Page 1 of 1

Parallel Line Level Inputs

Posted: Wed Apr 15, 2015 10:13 pm
by ashleysmithd

I'm working on an audio system however I've hit a bit of a problem. I'm trying to create a parallel bus on a patch panel, so operators can plug multiple line level inputs in to create a mix, and have them all DA'd by a single DA. I have attached a diagram to illustrate what I'm trying to achieve.

A colleague was telling me this would not work, that you cannot parallel multiple line level sources and connect them all to a single audio DA input. I didn't really understand this when he explained it, as typically line level operates at 100ohm and the DA we're using will take 20 k ohm, and we're only allowing 4 sources to be paralleled.

I'm obviously missing something here, if anybody can help explain why this would not work I'd appreciate it.

Re: Parallel Line Level Inputs

Posted: Wed Apr 15, 2015 10:57 pm
by Shane
There's a couple of things wrong with that idea but one of them is that several 100 ohm impedances paralleled do not add. 10 of them in parallel for example would make the impedance 10 ohms and the 20k input of the DAmight load way down when presented with that. Really, that's almost a short! Right. it's the output that gets loaded down, not the input. Duh. Ignore this. Read Jeff.

However, 100 ohms seems very low for output impedance on modern gear. You don't have much to lose but time by trying it. Generally though you want some isolation between the things you are mixing (That's part of what a mixer does), as well as some better impedance matching, and just "mult"ing them all together provides none of that.

Re: Parallel Line Level Inputs

Posted: Thu Apr 16, 2015 9:45 am
by PID_Stop

It might help to nail down exactly what the equipment is doing, and what the impedance specs really mean.

First, consider the case where you have an amplifier with 100 ohm output impedance feeding another amplifier with 20,000 ohm input impedance:


To the left is your source amplifier; without getting deep into the design theory, it's basically an ideal voltage source with a pair of 50 ohm resistors. An ideal voltage source has the ability to drive an infinitely great load; to put it another way, it can supply the given voltage at whatever current is necessary to feed the load. In the real world, there is no such thing: transistors will self-destruct if you try to exceed a certain amount of current. So practical amplifiers incorporate some amount of output resistance to protect the output device; those buildout resistors, together with the characteristic resistance of the final driving stage, becomes the amplifier's output impedance. 100 ohms is not an unusual value; I have frames of Leitch amplifiers with 66 ohm output impedance. So far, so good.

On the right hand is your destination: a distribution amplifier, say, with 20,000 ohm input impedance. Again to use an idealized model, that's equivalent to an ideal amplifier -- which just means that it places no load on the input -- in parallel with some load resistance. Nowadays, amplifiers really can approach an ideal case... and 20k is not an unreasonable value.

This model for interconnecting line-level audio has been pretty much standard practice since op amps became the norm back in the 1980s. Having a very low source impedance feeding a relatively high load impedance makes it much harder for induced electrical noise to overcome your signal, and tends to improve high frequency response because the driving stage has more ability to overcome cable capacitance. The much older notion of matching load impedance to source impedance was a throwback to early telephone company practices, where the most important issue was achieving maximum power transfer over long distances; but in a broadcast facility, the requirements are different.

So... how does this apply to your question about combining sources to feed one destination? Remember that this whole model is based on low impedance sources feeding high impedance loads... but what happens if you just tie four sources together? Think about what the first source sees electrically: it's trying to drive the 20k load, which is perfectly fine, as well as back-feeding the three other amplifiers' outputs. Those three amplifiers in parallel look like a 33 ohm load, which won't make your first amplifier happy at all. It probably won't smoke your amplifiers -- that's why they have those buildout resistors in the first place -- but the signal will probably be seriously distorted at higher levels.

What to do? The easiest solution is to build a passive combiner network that adds more buildout resistance for each source:


What does this accomplish? From each source's point of view, the other three sources plus the destination impedance look more like an 830 ohm load -- which the amplifier can drive very easily and meet its design specifications for distortion. The actual resistor values aren't terribly critical, and I'd probably choose something between 100 and 470 ohms (standard resistor values). Just make sure that they are all the same value. Personally, I like 300 ohm just because most equipment really likes to run into 600 ohm loads, and using 300 ohm combiner resistors means that you can add pretty much any number of combiner sources without worrying about overloading any of the sources. Also, it's a very common resistor value.

Is there a downside? Yes -- the more sources you try to mix, the more the level of a given source drops. For example, if you combine two feeds, each source represents only half of the combined sum... which is a 6dB loss. Combining three sources yields about 9.5dB loss, and combining four sources results in 12dB loss. An active mixer will have an amplifier stage to bring the mix level back up, but in your application you can probably adjust your distribution amplifier's gain to make up for the combiner loss.

The end result is that by adding a passive resistor combiner, you keep your sources operating within their design limits, and by adjusting the gain on your distribution amplifier, you maintain appropriate levels without raising the noise level unduly.

Hope this helps give you a better picture of what the issues are, and how to work within them!



Re: Parallel Line Level Inputs

Posted: Thu Apr 16, 2015 10:00 am
by PID_Stop
Oh, one other thing: you want to make sure that every input to the combiner is either fed by some source, or has an equivalent dummy resistance across it. Otherwise, your mix level will vary when people plug more or fewer sources into the combiner. Assuming that you are using a standard self-normalling patch bay, and that your feeds to the combiner come from four 'bottom' jacks, just hook a 100 ohm resistor between the tip and ring on the corresponding 'upper' jacks.

I have the tee shirt for this... we used to do a lot of passive combining back when we had to combine stereo sources to feed our mono plant. :wink:

-- Jeff

Re: Parallel Line Level Inputs

Posted: Thu Apr 16, 2015 12:43 pm
by Deep Thought
PID_Stop wrote:I have the tee shirt for this... we used to do a lot of passive combining back when we had to combine stereo sources to feed our mono plant. :wink:'s pretty much how every analog line-level studio console worked before digital started to take over. It was common to have every line-level input on a console come in through a resistive H or L pad to an isolation transformer which did a balanced to unbalanced transition, then to the pot, which connected everything to common the mix bus. Then a rather hefty amplifier brought that back up to line level for output.

Jeff's example should work fine, but since the DA's input isn't a virtual ground you might get some crosstalk back down the input lines. That might not be an issue in your application if the sources only feed this mix.

Re: Parallel Line Level Inputs

Posted: Fri Apr 17, 2015 7:24 am
The drop in level shouldn't be a problem when the commercial break runs! :lol:
Somewhere at home I have an old electronics book with the formula for a resistive build out network, that provides at least 30db isolation input-input. I had to make a few of those back in my phone-line distributed Muzak days. I'll see if I can dig it up this weekend.

Re: Parallel Line Level Inputs

Posted: Fri Apr 17, 2015 9:47 am
by PID_Stop
Deep Thought wrote:Jeff's example should work fine, but since the DA's input isn't a virtual ground you might get some crosstalk back down the input lines. That might not be an issue in your application if the sources only feed this mix.
This is true. A lot depends on the design of the source's output stage, but I calculate a worst case of around -40dB crosstalk, given 300 ohm combiner resistors and a commonly driven output. If you have a D/A that has separate drivers for each output (the Leitch ADA-880, for instance), crosstalk is pretty much a non-issue. You can also reduce crosstalk by increasing the combiner resistor value: going up to 1k brings the worst case crosstalk down to about -52dB; you can increase the resistance still more and lower the crosstalk accordingly, but bear in mind that the 20k load becomes more significant as your combiner resistors get bigger.

-- Jeff

Re: Parallel Line Level Inputs

Posted: Mon Apr 20, 2015 2:10 am
by BigRed
Ah! Analog audio! You're making me get into the "way, way-back" machine here . . .

First, you might want to give these guys a look? Check out their "Stick-On" product line. They've got some pretty inexpensive items that might do just want you need. [Passive networks start at around $60 US and include some RFI protection, and most of the terminal blocks and soldering, otherwise known as "the grunt work" . . . ;>)] If they don't have exactly what you need they may be able to suggest how to "gang together" a couple of their products to get to where you want to be. And they're not the only company that has this kind of stuff available, Remember, Google can be your friend here; just a suggestion . . . )

Another suggestion . . . why not consider using an active mixer to do the combining instead of a passive "network"? That will give you some "gain-in-hand" to tweak any levels that may be off and metering to see what your levels are to begin with? Just a thought . . . (I know, it's management's myopic view of the budget . . . don't want to spend the extra money . . . ) If the budget is really a consideration, and assuming that the DA has an active, differential input rather than a transformer input, the other cost extreme would be to put a 5.1-kohm resistor between each jackfield "tip" and the DA "+" input. Do the same for the "rings" of the jackfield to the DA's "-" input. That should give you a summing network that has somewhere around 3-dB of loss but provides good isolation between the inputs to be summed (and, hopefully, the DA has sufficient gain "trim" to make up that loss.) Just keep the resistors as close to the amplifier input as possible and make sure the cabling is "twisted pair, assuming balanced audio. (I've actually done that in the distant past with DA's from either McCurdy, JVE or SWA, don't remember which offhand; in fact, I think that the SWA amplifiers had the option to place the input resistor external to the PC board so that multiple input resistors could be tied together externally to form a summing network (in an adjacent card slot in the frame?) without affecting the gain. Anyhow, I can't find my notes from that far back so I would have to re-run the numbers or "breadboard" a test setup to verify so I'm not sure on the insertion loss that you'll see. For additional reference on summing networks you might want to look at any of Walt Jung's books on IC applications. ("Summing" is a quite common application for IC amplifiers.) I was actually going to suggest his Audio IC Op Amp Applications, 2nd Edition but it's going for over $675 US, used, on right now! (Time to dig out my copy and try to sell it.)

For the remainder of this discussion I'm going to make the assumption that you have an analog audio plant that employs "balanced voltage distribution", meaning that each audio source is low impedance (68-ohms to 75-ohms is just lovely to drive a 300-meter + run of typical twisted-pair cable with no degradation of the frequency response) and high termination impedances (10 X the source impedance, or greater is a good rule-of-thumb). Also, I'm hoping that a separate amplifier or DA output is used for each destination to prevent the shorting of one destination from taking out ALL the destinations for a given signal and also to prevent sources from "backfeeding" each other, the usual DA applications. And last, but not least, that you have a plant "standard" level for all equipment. [+4 dBu referenced to +4 dBm/600-ohms for "reference 0" became pretty common for solid-state plants after we figured out that solid-state gear just couldn't do the old "valve" standard of +8dBm/600-ohms and have any head-room left with the "standard" +/- 15-volt power supply rails. (The "voltage" distribution system for audio was researched extensively by Richard Hess, et. al. at ABC-TV, New York back in the late '70's and at NBC-TV and Radio, Washington, DC in the early '80's. Probably at other places too during that same timeframe. Mr Hess presented his findings to the AES in 1980. If you're interested I can try to find the article for you.)] But I digress . . .

If you do decide to go with a passive network of some variety instead of a mixer I suggest that you stick with the high Z stuff (10-kohm) and stay as far from the 600-ohm combiners as possible. The ONLY reason to do Z matching and use a 600-ohm network is if you have "iron" to deal with. In THAT case impedance matching and minimum-loss pads are critical to dampen transformer ringing and overshoot. Power matching/proper transformer termination was also essential in the good old days to insure that the tubes in tube-type amplifiers with transformer outputs maintained the correct plate loading. (And if you are using transformers make certain that your peak levels stay below the core saturation point of the transformers, which can be pretty low. As an example it is typically about +2-dB for the formerly ever-popular UTC A20, if I remember correctly from past measurements.)

There are several reasons I suggest the hi Z summing network instead of the low Z "combining" network for your application. One is level changes. The second is that impedance and isolation values change as sources are plugged into or removed from the network. Consider the 100-ohm source impedance noted elsewhere. If you terminate a 100-ohm source into a 600-ohm load when it is set up to work into a high impedance (like the 20-kohms of your DA) you will see a drop in level of around 1.5-dB. Not a HUGE drop, but a drop none-the-less. As you add sources they will add additional loading to the resistance network causing additional degradation in levels. In addition, unless the low-Z (600-ohm) network has all its ports properly terminated, the isolation numbers, as well as the power combining/division numbers (and it is a power combining/division network), go right out the window. The high impedance summing network provides greater isolation between inputs and is virtually unaffected by adding or removing sources, either low Z or high Z.