compare to hand calculations that are only as complicated as necessary to describe what’s seen in SPICE results. It’s also good to question the results because SPICE is as valid as what’s put in: for me, most issues have been “user error “ on my part. So start with a simple resistor and see if the noise floor matches the theory: a 1kOhm resistor should exhibit a noise density of 4 “nV per root Hz”.

Solid books/resources on the topic are:

Motchenbacher and Fitchen (or the similar book by Motchenbacher and Connelly)

Art Kay's book Operational Amplifier Noise, Techniques and Tips for Analyzing and Reducing Noise

Bill Whitlock’s chapter:
www.jensen-transformers.com/wp-content/uploads/2014/08/Audio-Transformers-Chapter.pdf

See 8-5 of this ADI app note which basically explains choosing optimum turns ratio for best SNR once the input referred voltage noise and input referred current noise generators (modeling opamp’s self noise) and source resistance Rs are known. Of course, with discrete transistors the designer can optimize first the discrete opamp’s noise and then follow up with best chosen transformer:
www.analog.com/media/en/training-seminars/design-handbooks/system-applications-guide/Section8.pdf

The design info on the 990 and twin servo preamp also helped me build a better understanding of these topics:
www.johnhardyco.com/pdf/990.pdf

www.johnhardyco.com/pdf/TSMP2015-03-01.pdf

www.johnhardyco.com/pdf/M1M2M1p.pdf

and Deane Jensen’s original AES paper:
www.technicalaudio.com/pdf/Jensen_Transformers/Jensen_OpAmps_990_and_related/Jensen_JE-990_opamp_JAES_reprint_1980.pdf

I find you being clueless on how all of this suppose to work.

Yes, it would make sense that Jensen Transformers is misinformed about how to best select transformers for an application using their transformers (kidding).

In case we are having some communication error, to clarify, the quote from Jensen above is regarding the use of a transformer on the input side, to interface between the (typically low-impedance, i.e. 150 ohms) mic and the preamplifier circuitry.

When "looking back" into the output of the input-transformer we want to "see" the most optimum resistance possible to keep noise floor to a minimum (best case being the noise floor of the ~150Ohm thermal noise generator of the mic if preamp were ideal, we can only make it worse in practice). The goal of the transformer selection method above is to minimize the impact of the added noise from practical preamp circuitry since the SNR can't get any better that what the mic exhibits.

I think the whole conversation got turned upside down because a transformer (apparently) designed for duties on the input side was being considered for duties on the output side --> it's of course possible to do this, but it wasn't optimized for that role (larger signal levels more likely to result in distortion, etc.).

Transformer input or output rating has nothing to do with the actual impedance it presents to the adjacent device in the chain.

Agreed. If transformer is ideal (ignoring core loss, saturation, etc.), I think this is referred to as “load reflection rule”, which goes in both directions since transformer itself is completely passive and has no knowledge of input vs. output (designation of primary vs. secondary can be reversed if we simply apply our “input” from one side or the other):
hyperphysics.phy-astr.gsu.edu/hbase/magnetic/refload.html



(This was also covered in some of the resources I listed in a post above, i.e. Bill Whitlock’s book chapter)

Transformers are designed to go between certain source and load impedances. An input transformer is expecting to be fed from a certain source impedance and be loaded by a certain load. The turns ratio is only a part of the transformer design. Total inductance and the expected load play a role as well.

For starters, if you look at the JT-11P datasheet it shows the typical source impedance is 14k. An op amp has near-zero output impedance. And, in a 14k source impedance to 10k output, it has a 3 dB loss. If you look at an output transformer like a JT-123 set up for 1:1 the expected input impedance is ~700 ohms, and for the rated maximum output level the input impedance Rs is zero ohms - exactly what you would expect for a transformer being fed directly from an amplifier output. The 11P is also designed for very high common mode rejection, because it is designed for a different purpose than output transformer. Even worse, maybe the real problem, is that if you use an 11P as an output transformer you're presenting a source impedance to the next device of 3.2k ohms. If you use a 123 in 1:1 your source impedance is only 112 ohms. What that translates to is a loss of signal level as you go from less bridging to more matching configuration, as well as potential high end loss due to cable capacitance. And, if we're talking about noise - a higher impedance circuit will have more noise in and of itself.

All that adds up to - wrong tool for the job, and you'll end up paying twice or three times as much for worse performance.

Jensen knows what they're doing. There is a reason they have different transformers for output and input duties.

If you want to pick an output transformer you need to consider

- the current drive capacity and maximum voltage output level of your driving amplifier
- the source impedance of your driving amplifier
- the range of loads expected for the transformer

From there, you have a box to pick from.

For example if you're designing a 312 style preamp with a beefy 2520 output, you have a maximum of 60 mA current capacity into 75 ohms and 20 dBu output voltage from your op amp. Your transformer source impedance is near zero, and you should expect a load range anywhere from 600 ohms to >10k ohms. We have gobs of current drive so we can consider 1:2 or even 1:3 output. That changes our load range minimum from 600 ohms to potentially 150 or even 67 ohms. (Aside - when you add the DC resistance of the transformer windings, you'll get a minimum load from a 1:3 step up driving a 600 ohm load to around 75 ohms - not coincidence that API puts the 75 ohm load capacity on the 2520 datasheet). Since we can handle the current required for step up, going 1:2 gets us a 'free' 6 dB of gain, increasing our maximum output voltage to 26 dBu. A quadrifilar output transformer gives the designer the freedom to select ratios from the same windings (1:1, 1:2, 1//1:2, 2:2, 1:3), which gives a range of inductance and load to match different driver and load scenarios.

Transformer windings have their own imperfections, and as you load them there will be losses and frequency response distortions. A transformer rated for 600 ohms will usually have a bit more high end when loaded with 10k and a bit higher than nominal voltage gain. This is why you'll often see a load resistor selected, to narrow the effective operating range to minimize the potential difference in performance when you plug a different device in on the other side of the chain.

A transformer's DC parameters are very nearly irrelevant in audio. Yes, their DCR is added to the impedances, but in most cases they're negligible compared to the impedance of the circuit. Transformers are fundamentally AC devices. They exist to transform AC signals.

You're not wrong to examine the distortion curve at rated output into different loads. Most IC op amps cannot happily drive a 600 ohm load, though 2520 style DOA will do it all day (API 2520 datasheet says 0.2% THD 20-20kHz at rated output, meaning 20 dBu into 75 ohms!). However, the transformer does not present the load to the op amp! It only reflects the load, through its turns ratio plus its own DCR / impedance, back to the driver.

It may be helpful here to return to first principles - what is a transformer. An ideal transformer has no impedance of its own. There is no leakage between the two coils - so no parallel capacitance etc. No core loss, and infinite permeability, and perfect coupling between the primary and secondary coils. If you hook an ideal transformer up to a circuit, the driver will see ONLY the load reflected by the square of the turns ratio. That is to say, a 1:1 transformer will present ONLY the load to the driver, and ONLY the source impedance to the next device. A 1:2 will present the load impedance divided by four, and the source impedance times four, and so on.

Transformers are always AC coupling devices, as they by definition do not pass DC. So all transformers are "coupling" transformers. And, likewise, all transformers are at least potentially impedance changing devices.

You are correct that transformers can also balance a line - but you don't need them to get common mode rejection. You can simply use a single-ended circuit with identical impedances on both lines for the same purpose. Transformers just happen to accomplish this along with presenting an inverted version of the signal on the low side. What does the "rejecting" is the receiving device, not the sending - by summing the two signals after inverting one. Op amps can do this, and so can transformers.

I don't really know what you're trying to describe in your coupling transformer paragraph, but I must admit it seems more than a little confused. In the end, the entire impedance of the load - whether from a resistor termination or series capacitors or parallel capacitors or inductance (ie from a cable) - can be rewritten as a single terminating impedance across the transformer windings, and if necessary a second impedance to ground. Transformers are designed for a certain terminating impedance; this is the "rating". So a transformer designed for 600 ohm loading is designed for a 600 ohm impedance within the specified bandwidth, regardless of what makes up that impedance (capacitance, resistance, and inductance can all factor in).

Saying a transformer can be used for input and output is like saying you can use any piece of glass for a door or car windshield or a window. It's technically true, but results may vary. Best to use the right tool for the job.

For optimum noise you need to present the correct source impedance to a circuit. This, along with clean gain, is the purpose of a mic input transformer - it increases the source impedance from the microphone to present the correct source impedance to the amplifier, which minimizes noise.

For optimum voltage transfer you use bridging impedance, which is used in ALL modern audio devices. Therefore you ALWAYS want to keep your output impedance as low as possible, and at a minimum 1/10 the amount of your input impedance.

XLR cables can have capacitances as high as 120 pF per foot. Even a modest 12 ft run with a 3k ohm source impedance would result in a -3 dB point of 36 kHz. Audible? Probably not. Degrading the bandwidth of the system? Absolutely. Never mind that a 3k source impedance into a 10k load also carries a -2.25 dB level penalty.

"Op amps are current out devices" - I fundamentally disagree with this based on EE convention. An "ideal opamp" (operational amplifier) is typically designed and configured (via feedback connections) to have high (ideally infinite) input resistance (drawing zero current ideally) and low (ideally zero) output resistance --> thus it can ideally drive any load current necessary while FORCING the output voltage at its output. It's a voltage output device because it's forcing a voltage at the point where the feedback connection is applied (sometimes called the "Kelvin connection", but this term has another similar meaning in the field). So it's a voltage input, voltage output device.

Even before feedback is applied, with the ideal opamp, the output voltage is always expressed as a voltage equal to the product of the differential input voltage (Vin+ - Vin-) and an ideally large scale factor (the "open-loop gain"). When designing actual opamps we try to mimic this behavior with exceedingly high open-loop gain (at least at DC) and by including an output stage often with inherantly low output resistance before we even apply feedback (with "rail-to-rail" opamps often being an exception). So, Opamp = voltage output device.

Also in EE convention we have something called an ideal OTA (operational transconductance amplifier) which has a similar high input resistance but simultaneously has HIGH output resistance and it forces a current at it's output, so in this case the voltage at the output can take on whatever value but it's the current that is being forced by the conceptual device. This is a voltage input, current output device.

An opamp with appropriate choice/connection of feedback elements can be configured to mimic an OTA in practice - see the Howland current pump or opamp "V-to-I converter" (voltage-to-current converter) using a FET or BJT.

We can also configure an opamp to mimic a "current amplifier" (current input, current output device) or even a "transimpedance amplifier" (current input, voltage output device). This has to do with the nature of feedback connections having the ability to modify both input and output impedances.

In practice, with finite input resistance and non-zero output resistance none of these conceptual devices exist perfectly, but of the four I think the opamp acting as a voltage in, voltage out device might be the easiest / most natural concept to apply based on behavior of most commercial IC opamps in the non-inverting feedback configuration.

Johan Huijsing's book on Operational Amplifiers is probably one of the best resources to understand these details, especially his chapter on Macro models.

I think to address this we need to move to Maxwell's equations, asap!