Balanced to unbalanced converter

[levimax]

For low noise and high CMRR transformers are the gold standard. Good transformers can be low distortion and what distortion there is is at low frequecies. Very easy and reliable as well. Only down side is cost .... sometimes you get what you pay for.

 

[Thorsten Loesch]

For low noise and high CMRR transformers are the gold standard. Good transformers can be low distortion and what distortion there is is at low frequecies. Very easy and reliable as well. Only down side is cost .... sometimes you get what you pay for.

Indeed. I like good transformers. But really GOOD transformers (that beat active circuits) are neither cheap, small or light.

I'm going to post "what do discrete circuits offer?" next. Wait for it.

Thor

 

[horias2000]

All the single chip balanced receiver chips have relatively high noise, THAT included.

The OP wants less noise than that.

Using a suitable Op-Amp (e.g. OPA1611) and 4 resistors of the right value can reduce noise by 10dB over the IC balanced receivers and approach the complex circuit given by D. Self within 2-3dB, at the cost of 2.4kOhm input impedance which is rarely an issue with well designed gear.

The problem with Mr. Self is that he suffers from an acute phobia against using any op-amps that are not NE5532 or TL072, so he would rather use large numbers of NE5532 costing more than an actually objectively better and once the full circuit is implemented more expensive than an appropriate part.

Mind you, he shows good Design techniques to achieve really low noise if needed.

Thor

Well D. Self looks at the LM4562 and at the AD797 and lists these in his measurements. Indeed the OPA1611 or OPA1612 op-amps are great but the price is considerably higher than the LM4562 and the improvements are not that big in my opinion.

 

[horias2000]

OpAmp = OPA1611/OPA1612, LT1028/1115, AD797
R1 = 1.2k
R2 = 600R (2 x 1.2k in parallel)

Resistors should be 0.01% types for best CMRR.

Alternatively to improve CMRR, multiple inexpensive resistors may be paralleled. In full reels Yageo 0.05% 10k SMD thinfilm resistors cost 8 cent per resistor.

If we use 8 x 10k for R1 and 16 x 10k for R2 we get in effect 0.01% tolerance in circuit (in practice better) for 2.56 USD in resistors and 2.60 USD in the Op-Amp (prices from mouser, full reels).

The following assumes the use of 32pcs 0.05% resistors 10k and OPA1611.

In practice just using 0.05% tolerance resistors 1.2k x 6 will deliver excellent CMRR anyway and reduce BOM cost to $ 3.08 materially outperforming "canned" options in all metrics including cost, except worst case (garantueed) CMRR.

BOM Cost: $ 5.16
Gain -6dB
Input Impedance 2.4kOhm
CMRR > 86dB

Resistor noise ~ 0.5 uV
Op-Amp noise ~ 0.2 uV
Noise gain ~ 1.5 (3.5dB)
Total noise ~ 0.81 uV

Ein ~ -116dBV
SNR @ 0dBu out ~ 119dB
SNR @ 0dBu in ~ 113dB

THD & N @ +18dBu in (~6V in) < 0.00005%
SINAD @ +18dBu in (~6V in) ~ 125dB

QED

With similar BOM Cost as a "canned" receiver chip we can outperform it by a large margin.

An NE5534 with an added external discrete class AB buffer driven from from pin 5 could be used if a modest reduction in SNR is acceptable.

A 2N4401 pair may be used as external input stage on the above NE5534 to provide lowest noise, add 2mA CCS.

Overall this, just like Mr. Self's proposal adds a lot of parts. And we might as well ditch the 5534 and go fully discrete.

Thus paying the 2.60 USD for a superior performance Op-Amps instead of paying 1.1 USD for six 5532 and adding a lot of extra parts is often the better choice.

As is paying 3.08 USD for the OPA1611 & 6 x 0.05% tolerance resistor compared to paying the same or more for "canned" receivers.

Thor

I totally agree with this approach. The extra cost of the precision resistors is trivial and the performance is there. I will most probably use LM4562 and 0,05% resistors. I will also try the 1611/1612 op-amp as soon as I can get my hands on a few pieces.

 

[Thorsten Loesch]

I totally agree with this approach. The extra cost of the precision resistors is trivial and the performance is there. I will most probably use LM4562 and 0,05% resistors. I will also try the 1611/1612 op-amp as soon as I can get my hands on a few pieces.

Mouser has stock and ship worldwide.

Thor

 

[majingotan]

Indeed. I like good transformers. But really GOOD transformers (that beat active circuits) are neither cheap, small or light.

I'm going to post "what do discrete circuits offer?" next. Wait for it.

Thor

What do you think of Cinemag Input Transformer? https://www.don-audio.com/CineMag-CMLI-15-15B-line-input-transformer

 

[Thorsten Loesch]

Well D. Self looks at the LM4562 and at the AD797 and lists these in his measurements. Indeed the OPA1611 or OPA1612 op-amps are great but the price is considerably higher than the LM4562 and the improvements are not that big in my opinion.

Mr. Self has a a bit of mission that makes him less than amenable to what he considers audiophile nonsense and in the process also kills a lot of useful stuff that he just dislikes.

If you do things DIY and you account for your time, you suddenly realise that that 5 USD Op-Amp is A LOT CHEAPER than six 1 USD Op-Amp's as you only do 4 solder joins, not 24. Similar for resistors.

Thor

 

[Thorsten Loesch]

What do you think of Cinemag Input Transformer? https://www.don-audio.com/CineMag-CMLI-15-15B-line-input-transformer

In my experience it is too small to have low distortion at low frequencies. So it will have the classic so-called "Iron Sound" beloved by many studio folk.

The kind of transformers that keep LF distortion under control weigh around a kilogram and have around ten times the material cost of Cinemag.

If you want small and cheap (yes, 100 USD is cheap for a quality audio transformer) go for a zero-field setu, but that is an passive-aggressive, ooops, passive-active setup and needs some design expertise in such circuits.

Thor

 

[horias2000]

Mr. Self has a a bit of mission that makes him less than amenable to what he considers audiophile nonsense and in the process also kills a lot of useful stuff that he just dislikes.

If you do things DIY and you account for your time, you suddenly realise that that 5 USD Op-Amp is A LOT CHEAPER than six 1 USD Op-Amp's as you only do 4 solder joins, not 24. Similar for resistors.

Thor

Yes. I use Mouser a lot and they have stock. And I also agree that for Diy it makes a lot more sense to use higher quality / performance parts. That is my philosophy as well. I think that D. Self has a more "series producton" approach to his designs and not diy oriented.

 

[Thorsten Loesch]

Yes. I use Mouser a lot and they have stock. And I also agree that for Diy it makes a lot more sense to use higher quality / performance parts. That is my philosophy as well. I think that D. Self has a more "series producton" approach to his designs and not diy oriented.

Yes, DS is production oriented, as am I incidentally.

For DIY within reason, anything goes.

I'd probably have a fully discrete SMD design assembled by a PCB/PCBA vendor from Asia.

Thor

 

[Hayabusa]

What about adding a small trimming pot to trim the cmr to perfection?

I totally agree with this approach. The extra cost of the precision resistors is trivial and the performance is there. I will most probably use LM4562 and 0,05% resistors. I will also try the 1611/1612 op-amp as soon as I can get my hands on a few pieces.

 

[Thorsten Loesch]

What about adding a small trimming pot to trim the cmr to perfection?

Yeah, well, I always think "production engineering.

Being able to make something meet spec without manual trimming or selection of components is worth a lot. More reliable and less expensive overall.

So if using 32 cheap tiny SMD resistors makes the design reliably meet spec without trimming, hell yeah.

Imagine an 8 In + 8 Out Audio interface, trim each input and output for CMRR of > 80dB by hand? Consider the skilled labour cost and the cost of test gear etc. Why not make sure you get this by design for PCB after PCB without any need to even check (guaranteed by design spec)?

Thor

 

[horias2000]

Ok. As a short conclusion I will be building the PCB for the circuit presented in the original post (the one with 4x dual op-amps) as this PCB will give me the flexibility to try several configurations. I can easily short the input buffers and use the basic one op-amp differential amplifier. I can also use the input buffers if I decide I need higher input impedance and I can also parallel the stages for lower noise. I will modify the schematic so that I can add several resistors in parallel for better accuracy and hence CMRR. Adding a multi turn trimmer might also be a good idea. I can later assess the noise performance and the CMRR.

Thank you all for the valuable input. I'll keep the thread updated with the measurements once I get the PCBs from China.

 

[Thorsten Loesch]

Ok, well, the OP has decided to make the Monster anyway.

But I promised a further installment. Namely, what can discrete circuitry do for us. Here my candidate for a Balanced Input:

Yes, I notice that transistor types are unmarked, as are all resistor and capacitor values. Anyone (not a chinese engineer) who can successfully implement the circuit can work these out anyway and I am not in the habit to give free design work to China.

This circuit uses only "jelly bean" transistors, nothing is "special", everything is generic with multiple vendors and large stocks available. That offers a backstop for both obsolescence and availability issues as we now face frequently. Cost for all semi-conductors for two channels and power supply is USD 1.25 (buying full reels).

The 0.05% resistors and electrolytic capacitors are the other main cost drivers, at 1 USD for resistors and 2 USD for electrolytic capacitors.

So in volume BOM cost for 2 Channels is around 4.25 USD.

Note that at least some of these costs accrue for both Op-Amp + resistors and even with "canned" receivers.

Now what are the expected performance parameters? They are all based on the method for the test results in the THAT1246 Datasheet. They should be directly comparable:

Input Impedance 2.4kOhm

GAIN -6dB

SNR ~ 119dB

THD&N ~ 0.0003% (+10dBV [3.16V] input, RL 2k)
~ 0.005% (+26dBV [20V] input RL 2k)

CMRR ~ 70dB @ 60Hz guaranteed, > 80dB typical

As we can see, the discrete solution compares highly favourable in all audio metrics to the IC solution.

The downside is the need for +/-24V (~ +/-30V unregulated before the PSU) Power (it may be seen as benefit by some) at some 50mA of current. This is substantially more power than used by the IC's.

Still, the otherwise much lower cost, lack of supply chain risk and improved performance in my view makes this discrete alternative at least worth serious consideration.

Thor

 

[KSTR]

@Thorsten Loesch, derogatory blanket statements are not welcomed here, please watch your language.

As for your circuit proposal (a standard subtractor with a standard OpAmp core in the end), you're comparing to the wrong IC here. 124x is deprecated, nobody uses them. 120x is the series to compare to.
The input impedance is very low (not ideal to begin with) and CMRR goes out of the window with any small amount of source impedance unbalance, much quicker than with a higher input impedance for that standard subtractor circuit. Both of which could be issues for a truly universal balanced input.

 

[KSTR]

@horias2000

If your front-end can tolerate some gain (>= 2x) and a large common-mode input range is not required then you might want to look into this circuit by Scott Wurcer, a giant in the world of integrated circuits design: https://www.diyaudio.com/community/...oners-ln-measurement-amp.253174/#post-3864395
JFE2140 from TI would be choice for the dual JFET and would yield very low noise together and have high input impedance. The circuit can be tweaked (cascodes for lower distortion) and has a few things to consider to squeeze out the maximum performance from the JFETs at low gains (CM operational point etc, feedback resistor noise, opamp drive capabilites, etc) but the basic idea is brilliant.

 

[Thorsten Loesch]

@Thorsten Loesch, derogatory blanket statements are not welcomed here, please watch your language.

??? You mean my note of Mr. Self's bias is derogatory instead of observational?

As for your circuit proposal (a standard subtractor with a standard OpAmp core in the end), you're comparing to the wrong IC here. 124x is deprecated, nobody uses them. 120x is the series to compare to.

https://www.mouser.com/ProductDetail/THAT/1246P08-U?qs=9Udfh7QmL4tv6UPu0jt%2Bdg==

It is an active product, stocked by Mouser and actually quite widely used, because it has pin compatible, functional and performance near equivalent parts from a number of vendors.

On the other hand I do not see a widespread use in products of the 12XX.

Mind you, we could easily make a discrete "super 12XX", the relevant patent is expired since 2013. The extra cost would be quite nominal, say another 50 USD cent maximum.

The input impedance is very low (not ideal to begin with) and CMRR goes out of the window with any small amount of source impedance unbalance, much quicker than with a higher input impedance for that standard subtractor circuit.

Everything is a tradeoff. More CMRR with unbalanced source means more noise, or extended circuit complexity.

It would be trivial to increase the relevant resistor values if we feel the need to emphasise CMRR with any unbalanced source

Both of which could be issues for a truly universal balanced input.

Yes, one needs to precisely define requirements. I do not think there is really any kind of "universal" circuit. Every circuit is a tradeoff and something "universal" is usually suboptimal in some metrics in any real world application,

Thor

 

[KSTR]

??? You mean my note of Mr. Self's bias is derogatory instead of observational?

No. It is this statement:

Anyone (not a chinese engineer) who can successfully implement the circuit can work these out anyway and I am not in the habit to give free design work to China.

Overall, I would agree to "everything is a [n engineering] compromise" as well as an economic tradeoff.

 

[Thorsten Loesch]

No. It is this statement:

"Anyone (not a chinese engineer) who can successfully implement the circuit can work these out anyway and I am not in the habit to give free design work to China."

That is a statement of fact.

Thor

 

[KSTR]

So you are saying that all Chinese engineers are generally incapable? Not true.
Also, that publishing full circuits will invite only the Chinese to copy/build them? Not true either. Stuff put in the public domain can (and will) be used by anyone.
Those are clearly derogatory blanket statements in my book and should not be used, no matter if you got bitten in some specific incidents or not. (I've got bitten too, for that matter, mind you).