Hi,
I'm building a couple attenuators, one for unbalanced lines and another for balanced lines.
I have my unbalanced design worked out using a PGA2310, but haven't seen much discussion of designs/chips for balanced (differential) lines. Except this brief thread.
Does anybody have any experience or recommendations for chips and designs for balanced lines?
If there isn't a balanced equivalent of the PGA2310, I'd consider making a "balanced bridged T" with 10-bit 1%-tolerance digital potentiometers (something like these) in the requisite locations.
Thoughts?
Thanks!
John
I'm building a couple attenuators, one for unbalanced lines and another for balanced lines.
I have my unbalanced design worked out using a PGA2310, but haven't seen much discussion of designs/chips for balanced (differential) lines. Except this brief thread.
Does anybody have any experience or recommendations for chips and designs for balanced lines?
If there isn't a balanced equivalent of the PGA2310, I'd consider making a "balanced bridged T" with 10-bit 1%-tolerance digital potentiometers (something like these) in the requisite locations.
Thoughts?
Thanks!
John
This design DIY, audio, electronics... uses a balanced line receiver INA2134 and driver DRV134 in conjunction with a PGA2310.
Well documented but oy vey I was hoping to keep the design simpler and not put so many chips in the signal path.
Well documented but oy vey I was hoping to keep the design simpler and not put so many chips in the signal path.
The right thing to do is almost certainly to unbalance the line, attenuate then do the balanced line driver thing, that way the critical components for interference rejection are restricted to the line recever.
Remember 1% parts in a balanced circuit potentially equal only 40db CMRR, and I would be surprised if the digipots were that good by the time you dial them down from full scale by 20dB or so.
I would advocate the THAT corp line recevers as being good parts, some with really nice real world CMRR that you will not get by staying balanced all the way (Almost always a mistake IMHO).
Regards, Dan (Who does not at all mind more chips in his signal path as long as they serve a purpose).
Remember 1% parts in a balanced circuit potentially equal only 40db CMRR, and I would be surprised if the digipots were that good by the time you dial them down from full scale by 20dB or so.
I would advocate the THAT corp line recevers as being good parts, some with really nice real world CMRR that you will not get by staying balanced all the way (Almost always a mistake IMHO).
Regards, Dan (Who does not at all mind more chips in his signal path as long as they serve a purpose).
Do a search for "Muses 72320". That is what Mr. Pass uses in some of his high-end gear and there are semi-DIY offers for both balanced and SE around.
Not shy of saying that IMO keeping the signal balanced all the way and having as few as possible chips in the path is obviously always best😉
Not shy of saying that IMO keeping the signal balanced all the way and having as few as possible chips in the path is obviously always best😉
I would :
1, Convert to SE, do the attenuation and then back to balanced.
OR
2, Build an R2R relay Attenuator, use 0.1% resistors and you get almost perfect balanced signal.
1, Convert to SE, do the attenuation and then back to balanced.
OR
2, Build an R2R relay Attenuator, use 0.1% resistors and you get almost perfect balanced signal.
Why not make a shunt, two fixed resistors and then shunt the signal between rails, that way the signal remains truly balanced.
the muse datasheet has a matching plot - looks like ~0.5% so not much better than 40 dB CMRR if you keep the signal balanced
sometimes you have to choose between real performance and "conceptual art" - where in this case "fully balanced signal path" is on the conceptual art side of the line
sometimes you have to choose between real performance and "conceptual art" - where in this case "fully balanced signal path" is on the conceptual art side of the line
Why not use two PGAs, one in each line?
Seems the simplest sulotuion, the control lines would just be connected together.
jan
As I understand, a shunt does not present a constant input and output impedance as the attenuation is changed.
Preserving a constant input and output impedance would be a design goal for me.
40 dB CMRR isn't actually "bad" - may be way more than adequate in a home system with few meter runs, common branch mains power (which could be fine with SE IC from the start)
but if you actually need CMRR, or are chasing it as a spec - then you can do much better with the diff-to-se conversion up front
nowadays there seems little point in analog attenuation if you are fully digital source - the "lost bits" heuristic is outdated in a world of 24 bit, >120 dBa S/N audio DAC with the electronic's noise way above the digital word lsb
but if you actually need CMRR, or are chasing it as a spec - then you can do much better with the diff-to-se conversion up front
nowadays there seems little point in analog attenuation if you are fully digital source - the "lost bits" heuristic is outdated in a world of 24 bit, >120 dBa S/N audio DAC with the electronic's noise way above the digital word lsb
Last edited:
I'm inclined to agree, however…
These attenuators I'm building are for precision attenuation of theatrical cinema audio mixes in a calibrated (-20dBFS=85dBspl per SMPTE RP200) listening environment, for the purpose of loudness matching.
They need to be outboard gear, hence I'm working in the analog domain.
Last edited:
Hi all,
I have some MUSES 72320 for sale.
25 Euro each plus shipping from Germany.
Please PM me.
I have some MUSES 72320 for sale.
25 Euro each plus shipping from Germany.
Please PM me.
If you want constant input and constant output impedances, take a balanced stage before the muses and a balanced stage after the muses.
You do this also with the classical bal-se conversion design.
With a OPA1632 , it is very simply, the design will be full balanced and the gain stages and buffers in the signal chain will be less then in normal bal-se -pot -se-bal applications. To lower input impedance error I use also a 4.7k in parallel to the muses input. But I have used over hundred muses in several designs and the impedance matching on a single chip, used for balanced, is very good.
If you use the Muses direct at the input, you can get over 40dB CMRR.
You do this also with the classical bal-se conversion design.
With a OPA1632 , it is very simply, the design will be full balanced and the gain stages and buffers in the signal chain will be less then in normal bal-se -pot -se-bal applications. To lower input impedance error I use also a 4.7k in parallel to the muses input. But I have used over hundred muses in several designs and the impedance matching on a single chip, used for balanced, is very good.
If you use the Muses direct at the input, you can get over 40dB CMRR.
- Status
- Not open for further replies.