Not quite. Once upon a time, we really did send music signals down 600 ohm transmission lines. Given that audio is only low frequency, the line has to be rather long to qualify as a transmission line, but there was an 80lb per mile (copper weight) twisted pair without any amplifiers from central London to Daventry (60 miles or so) that definitely qualified. But the real point about 600 ohms was not that audio circuits were really transmission lines (most weren't long enough), but that they were driven from a defined impedance and so long as that occured, the equaliser at the destination would give its designed response. Equalisers at that time were complex passive constructions of L, C, and R, followed by a line receive amplifier. At the BBC, we expected the output of an equaliser to be -45dBu in order to match the 45dB gain of a line receive amplifier.
Transformers need quite careful experimentation and testing to give of their best. Although driving from a low source impedance reduces low frequency problems, it easily causes high frequency problems.
I presume you mean the optional coupling cap "C7" which some builders choose to omit and replace with a jumper.
C7 is only there to soothe nervous people and assuage their fears that "DC current flowing into the transformer primary is automatically and always EVIL, even single digit microamps! EVIL I tell you!". Install C7, presto, no DC can flow into the primary. The problem she is solv-ed, Inspector Clouseau.
If you agree with this sentiment, include C7 and sleep soundly. If you don't agree, calculate how much DC current might flow into the primary under worst case conditions of AD744 max input offset voltage and also pathological input signals with terrible DC offsets. Maybe you'll conclude that you can sleep soundly without C7.
Last edited:
I agree about the coupling cap at the primary side. Its not necessary if the input offset is reduced. Thats one of the areas where I pay alot of attention. Not because of the actual DC on the primary winding causing issues, more for the sake of linearity and CMR.
All this isn't so much a problem with FET input devices. Bipolar op amps need careful attention paid to input offset for several reasons. Doing things right at the inputs from a DC perspective avoids alot of issues, mainly DC output offset and the problems added by the obligatory cap on the output.
Capacitor distortion isn't a big deal when mainly coupling AC signals that create low voltage drop across the cap without DC differential. Once you have DC offset across the cap, distortion goes up fast and the type of distortion added isn't pleasant sounding.
There's little point to using a transformer if you're putting a cap in the primary signal path. I'm not referring to interfacing different impedances here. Its moreso for retaining the low end linearity and mostly adding the specific flavor of residual distortion the transformer usually adds.
The dominant kind of HD a capacitor adds doesn't sound nearly as good in comparison. Mind you we're talking about very small.amounts of distortion with properly selected, good quality devices.
With decent high sensitivity drivers, there is more of a perceived grainy, etched type of sound that ruins low level detail. I describe this sort of sound like comparing the sound of cheap ear buds vs open back full size headphones. Maybe not quite as extreme, but similar in character. In the measurement world, this looks like odd order HD dominating instead of mainly small amounts of even order HD..
The AD797 is a good op amp, but very dated and a little on the veiled sounding side. The OPA1692 is by far a better device. It needs careful decoupling with a ground plane PCB layout to sound its best, as do many high speed op amps. As an example, the LM4562 sounds absolutely awful when not carefully decoupled and bypassed. It will squeal like a pig in the HF and make your ears bleed. Even a lowly NE5532 will sound rough when not properly decoupled.
How much offset is too much on a 40 ohm DCR winding? I'd venture to say a few mV (less than 3-4 mV) is tolerable.
All this isn't so much a problem with FET input devices. Bipolar op amps need careful attention paid to input offset for several reasons. Doing things right at the inputs from a DC perspective avoids alot of issues, mainly DC output offset and the problems added by the obligatory cap on the output.
Capacitor distortion isn't a big deal when mainly coupling AC signals that create low voltage drop across the cap without DC differential. Once you have DC offset across the cap, distortion goes up fast and the type of distortion added isn't pleasant sounding.
There's little point to using a transformer if you're putting a cap in the primary signal path. I'm not referring to interfacing different impedances here. Its moreso for retaining the low end linearity and mostly adding the specific flavor of residual distortion the transformer usually adds.
The dominant kind of HD a capacitor adds doesn't sound nearly as good in comparison. Mind you we're talking about very small.amounts of distortion with properly selected, good quality devices.
With decent high sensitivity drivers, there is more of a perceived grainy, etched type of sound that ruins low level detail. I describe this sort of sound like comparing the sound of cheap ear buds vs open back full size headphones. Maybe not quite as extreme, but similar in character. In the measurement world, this looks like odd order HD dominating instead of mainly small amounts of even order HD..
The AD797 is a good op amp, but very dated and a little on the veiled sounding side. The OPA1692 is by far a better device. It needs careful decoupling with a ground plane PCB layout to sound its best, as do many high speed op amps. As an example, the LM4562 sounds absolutely awful when not carefully decoupled and bypassed. It will squeal like a pig in the HF and make your ears bleed. Even a lowly NE5532 will sound rough when not properly decoupled.
How much offset is too much on a 40 ohm DCR winding? I'd venture to say a few mV (less than 3-4 mV) is tolerable.
AD797 has bipolar inputs, AD744 has JFET inputs. AD797 has very little flexibility to adjust compensation externally, AD744 has extremely flexible external compensation.
This thread is not making any sense.
Are there no 600:5k or 1k:5k transformers available?
Or if you don't want to bother with discrete: LME49600 buffer. Plenty available. You can buy plenty to get a discount and use them everywhere.
Yeah, if we simply need a low-distortion 600-ohm output, THAT 1646 is an excellent choice.
Absolutely Im surprised this has not come up sooner. Perfect solution or the solution.
It allows the use of very high performance opamps, then the buffer gives the current capabilities.
Aside from the classic LME49600
TI has newer units with very low noise and even wider bandwidth. Has adjustable pins to raise bias current as well.
I believe the BUF802 can drive up to 50 ohms loads, might have limited voltage.
The higher voltage BUF634 has 250ma of current and supports slew rates way way above audio opamps, maybe 3000 or more check the datasheet.
This would allow the wide bandwidth low noise op amps to be used.
Should look at OPA2828 should be well into 45 MHz bandwidth and 150 V/us
Depending on gain your distortion levels at high frequency will hold up way way way above the transformer.
Point is that bandwidth can turn into a 250ma opamp with the buffer, and most those buffers hold up well above 200 MHz.
I also think @campsquire has a fun idea doing the same with a discrete diamond buffer.
Trick to diamond is they all need to thermal track. Ironic THAT was mentioned. They have quad matched transistors on one package.
They would all thermal track. 2x PNP and 2x NPN on single chip. 70 volt transistors rated at 350 MHz maybe super beta.
Only other 2x PNP and 2x NPN is OnSemi. Is 3904/3906 probably 3 dollars the THAT shoots up around 13 dollars.
Think the TI buffer be easier and cheaper, though diamonds are fun, likely thermal track BD139/40 buffer as well.
The new TI buffers and old same thing, diamonds. New ones allow the bias to be turned up, and external flybacks you
added to old buffers and a fet input external, are now all on the chip.
All excellent suggestions. Thank you. I didn't think of the LME49600... thats an obvious one. I do have to choose a load impedance on the secondary side, mainly because the amps have a 20k input load x2 (one on each leg of the XLR inputs). There are attenuators on the inputs and they are wired with the wiper on the signal in side.
I see the HD caused by the coupling cap. Yes, symmetrical will require a bipolar or non polar cap. I understand the distortion will be minimal if the voltage drop across the cap is low.
I see the HD caused by the coupling cap. Yes, symmetrical will require a bipolar or non polar cap. I understand the distortion will be minimal if the voltage drop across the cap is low.
I think that mixing LME49600 buffered opamp with a zero impedance type drive mentioned before might be the most optimal way. I would avoid THAT 1646 in this case. It has differential outputs, which means the effective impedance of the load will be halved to 600Ohms/2=300 Ohms. This will increase distortion. Even in perfect conditions 1646's distortion is still a bit higher than of a good opamp.
"Boost Amplifier Output Swing With Simple Modification" is an old AppNote from Burr Brown that does this, only better.
https://www.ti.com/lit/an/sboa009/sboa009.pdf
😉
Depending on the needed output current my 2nd choice would also be a diamond structure with current mirror - one of the most linear buffer circuits with very low noise. I have used that without feedback as microphone buffer.
But to be honest - I doubt that that much output current is needed in an "closed", active system. You know which amps you use, there are not 100s of m of microphone cables to drive. I would try without buffer.
I know that feeling ... these top performance parts are not cheap ...
Can you write more about your system? Do you have "important" analog sources that you want to stick with a 100% analog signal path?
An analog crossover would for sure be a bottle neck in terms of S/N. You would need go to very low resistor values to keep noise down, probably would need a buffered OPA design for this part of the circuit.
Depending on your amps and their input sensitivity you could use a step down output transformer after the crossover circuits to shift signal range! Transformers are the coolest tool for gain staging, could be very helpful here.
I'm done with analog crossovers. Get a nice DSP, good D/A converters and proper amps and feed that digital. Optimise the whole gain staging so that digital artefacts and noise is below the hearing threshold. Works great!
@SomekPoland I was under the impression the THAT1646 was identical to the SSM2142 and DRV134. While its a decent device, it's not quite on the level of an op amp, i5n which case I'd prefer a discrete buffered FET op amp instead.
@IamJF I play alot of vinyl and analog tape (1/2" 2 track and even cassette). When I'm lazy, I listen to CDs and other optical disc formats.
DSP based EQ and crossover would make sense if I listened to mostly digital formats, but I wanted to keep the signal chain fully analog to retain the headroom capability and not have a digital conversion step.
DSP based EQ and crossover would make sense if I listened to mostly digital formats, but I wanted to keep the signal chain fully analog to retain the headroom capability and not have a digital conversion step.
THAT 1646 is only pin compatible with DRV134/135. It isn't a direct clone.
Not every application is/was a transmission line where signal reflections were an issue (echoes).
Indeed it wasn't. 600 ohm was used within studio centres. But once you needed a bit of flexibility and more than (say) ten inputs on a desk, 600 ohm was hopeless. The test equipment manufacturers persisted in supporting 600 ohm long after it had gone. But my point is that almost everyone who talks about driving a 600 ohm line has it wrong. Just like balanced audio; most people get that wrong, too. I was in Lines Department where 600 ohm and balanced lines needed to be properly understood if you were going to get source material for R3 from Glasgow to London with any reasonable quality.