“H / Balanced-T” Attenuator structure

[systemtruck]

Hey all,
I’m building a pair of H attenuators because I can’t find what I want anywhere on the market. These will be used for 600ohm mixer outputs and will be (hopefully) designed to feed modern inputs that are generally 10K input impedance.
Two questions..

1: In practical use.. is Shorting or Non shorting going to give the cleanest listening experience? Shorting says to me that the signal will be momentarily double attenuated, non shorting says to me that the signal will be momentarily disconnected. I’m thinking shorting is way better but I’d hate to make the decision and only realize I was wrong after they’re built.

2: In some H circuits there is a common/ground point in the middle/center of the “resistor” that crosses between the positive and negative. In other H circuits there is just a “resistor”.
First image here is from the Daven attenuator catalog. https://www.technicalaudio.com/pdf/Daven/Daven_Catalog_1967.pdf
Note the common point in Balanced H example.
Also see this tutorial here (half way down page), which contains an example near the bottom that like the Daven has a common point.

But if you go to places like here and here, which are H attenuator circuit calculators, you’ll see that both do NOT have a common/ground point in the center of the resistance crossing between positive and negative.


Anybody have any suggestions?
Thanks!

 

[emrr]

Shorting. And FWIW there are absolutely Zero cases of H attenuators being used in vintage consoles, UNLESS it's feeding a miles long Telco line, and even then it's probably a T feeding a 111-C for balancing.

 

[systemtruck]

Shorting. And FWIW there are absolutely Zero cases of H attenuators being used in vintage consoles, UNLESS it's feeding a miles long Telco line, and even then it's probably a T feeding a 111-C for balancing.

Yeah i figured this is overkill, but it’s an unusual set of outputs I’m patching/creating and this is what makes sense to me.
Shorting, sounds good thanks.
How about that common point in the middle of the cross resistance?

 

[emrr]

Meant to be grounded, though doubtful you'd need to do so.

 

[ccaudle]

used for 600ohm mixer outputs and will be (hopefully) designed to feed modern inputs that are generally 10K input impedance.

What is the output level of the mixer? Is the level so high that you need to attenuate or a modern device will clip? Even most USB interfaces can handle 20dBu on the balanced inputs.
Just making sure you are not under the mistaken impression that a device with 10k Ohm input impedance needs the driving device to have 10k Ohm output impedance.

non shorting says to me that the signal will be momentarily disconnected.

Or worse will be momentarily at full level with no attenuation at all.

How about that common point in the middle of the cross resistance?

There is no single answer for which way is best which is always accurate. The best scheme depends on the particular input circuit and the level of common mode noise present. Including a jumper to let you choose connected or not would be the most flexible.

 

[systemtruck]

What is the output level of the mixer? Is the level so high that you need to attenuate or a modern device will clip? Even most USB interfaces can handle 20dBu on the balanced inputs.
Just making sure you are not under the mistaken impression that a device with 10k Ohm input impedance needs the driving device to have 10k Ohm output impedance.



Or worse will be momentarily at full level with no attenuation at all.



There is no single answer for which way is best which is always accurate. The best scheme depends on the particular input circuit and the level of common mode noise present. Including a jumper to let you choose connected or not would be the most flexible.

Thanks. Yes the idea is to be able to drive the final stages of a stereo tube summing amplifier to high levels, to intentionally cause heavy distortion. The output transformers have 600 Ohm secondaries. And they can handle very high levels, so they’ll all be fine. The attenuation is to be able to have the freedom to drive the summing and turn down everything back down, post transformers, before hitting digital converters.

I do also have a handful of nice 600:600 transformers on hand unused so far, so potentially could stick with a simpler unbalanced attenuator design as opposed to the H design, take an unbalanced output from the output trans secondaries and go through the attenuation and then into the 600:600 transformers to be balanced again going out into the real world.

Main goal with the attenuation design overall is to avoid having insertion loss, to stay as close to near zero as possible. I suppose with the method just mentioned, going back into a 600:600 transformer would provide the same makeup gain as was lost by going unbalanced from the output trans and into attenuator stage. So maybe that’s the easiest solution to build.

 

[ccaudle]

Main goal with the attenuation design overall is to avoid having insertion loss, to stay as close to near zero as possible.

I don't understand this statement. The purpose of an attenuator is to add loss, can you expand on what you mean by avoiding insertion loss?

 

[systemtruck]

I don't understand this statement. The purpose of an attenuator is to add loss, can you expand on what you mean by avoiding insertion loss?

Was referring to attenuator designs overall. Ladder attenuators have an inherent insertion loss. Put a ladder type attenuator in the signal path and even with the attentuor itself turned to “zero” attenution, and you’ll still have decibels of loss. Not the nicest idea for a mixer master bus output, especially when mixing down to tape when you want to subtely adjust how hot the stereo signal is going to tape. For an example of the loss.. In the case of a 600 out / ladder / 600 in, you get 6 dB of level drop even with attenutor set to zero. This is charted in the bottom of page 10 of that Daven catalog linked in first post. The ideal scenario is being able to switch ON the attenuator and start from full level and go down in small increments, not drop suddenly 6dB as this starting point.
According to that Daven catalog, the T (unbalanced) and H (balanced) types allow for zero loss when attenuator is put in the pathway.
But that is in the perfect world where my output and input, on either side of the attenutor, are in fact perfectly 600 ohms. Interfacing with other gear is gong to present a higher load to the output of the attenuator. An expensive solution would be to find a pair of 600:10K transformers for post-attenuator. Then in theory I’d have “zero” loss from the attenuator when they are set to zero attentuation. But maybe it’s possible to adjust all of the resistor values in the attenuator design to be meant for seeing 10K. I plan to experiment with those online calculators a bit.

 

[ccaudle]

Interfacing with other gear is gong to present a higher load to the output of the attenuator.

Just put a switchable load resistor across the output. If you are going into a 10k load but want the same response you get driving gear with 600 Ohm input impedance then switch a 649 Ohm resistor across the output.

 

[systemtruck]

Just put a switchable load resistor across the output. If you are going into a 10k load but want the same response you get driving gear with 600 Ohm input impedance then switch a 649 Ohm resistor across the output.

Oh right !!

 

[Brian Roth]

Just put a switchable load resistor across the output. If you are going into a 10k load but want the same response you get driving gear with 600 Ohm input impedance then switch a 649 Ohm resistor across the output.

604 Ohms is a standard 1% value....620 standard 5%.....Just Sayin'! <G>

Bri

 

[ccaudle]

604 Ohms is a standard 1% value....620 standard 5%

649 is also a standard 1% value. To get right at 600 load with a 10K in parallel was something like 640 Ohms, so I went to the next standard 1% value higher so that the parallel combination would not be lower than 600.

 

[Brian Roth]

649 is also a standard 1% value. To get right at 600 load with a 10K in parallel was something like 640 Ohms, so I went to the next standard 1% value higher so that the parallel combination would not be lower than 600.

Point taken! It all gets "goosey" when dealing with unknown bridging load impedances, especially with a patchbay and possible mults involved. I recall that Ampex chose a 620 Ohm value (1/2 Watt carbon) for the "TERM" switch output load on the AG-440 machines.

I believe that we both would agree to a minimum 0.5 Watt rating for a "600-Ohm-ish" load resistor to handle worst case conditions. Or (as is often the case!) am I over thinking all this?

Bri

 

[ccaudle]

minimum 0.5 Watt rating for a "600-Ohm-ish" load resistor

That is a good point, 24dBu into 600 Ohms is right at a quarter W, definitely not good practice to run right up to the rated power

 

[Graham Nystrom]

I’m confused about the insertion loss and makeup gains you propose from the attenuator and 600:600 transformers. A Daven ladder attenuators lowest step is a straight wire with no insertion loss and a 600:600 will have significant insertion loss in my experience. Even wired 1:2 a transformer has very little free gain once you subtract it’s DCR.

A term resistor is often needed to flatten the hf response of a vintage 600 ohm transformer output that feeds a high z input. It will drop overall level too.

I wish I had the perfect answer for you but I only have questions. Also it’s early the am and haven’t had coffee so my apologies if I missed something or got some info wrong.

 

[emrr]

A Daven ladder attenuators lowest step is a straight wire with no insertion loss and a 600:600 will have significant insertion loss in my experience. Even wired 1:2 a transformer has very little free gain once you subtract it’s DCR.

A term resistor is often needed to flatten the hf response of a vintage 600 ohm transformer output that feeds a high z input. It will drop overall level too.

A Daven ladder is indeed -6dB at max output, the catalog mentions it and I've measured it many times. The trick around that is to use one with a 'cue' output, that is indeed unity throughput. I think I drew that out on a Langevin 117A post here.

Many quality 600:600 in my experience don't show much change with or without a termination resistor, outside of the loading loss, also variable depending on the drive Z. If there's change, it's frequently well above 20kHz, that and phase differences usually swamped by the realities of the preceding tube amp response. Usually you expect roughly 1dBish transformer loss in a 600:600, apples to apples. Every little change does something though, and there can be surprises.

Curious where you've seen otherwise.

 

[limo]

Tokyo Ko-on Denpa bridged-H has no common/ground

https://www.tkd-corp.com/en/products/att/type-d-spec.html

 

[systemtruck]

Actually, that’s an interesting site because it illustrates the two side by side. There is a common in the final HN version. So it seems optional, “as needed” for an application.
Anybody able to explain what the application difference is for the third vs fourth option they show here?

P : Potentiometer (Unbalanced circuit)
T : Bridged-T (Unbalanced circuit)
H : Bridged-H (Balanced circuit)
HN : Balanced Bridged-H (Balanced circuit)

Tokyo Ko-on Denpa bridged-H has no common/ground

https://www.tkd-corp.com/en/products/att/type-d-spec.html

 

[limo]

This is the website for a shop in Akihabara Sanei-musen, but there is no English version.
http://www.san-ei-denpa.com/jyucyuu/013/indexi.html

Google translate and summarize...

The H type and HN type are connected as shown in the diagram above, with an insertion loss of 0 dB and input/output impedance constant, and the attenuation level adjusted in stages.
It can be changed without disturbing the alignment of the line.
The H type and HN type are balanced types and will operate reliably when inserted into a balanced circuit, but they should not be inserted into a circuit where the balance is disrupted.
If used, errors will occur.
For example, if the line is long, use a bridge balanced H type (HN type) connection and connect the midpoint N to the "ground".
You can get good results by doing

 

[Graham Nystrom]

A Daven ladder is indeed -6dB at max output, the catalog mentions it and I've measured it many times. The trick around that is to use one with a 'cue' output, that is indeed unity throughput. I think I drew that out on a Langevin 117A post here.

Many quality 600:600 in my experience don't show much change with or without a termination resistor, outside of the loading loss, also variable depending on the drive Z. If there's change, it's frequently well above 20kHz, that and phase differences usually swamped by the realities of the preceding tube amp response. Usually you expect roughly 1dBish transformer loss in a 600:600, apples to apples. Every little change does something though, and there can be surprises.

Curious where you've seen otherwise.

Perhaps I’m thinking of a Daven that had the cue step. They have many versions of those attenuators. I sold the last of mine a while back but I remember measuring no resistance on last step. I’ll have to consult the Daven catalog to figure out which one I was using and measuring. And it’s possible I’m just misremembering!