Stepped attenuator (potentiometer) giving unexpected results

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I had posted this under the Analogue forum and didn't get any response. Someone there recommended that I repost here. Please help. I have run out of ideas....

I need help diagnosing and correcting a problem I am having with a DIY stepped attenuator.

I use a Digidesign 002 and ProTools LE for recording. Playback is through a Mackie HRS-120 powered subwoofer that picks off the low frequencies from each channel and then sends the rest of the signals to a pair of Mackie 624 powered monitors. Until now I have been taking the output signal from digi 002's monitor outputs, which have a volume control pot. It works well, but it is impossible to reset the volume to exactly the same level due to the absence of detents.

So I built a stepped attenuator for volume control with the intention of connecting it between the main outputs of the digi 002 and the HRS-120. The attenuator is a series resistor design with both balanced inputs and outputs. The input impedance is 6.3 kOhms, and the output impedance ranges from 0 to 6.3 kOhms over 19 steps. Inputs and outputs are balanced via 1/4" TRS connectors.

The attenuator checks out fine. Impedances match within <<0.2% over most of the steps. I also tested it at several frequencies using a signal generator and oscilloscope, and the attenuation results matched expectations at all conditions.

When I connect it into the system, however, results are not good. First, instead of being silent at the first position, there is sound from the monitors (mostly the right monitor). Also there is a gross imbalance in the sound level coming from the left and right monitors over the next 4 switch positions (with the right being louder). It isn't until the fifth switch position (about 100 Ohms output impedance) that the sound levels match.

To diagnose the problem, I swapped the left/right inputs and outputs from the attenuator to see if one pair of decks was bad. There was no change...the right speaker remained the louder. I replaced all of the cables in the signal path, and again no change. I swapped just the inputs and then just the outputs to the attenuator, and again no change. Only when I swapped the left/right outputs from the subwoofer did the louder sound shift to the left monitor. So the attenuator seems fine.

I then removed the subwoofer from the signal path and connected the attenuator outputs directly to the monitors, and they worked perfectly. Silent at position zero, and equal in left/right volume balance at each higher position. So the problem would seem to be with the sub....but the sub works fine when connected to the digi 002's monitor outputs???

Based on these results, there is an interaction between the attenuator and the subwoofer. What could be causing this behavior, and how can I fix it?

For further background... All other devices in the signal path have balanced inputs and outputs. Both the sub and monitors have input impedances of 20 kOhms. Also, I tried inserting a 56 Ohm resistor into each signal line of the attenuator output (after the stepped switch) to increase the output impedance to a minimum of 112 Ohms (based on the observation that the system worked well when the switch was in the 5th position where the output impedance had reached about 100 Ohms). Doing this did not correct the behavior.

Thanks to any who have read this far. And thanks in advance for any insight.
 
I think that the inputs are already buffered

Thanks for the quick response and the suggestion. I am ok with passive circuits, but my facility with active circuits is limited. That said, though, I could probably work out what I need to buffer the inputs to the HRS-120 subwoofer.

Before I do that, though, I could use some help in determining if the inputs are already buffered. Attached is a copy of just the input area of the HRS-120 schematic. Are those first devices after the input connectors Op-amps? If so, does that mean that the inputs are already buffered? If not, what are they and what are they doing?

Thanks again for your help.

Barry
 

Attachments

  • hrs120 input.jpg
    hrs120 input.jpg
    53.3 KB · Views: 567
The block diagram would suggest that the two channels are symmetric, so either you can buffer (there are some excellent opamp buffers available) or you can dig around and figure out why one channel is off. First thing to determine is which channel is working properly so you know where to look; with an input test signal of 50-100Hz, use an AC voltmeter and plot output voltage versus control position for each channel.

Is anything inadvertently connected to the input phono jacks?
 
A quick diagram of the whole system and schematic of the 'balanced
step attenuator' would be helpful. This can be done with passive
elements but in most cases a buffer stage can solve the problem
since it isolates and controls the impedance and thus the actual
voltage drop that will be seen by the receiving input.

A good reference for attenuators:
http://www.rfcafe.com/references/electrical/attenuators.htm
 
There are several different types of 'balanced' inputs and outputs:

Good Reference

TABLE 1: Output arrangements
1. Unbalanced output
2. Impedance-balanced output
3. Ground-cancelling output (or ground-compensated output)
4. Balanced output
5. Quasi-floating output
6. True floating transformer output


TABLE 2: Input arrangements
1) The standard differential amplifier
2) Switched-gain balanced amp
3) Variable-gain balanced amp
4) The "Superbal" amp
5) Hi-Z balanced amp
6) Microphone preamp plus attenuator
7) Instrumentation amp.

Not all balanced circuits act the same to different loads. . .

Rane Note 110

You need to know what type of balanced input and output you have
before you design the attenuator. . .For testing, you could design a fixed balanced attenuator of 6dB (T-attenuator and Pi-attenuator). . .watch your resistor values and tolerances. . . find out what type work
best with your Mackie equipment. Some balanced equipment is
very compatible with unbalanced equipment (shorting pin 3 to ground) while other balanced equipment is not. . .read over the
Rane Note 110. . .it is very good. . . .
 
Thanks to both of you

First to Sy. No there is nothing connected to the phono inputs. I am using TRS to XLR cables between the attenuator outputs and the sub inputs. And TRS to TRS cables between the outputs of the Digi 002 and the inputs of the attenuator. I have already done the experiment I think that you are suggesting. I used a signal generator at 200 Hz, 1000 Hz and 2000 Hz input to the attenuator and measured the output voltage from the attenuator. The results were as they should have been zero output at position 1, rising stepwise from there with each switch position. The outputs of the left and right channels matched within about <0.2% difference. I think the attenuator is ok.

To GNI, I have prepared four drawings in response to your question about the attenuator design and the system configuration. I don't know how to post more than one in a note, so I will post one with this note and then 3 more notes.

One drawing is the system configuration prior to adding the attenuator. The second is the system configuration with the attenuator. The third is the configuration with the attenuator outputs bypassing the subwoofer and connected directly to the monitors. The fourth contains the design of the attenuator. Let me describe each below. All connections are balanced. All sound comparisons were done using pink noise from the Digi 002.

1) System prior to attenuator - Signal originates from the Monitor Outputs of the Digidesign 002. The output level is controlled by the output potentiometer of the 002. It passes from there to the L and R inputs of the sub. In the sub, the frequencies below the crossover are routed to the subwoofer amp and the higher frequencies are sent to the output connectors and then to the left and right monitors. Everything works fine when set this way, I just don't have reproducible enough control of the volume.

2) System with the attenuator - Signal originates from the main (uncontrolled line level) outputs of the Digi 002. The left and right channels go to the attenuator, and from there to the sub inputs, and then on to the monitors as above. There is no difference in the sub settings with or without the attenuator. In this configuration I have the problem of sound even at the 1st attenuator position (which should be silent) and a gross imbalance (5 dB) in the SPL between the right and left monitors at the first 4 attenuator positions.

2) System with the attenuator bypassing the sub - Signal originates from the main Digi 002 outputs, goes through the attenuator, and bypasses the sub to go directly from the attenuator to the left and right monitors. This works perfectly. Silent at position 1 of the attenuator and then the volume rising in even steps as the switch position is increased. SPLs between the L and R speakers match exactly.

That is my quandry. Why does the sub/monitor combination work fine without the attenuator, and the attenuator/monitor combination work fine without the sub, but the triple combination of attenuator/sub/monitors work poorly?

The last drawing is of the attenuator. The upper drawing shows the overall circuit layout and the lower drawing shows the attenuation circuit. There are 2 sets of TRS inputs controlled by a 4PDT switch. Both sets of inputs give the same symptoms. The 4 signal lines each go to their own deck of a 4 deck rotary switch. I have resistors in 19 of the 24 positions of the switch. The remaining spots are empty. The outputs from the Left and Right channels each go to their own DPST switch that can be used to mute the outputs individually (they have been on throughout this testing). The output from the switches go to a pair of TRS output jacks. The signal ground bypasses the attenuator and directly connects the sleeves of each input with its respective output jack.

The attenuation circuit is a stepped potentiometer. Each pair of decks representing either the left or right channels are connected at one end. The input signal is at the opposite end of the resistor chain from where the two decks are connected. The output signal is taken from the wiper of the switch. At switch position 1 there is a short between the outputs signal legs. As the wiper moves to the left (higher switch position numbers) it inscribes greater and greater resistance, sending a greater fraction of the input voltage to the output. In effect, it is a variable "U" pad. I have also connected 56 Ohm resistors to each leg of the outputs to increase the minimum output impedance from zero to 112 Ohms (making the circuit equivalent to a variable "H" pad).

The total input impedance is 6.3 KOhms. The output impedance varies from 0 to 6.3 KOhms.

Sorry this is so long, but there is lots of information and I don't know which is important.

Also, I started writing this before I saw your last note. So I haven't yet read the reference. I apologize in advance if it answers my question and I have wasted your time posting this.
 

Attachments

  • systemlayoutnoatten.jpg
    systemlayoutnoatten.jpg
    50.6 KB · Views: 512
Sy, sorry I misunderstood

Let me try and restate what I think you're suggesting I do. Leave the attenuator connected to the Digi outputs and sub inputs. Then feed it with a sinewave signal of 50-100 Hz at line level. Then measure the output signal from the subwoofer outputs that would normally be connected to the monitors. This is where the sound level problems begin to show up.

I think I can do that. I will have to go to a higher frequency because the sub crossover is set at about 120 Hz (or thereabouts...I don't recall exactly). But I can do that, and monitor the output signal with an oscilloscope.

Is that what you're suggesting?
 
Higher freq. with oscilloscope?

Shouldn't I be able to go higher in frequency using an oscilloscope to monitor the outputs instead of a meter? I'm afraid that even at 200 Hz I might catch some tail of the crossover. And that it might be slightly different from one channel to the other. If I go to 1000 Hz I'm sure to be past the crossover.

Thanks again for your graciousness and patience.
 
Read the references

GNI, I've read the 2nd reference and skimmed the first. The sections on op-amps are a little dense for my level is expertise. But even in that first reference near the end section 6 (figs 14a and 14b) show a balanced system similar to my attenuator. Mine isn't optimum (not grounded at the centerpoint), but the the weakness is in potential hum and Common Mode Rejection. I have no hum, and I don't think that CMRR is my problem (or is it and I am missing the rationale).

My attenuator design also seems to fit the points raised in the 2nd reference. Mine is simpler than the other options, but seems ok (again, unless I am missing something).

Thanks again for your help. I have my fingers crossed that you (or anyone) will look at my drawings and see the flaw causing this behavior.
 
The Mackie HRS-120 uses balanced differential 20kohm input and
looks like (from the schematic) an impedance balanced output. . .

The 624 also uses a balanced differential 20kohm input.

I would not use a series attenuator but a balanced T or balanced Pi.
There is some sort of impedance mismatch that is interacting with
the subwoofer in/out circuit. I would build from scratch one or the
other. Insure that the output impedance of the balanced T or
balanced Pi attenuator is low (under 600 ohms). I would also
keep the output Z as close to the same at all attenuation levels.

It is necessary to keep the output Z of the source low to create
a good voltage transfer (we don't want an impedance match since
we are not trying to transfer power like the old days). This means
we always want a low source to high receiver impedance. . .more
than ten times would be great. . .

I really feel that using the series attenuator is the cause of the
problem; make a balanced T or Pi attenuator. . .I really think
that will solve the problem. I don't think the Mackie equipment
is at fault since it works well without the attenuator.


This will give you good equations for calculating correct attenuation
for balanced circuits. . .the more I look at the series attenuator
the more I feel it is wrong. . .
 
Thanks again GNI

Great reference. I am going to have to apologize in advance (again), because your response triggers another batch of questions from me.

Isn't my attenuator already a balanced "T" design? The nomenclature is different than where I was trained, but what I am calling an balanced "H" attenuator looks to me to be the same as what the rcafe site is calling balanced "T". See the drawing attached to this note. It describes my attenuator right now. I have inserted two 56 Ohm resistors in each channel (one in each signal line). To me this looks exactly like the balanced T design, except for the variable nature of R3 (using the designation from the rcafe bal T drawing). Am I misinterpreting rcafe or my design?

As for the output source impedance needing to be low, I recognize that the peak output impedance of my attenuator is too high for the Mackie's inputs. But the problems happen only at the lowest output impedances. From zero to about 100 Ohms (switch positions 1 thru 5). Starting at about 140 Ohms and up until painful SPLs, the sound is excellent and balanced left to right. At that point the output impedance is still under 2000 Ohms (< 1/10 the Mackie's input impedance). I never get to a position where the output impedance of the attenuator is >2000 Ohms. To me that says that too high an output impedance is unlikely to be the issue. Again, do you think I'm over-reaching in that conclusion?

I thought that with a passive stepped attenuator one has to make a choice between keeping the input impedance constant or the output impedance constant...and that the norm is to keep the input impedance constant so as to not variably load the source. And to keep the output impedance low such that its variation will not affect the load. That said, though, in my current configuration with the 2, 56 Ohm resistors in each channel output, the attenuator output impedance is relatively constant at the first 2 switch positions (112 Ohms and 122 Ohms respectively...R3 is 10 Ohms at switch position 2). And the problem is still there.

One final question. Could I be looking at the wrong impedance? Could a 6.3 KOhm input impedance on the attenuator be too low for the output of the Digi 002? What if its right on the edge of being too low in parallel with approx. 20 KOhms. And what if the subwoofer input impedance were slightly low (say 15 KOhms), and the monitors were slightly high (say 25 KOhms). Then when connected to the sub, the Digi would be seeing an impedance of about 4400 Ohms, and when connected to the monitors it would see an impedance of about 5000 Ohms. Are they low enough and different enough to affect a line level source differently?

Finally, are you recommending that rather than designing a variable circuit for the attenuation that I should build a series of discrete bal T attenuators each with the same input and output impedance, but different attenuation ratios (I'm not sure if that's possible, but I'll solve the equations and see if it is...I recognize that K can float since power transfer is not critical)? Then switch from one to another to achieve different output levels?

Thanks again for your support and the excellent references.

Barry
 

Attachments

  • attenuatordesignbalh.jpg
    attenuatordesignbalh.jpg
    59.4 KB · Views: 372
Did the suggested signal test...results below

I did a signal test of the system. Sent a 1 kHz sine wave from the Digi 002 outputs and monitored the signal levels at different points in the signal path:

1) Signal at the Digi 002 main outputs with nothing connected downstream - 4.1 V p-p on both channels (actually read about 4.11 on right and 4.14 on left). Waveforms appeared normal.

2) Signal at the Digi 002 main outputs with all downstream connections in place...stepped attenuator, subwoofer, and then monitors - 4.1 V p-p on both channels (actually 4.08 V on both channels...is the output voltage difference between without and with load significant, even though both were equal?). Waveforms appeared normal.

3) Signal at the output of the stepped attenuator when not connected to the subwoofer - zero at position 1 and increasing uniformly and equally between channels as the rotary attenuator switch was advanced. It reached about 300 mV at position 6. Waveforms appeared normal.

4) Signal at the output of the subwoofer with the attenuator in the circuit - The voltages behaved just like the Monitor sound levels that exist when the attenuator is in the signal path. Namely, a few mV (low but discernible sound) at the right channel and about half that on the left when the attenuator switch is in position zero. As the switch is advanced the voltages increase, but the right channel rises more quickly until about position 3 when the left channel begins to rise quickly and by position 5 the 2 signal levels matched. Again, the waveforms appeared normal.

In hindsight I probably should have tested the output of the stepped attenuator when it was connected to the subwoofer. I will try and do that this weekend. My bet is that it will be normal, and that the issue is with the sub....But why doesn't it happen when the Monitor outputs are connected to the sub?....the block diagram of the Digi 002 indicates that the Monitor output is taken from the same signal lines as the Main outputs, except that a potentiometer is added into the signal path of the Monitor outputs.

Any further suggestions would be appreciated.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.