Determining input impedance of audio circuit

Advertise with us
Shop deals on ebay
shop deals on amazon
GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
P

Potato Cakes

Well-known member
Joined
Jul 1, 2014
Messages
2,269
Location
Nashville, TN
Hello, everyone,

I am using a BA283AM circuit as a summing amp for a project. Since all of the signals will already be unbalanced when they are passively summed before the amp, there will be no transformer between the buss and the make up circuit. I am trying to figure out how to determine the input impedance for the BA283AM circuit and for any other audio schematic that I come across. The input for the BA card being used as a 10uF tantalum capacitor followed by a 2k2 resistor and then it enters the circuit. Is the answer I am looking for 2k2 or do I need to calculate for the capacitor as well? This is probably a seemingly easy thing to do but I have never understood how one goes about finding this value to be able to do the related math.

Thanks!

Paul
 
Potato Cakes said:
Hello, everyone,

I am using a BA283AM circuit as a summing amp for a project. Since all of the signals will already be unbalanced when they are passively summed before the amp, there will be no transformer between the buss and the make up circuit. I am trying to figure out how to determine the input impedance for the BA283AM circuit and for any other audio schematic that I come across. The input for the BA card being used as a 10uF tantalum capacitor followed by a 2k2 resistor and then it enters the circuit. Is the answer I am looking for 2k2 or do I need to calculate for the capacitor as well? This is probably a seemingly easy thing to do but I have never understood how one goes about finding this value to be able to do the related math.
Click to expand...
The actual impedance is difficult to compute, since it encompasses three main elements: indeed, the 2.2k , but also, the input impedance of the 1st stage is about Beta times the emitter resistor, so between 300k-1Meg...but, the bias resistor (56k) introduces current negative feedback, which significantly reduces the apparent impedance. By how much is a little complicated to calculate, since the actual value depends on the Beta of the three transistors. Anyway, a simulation shows that the actual input Z is about 40kohms, but that's probably with +50/-30% accuracy.
 
That's not a great circuit for summing. That's a line driver. The 2k2 could cause problems with summing and noise. The preamp circuit might be better. You should post your exact circuit if you want better answers.

If you want to know the impedance of a circuit that is more complex like the neve discrete amps, you can measure it but driving the input through a series resistor of say 10K. Then use a scope to measure the voltage drop across the resistor. That divided by the resistance gives you current into / out of the input. Then just take the voltage at the input / current. That will give you impedance. Although if the input is high impedance, you might have to make the series resistor pretty big to get enough of a voltage drop to get an accurate current reading.

Or simulate in LTSpice but of course that is not 100% accurate. Empirical measurements are always better.
 
Thanks for help, guys!

I am actually using the AM circuit in some custom designs that have a low channel count summing buss and they've been working great with no noise or undesirable frequency response artifacts. However I am feeding unbalanced signal from active circuitry and not a passive setup from balanced connections.

I thought about a voltage drop test across a resistor as I use that for testing the power draw for circuits, but I didn't think the same would apply for audio. Also, I am really lazy so I was hoping that looking at the schematic would clearly reveal the answer. I suppose if an input transformer was being used then I would just use the resistance on the input side be it the primary or the secondary windings.

Even if the impedance is 40k +/-50% I would still be well within the 1:10 ratio from summing bus to summing amp input. So I will proceed as planned with the design I have in mind.

Thanks!

Paul
 
Last edited:
Potato Cakes said:
I am feeding unbalanced signal from active circuitry and not a passive setup from balanced connections.
Click to expand...

Not sure if you're aware, but the classic Neve desks that used the 1272  (or the later 3415) for voltage summing were also unbalanced, despite there being a 10468 transformer before the amplifier.    The primary of the transformer was shunted with a resistor to force the impedance of the mixing network low, close to the nominal microphone impedance of 150 - 200 ohms. 
It might not be the best practice as far as noise is concerned when the mix amp has some 40dB gain (which is enough for 100 channels and you're only summing, say, 24),  but there are still plenty of great sounding records recorded and mixed on those desks.

Actually, if you were mixing 24 channels and also wanted to have the usual 10dB "in hand" for the master/group fader, 40dB is only 2-3dB more than needed.
 
 
To measure the BA283AM input impedance you could fit a 100K pot in series with the input. Set the pot to zero, feed in a signal from a low source impedance generator and note the output level. The adjust the pot so the level drops by 6dB, Carefully take the pot out of the circuit and measure its resistance  which should be equal to the input impedance.

As others have pointed out, the Neve sunning buses are unbalanced. The transformer provide a handy almost noise free 6dB of gain.

One reason the Neve passive bus bus is slugged to 150 or 200 ohms (despite the noise implications) is crosstalk. In passive mixing two factors affect crosstalk; the source impedance driving a pan pot and the bus loss. Imagine channel A driving the bus via its pan pot (panned centre). The level on the bus it feeds will be 40dB below the level coming out of the pot (Neve use 15K bus feed resistors and 150 ohm bus = 40dB loss). Imagine channel B doing the same thing. The -40dB signal from channel A  returns to channel B pan pot and is attenuated by the ratio of the bus feed resistor and the output impedance of the channel B amp driving the pan pot. In Neve amps this is typically about 50 ohms so the attenuation is about 50dB. The signal from channel A pan pot therefrom appears as a signal from channel B pan pot but 40+50 = 90dB down. So the crosstalk is 90dB (a typical figure for a 1970s Neve).

If you do  not slug your bus then your crosstalk will be lower, especially for small numbers of inputs. Instead of 150 ohms your un-slugged bus impedance will be just R/N where R is the bus feed resistor value and N is the number of channels. A bit of maths shows the bus loss is equal to N. So with 10 inputs the bus loss is only 20dB. The cross talk will also increase to 70dB because the bus loss has dropped to from 40dB to 20dB. However, 70dB crosstalk is perfectly adequate for most applications so I would not be too concerned.

Cheers

Ian
 
First, that pic is amazing, PRR.

Second, I brought up unbalanced summing just to clarify for math/design purposes and differentiate between what I doing and a balanced passive network that needs a preamp for makeup gain. Trying to take the unbalanced signal then balance then sum then unbalance then make up gain then balanced output is way too much work and goes against my lazy nature. Just typing that made me tired thinking about it.

However, minimizing cross talk is of great interest to me as my first attempts (using 15k buss resistors) did have a higher level of cross talk than I wanted but I didn't seem to make a big impact on the overall performance for the particular application for which I was using it. I was not using a pan pot, but the stereo input did have have a higher than expected signal bleed from one side to the other when just sending tone. The design I'm currently scheming is a scalable PCB in bank of 8, which can combined in multiples then sent to a summing bus, which is not part of this design. This will allow for various summing amp circuits to be used. It will have mono/stereo switches for the pairs. Bus resistor network values will have to be calculated for the number of channels being used. I'm following along pretty good with the math presented, but the new unknown thing is the output impedance of the 283NV circuit which will be feeding the bus resistors. The output section of the NV differs from the AM section on the schematic, so I don't think I can use 50ohms as the value to use for crosstalk computation.

For my own initial use for this board, I will be using 16 channels which gives me about 24dB of loss and the bus resistance will be 937 Ohms. Would there be any value to slug the bus with a resistor somewhere between 300-500 Ohms that wouldn't be countered with an increase in noise? To calculate the output resistance of the NV circuit, could I do the same test used for the input resistance?

Thanks!

Paul

Thanks!

Paul
 
Last edited:
Potato Cakes said:
The output section of the NV differs from the AM section on the schematic, so I don't think I can use 50ohms as the value to use for crosstalk computation.
Click to expand...
You have to realize that the output Z of the AM is almost entirely due to the xfmr; the NV, having no xfmr, has a lower output Z, under nominal working conditions.

Would there be any value to slug the buss with a resistor somewhere between 300-500 Ohms that wouldn't be countered with an increase in noise?
Click to expand...
Ther is always a compromise between x-talk and noise in so-called "voltage-summing" mode. Your choice.

  To calculate the output resistance of the NV circuit, could I do the same test used for the input resistance?
Click to expand...
Yes; yoiu need to make sure the system does not go into non-linear operation, though. Typically, the output impedance of the NV is about 10 ohms at 1kHz (almost entirely due to the 10uF capacitor); you have to make sure the generator can drive such a low capacitive impedance. Also the impedance being capacitive and the injection resistor being, well, resistive, you can't use the 6dB method, but rather 3dB due to the 90° phase shift.


Thanks!

Paul
[/quote]
 
abbey road d enfer said:
You have to realize that the output Z of the AM is almost entirely due to the xfmr; the NV, having no xfmr, has a lower output Z, under nominal working conditions.
Click to expand...
Except Neve summing buses were not normally driven from the transformer but from pin F.
Typically, the output impedance of the NV is about 10 ohms at 1kHz (almost entirely due to the 10uF capacitor);
Click to expand...
I think you mean the 22uF output capacitor.

Cheers

Ian
 
ruffrecords said:
Except Neve summing buses were not normally driven from the transformer but from pin F.
Click to expand...
I didn't know that.

I think you mean the 22uF output capacitor.
Click to expand...
Correct, my mistake.
Anyway, the output Z would be in both cases largely determined by the output capacitor.
In the OP case, he uses the NV, with a 22uF cap, so the output impedance at 1kHz would be about 8 ohms at 1kHz.
A quick simulation shows that x-talk should be >80dB @1kHz for an 8-stem mixer.
I'm a tad surprized the OP worries about x-talk. That suggests a lay-out issue.
 
abbey road d enfer said:
I'm a tad surprized the OP worries about x-talk. That suggests a lay-out issue.
Click to expand...
Stray capacitance can be a problem. 500pF = 15K at 20KHz so 5pF of stray capacitance can lead to -40dB crosstalk. Switches can be quite bad in this respect.

Cheers

Ian
 
Regarding input impedance of a 183AM:  Penny & Giles faders in a desk were 5K log and responded exactly as to the designed law, IOW, there was no appreciable loading of the fader from the amp.
Crosstalk will never be as good as a desk riddled with zero output impedance IC buffers, but then the IC desk won't have the tone of the Neve so you picks yer poison..

There are a number of original Neve custom summing boxes I've fixed up or played around with that were originally built for mostly broadcast use.
For line level, most I've seen were 4 in, 1 out. 
Typically they were:
31267 input transformer into -
channel level control 1K log pot with // R so load on xnsform. then being close to 600 ohm into -
BA438 input amp set at nominal gain into -
Usually 6K8 mix resistors into -
BA440 output amp and transformer.
Naturally you can substitute BA183NV and BA183 AM for BA438 & BA440 respectively.

One other thing regarding voltage summing in old transistor desks:.  Dick Swettenham was also a proponent of voltage summing into a mic amp and then slugging down the impedance.  His reason for that method was that disconnecting a channel didn't change the gain of the mix amp to as great an extent.
Granted, no two Helios desks are the same, and the early Olympic desks were not particularly favoured by the likes of Glyn Johns due to headroom issues but, mostly, no one bothers too much about the particularly bad crosstalk or a bit more noise on their vintage desks.  Same with the 80 series Neves that are still intact.
 
> First, that pic is amazing, PRR.

It is an art-print that was recently auctioned for hundreds of bucks. There was another good hit for Ian's "sunning bus" typo but it was just a woman in a bikini, tourist-grade. While I don't get thrills looking at guys with their Speedo pulled low, I did think it was an interesting image.

> minimizing cross talk is of great interest

Buy and Read Doug Self's Small Signal Audio book. He does not tell all his tricks but it is a great overview of the problem. Also some old-old scans from a guy name JR. Also the mixer design section of Handbook for Sound Engineers.

But is crosstalk a REAL problem in studio mixers? It's all going to the same place. 40dB down should be masked by the bulk of the mix. Where crosstalk really bites is in radio network central consoles: very different programs flowing through and you don't want football cheer leaking into Flower Hour.
 
PRR said:
You don't want football cheer leaking into Flower Hour.
Click to expand...

Speak for yourself Paul.  'Flower Hour' has gone downhill in my opinion,  a few cheers would liven things up.

PRR said:
But is crosstalk a REAL problem in studio mixers?
Click to expand...

In a little stem-summing box such as Potato Cakes is building, no.  Even in a big desk with 24 - 64+ channels feeding the final mix, still no, not really.
In the old days when we used grouping and pan to send disparate signals to different tape tracks when recording, it was something to keep an eye and ear on.  But so was making sure not to put your lead vocal or solo cello track next to the SMPTE track on the machine. 

I 2nd the Self 'Small Signal' recommendation.
 
Hello, Everyone,

I should clarify that crosstalk is of great interest to me in the context of learning more about audio circuits but not a real concern in actual practice. Since it is something to consider in process of determining a bus resistor value to work with the input impedance of summing amp circuit I would at least like become more knowledgable about the subject for future designs in case there is a need/want from myself or client to minimize it. For the sake of simplicity and subjective audio preferences for the circuit I'm working on keeping the buss network passive and avoiding ICs would be best.

As for the subject of this post, I think I have the information I need to do some more math and testing to determine which design aspects and compromises to utilize.

Thanks!

Paul
 
Last edited:
Potato Cakes said:
Hello, Everyone,

I should clarify that crosstalk is of great interest to me in the context of learning more about audio circuits but not a real concern in actual practice. Since it is something to consider in process of determining a buss resistor value to work with the input impedance of summing amp circuit I would at least like become more knowledgable about the subject for future designs in case there is a need/want from myself or client to minimize it. For the sake of simplicity and subjective audio preferences for the circuit I'm working on keeping the buss network passive and avoiding ICs would be best.

As for the subject of this post, I think I have the information I need to do some more math and testing to determine which design aspects and compromises to utilize.

Thanks!

Paul
Click to expand...
Please note that the best way of minimizing x-talk in this application is to feed the injection resistors from a source of very low impedance, typically the output of a buffer. A further level of improvement is using Virtual Earth mixing.
 
Pretty much everything you'd need as reference for dealing with crosstalk is included in the Doug Self book PRR recommended. 
Abbey mentioned the very low output Z buffers and virtual earth mixing, and Ian mentioned capacitative crosstalk through switches etc.
Another related issue in mixer design is the "offness" of slugged linear taper pan pots at either end of the throw.  Due to the pull up resistor used for the slugging, the best that is usually achieved is some 70dB down.  Self explains this then reveals how, using a negative impedance technique, this 70dB figure can be improved by another 20dB.  Negative Impedance might sound like the stuff of an alternate universe but, once you see how he does it, it makes perfect sense.

 
> determining a buss resistor value to work with the input impedance of summing amp

Think "resistor ratios".

And it is a 3-stage (at least) process: source, mix R, sum point, other mix R, other "source" now acting as victim.

And "resistance" in practice includes capacitance everywhere and inductance of bus and summing node.
 
80hinhiding said:
I've experimented with both in one channel, and it's interesting.  The Neve circuit imparts its thing even after one amp, which is kind of a great thing that we can design and get the amount of poison we're after, whatever dose that may be.
Click to expand...
Thing is, what is commonly called "warmth", "punch", "sizzle"... is actually the addition of harmonics and intermodulation products. Once they're added to the signal, there's no way to remove them, so adding IC's to a signal path cannot "sterilize" it.
 

Similar threads

StarTrucker
Mastering Lab LA2As Cap Input
Replies
3
Views
330
beatnik
B
ERRP55
Understanding & Calculating In/Out Impedance
Replies
2
Views
1K
sodderboy
S
S
v672 (and more general) input impedance
Replies
0
Views
212
sea mouse
S
P
Passive low pass circuit on output transformer secondaries
2
Replies
32
Views
2K
Potato Cakes
P
Nidare
Confusion about output driver and signal levels
2
Replies
20
Views
926
Newmarket
Newmarket

Latest posts

Back
Top