I am having a problem understanding input/output impedance and which type of attenuator to build along with which values of resistors to use.
My problem is that my preamp/headphone amplifier never goes past 10 o'clock where sound quality problems occur and channel imbalance are apparent.
Here is the setup:
DAC (Topping D10): 100 ohm output impedance
Preamp/Headphone amp (JDS Atom) 0.7 ohm output impedance (unsure of input impedance ?)
Power Amplifier (3E tpa3255): 10K ohm input impedance (unbalanced)
I am looking for a 12 - 16db level of attenuation. Which type of attenuator should I use, L, T or Pi? What values of resistors should I use for 12b and 16db.
I would prefer the attenuator to go between the dac and preamp so that when using headphones I can still turn up the preamp/headphone amp past 10 o'clock.
I have seen options such as Rothwell In-Line Attenuator -15db but there are no specifications on impedance so I am not sure if it would be applicable to my setup. I believe they are more so designed to be used at the power amp inputs.
I will not be using digital attenuation as an alternative.
Thanks!!!
My problem is that my preamp/headphone amplifier never goes past 10 o'clock where sound quality problems occur and channel imbalance are apparent.
Here is the setup:
DAC (Topping D10): 100 ohm output impedance
Preamp/Headphone amp (JDS Atom) 0.7 ohm output impedance (unsure of input impedance ?)
Power Amplifier (3E tpa3255): 10K ohm input impedance (unbalanced)
I am looking for a 12 - 16db level of attenuation. Which type of attenuator should I use, L, T or Pi? What values of resistors should I use for 12b and 16db.
I would prefer the attenuator to go between the dac and preamp so that when using headphones I can still turn up the preamp/headphone amp past 10 o'clock.
I have seen options such as Rothwell In-Line Attenuator -15db but there are no specifications on impedance so I am not sure if it would be applicable to my setup. I believe they are more so designed to be used at the power amp inputs.
I will not be using digital attenuation as an alternative.
Thanks!!!
You need a better volume control, but input attenuation will at least let you run the control at a higher setting. Try a 10k resistor in series with each amplifier input, which will give you 6dB of attenuation.
If that is not enough, use a 30k series resistor for 12dB loss.
If that is not enough, use a 30k series resistor for 12dB loss.
You're wrong! When the LINE word is used it means that the signal is at line level, which might be between 0.1 and 2 V, say; typically 0.3 V. So that might work - just in case you don't have the time to find some spread sheet/calculus or web page that explains a resistor net.
That would be important because you have to connect a potentiometer ( logarithmic scale, not linear ) of the right value, and the lower the better, say 10 kΩ.
What nominal levels are the various amps inputs and outputs?
You have no issues with impedance, in each case the output impedance is way below the next device's input impedance.
Sounds like a fault with the preamp, perhaps one channel is going into parasitic oscillations at higher volume. An attenuator is likely to require more amplitude from the preamp and thus more problems.
There might be other issues, more investigation is needed, but I don't think attenuation can ever compensate for an amplifier fault.
A correctly set up chain of amplifiers only limits when the power amplifier clips, and you wouldn't see a channel imbalance with clipping (unless using two completely different monoblocs).
You have no issues with impedance, in each case the output impedance is way below the next device's input impedance.
Sounds like a fault with the preamp, perhaps one channel is going into parasitic oscillations at higher volume. An attenuator is likely to require more amplitude from the preamp and thus more problems.
There might be other issues, more investigation is needed, but I don't think attenuation can ever compensate for an amplifier fault.
A correctly set up chain of amplifiers only limits when the power amplifier clips, and you wouldn't see a channel imbalance with clipping (unless using two completely different monoblocs).
Last edited:
Wouldn't the power amp input be a line level?
Voltage out of the dac is 2v while voltage out of the preamp is obviously lower. I wasn't sure if that item was designed to be used before or after preamp.
Thanks
I'll give that a shot. Is that all the RCA attenuators are, one resistor? Which wattage resistor would be suitable? Would it go before or after preamp?
Thanks
You want an L-pad for most hi-fi situations.
You *can* design an L-pad using the load (input impedance of the next box) as the bottom of the L. Problem is that this tends to lead to high values, and the input Z number in the specs is only nominal and sometimes just wrong.
High values and long lines or caps on amp inputs lead to treble loss.
In your case you could aim for say 2k into the L. 1k+1k loaded in "10K" gives a bit more than -6dB and vast tolerance for mis-specs and stray capacitance. 1.8k+680 gives about -12dB. A string of ten 220r resistors lets you try taps from -1dB to -20dB.
For quick-test you buy a 100-pack of resistors and a 99c RCA cord. Cut and strip the cord, twist resistors in. The resistors out in the open (also cheap connectors) will attract hum/buzz and twist-connect is not reliable.... just do this to zone-in on a happy amount of loss, then do it up right with solder. In an adverse room (nasty lighting, wifi/cell) you may need a metal box bonded to RCA shell. But in benign rooms I have often go away with short leads (minimum antenna length) just taped-up.
You *can* design an L-pad using the load (input impedance of the next box) as the bottom of the L. Problem is that this tends to lead to high values, and the input Z number in the specs is only nominal and sometimes just wrong.
High values and long lines or caps on amp inputs lead to treble loss.
In your case you could aim for say 2k into the L. 1k+1k loaded in "10K" gives a bit more than -6dB and vast tolerance for mis-specs and stray capacitance. 1.8k+680 gives about -12dB. A string of ten 220r resistors lets you try taps from -1dB to -20dB.
For quick-test you buy a 100-pack of resistors and a 99c RCA cord. Cut and strip the cord, twist resistors in. The resistors out in the open (also cheap connectors) will attract hum/buzz and twist-connect is not reliable.... just do this to zone-in on a happy amount of loss, then do it up right with solder. In an adverse room (nasty lighting, wifi/cell) you may need a metal box bonded to RCA shell. But in benign rooms I have often go away with short leads (minimum antenna length) just taped-up.
It figures, just after reading 3 pages of someone who built a T Pad Attenuator, who also used about 2k. New toys: Parasound JC3 phono preamp & Lyra Delos cartridge | Page 4 | Steve Hoffman Music Forums
I emailed the preamp manufacturer to see if I can get the input impedance, but I am thinking the output impedance of .7 measured (.1 is spec) is the headphone output and not the preamp output impedance...
The high end resistors aren't cheap so I'll probably get some cheap ones to experiment with and I have rca connectors with screw terminals I can use for testing.
What about values for between the dac and preamp/headphone amp?
I'd ideally like to attenuate the dac signal (not digitally) so I can turn up the preamp/headphone amp and eliminate the issues with the pot at such a low level.
I emailed the preamp manufacturer to see if I can get the input impedance, but I am thinking the output impedance of .7 measured (.1 is spec) is the headphone output and not the preamp output impedance...
The high end resistors aren't cheap so I'll probably get some cheap ones to experiment with and I have rca connectors with screw terminals I can use for testing.
What about values for between the dac and preamp/headphone amp?
I'd ideally like to attenuate the dac signal (not digitally) so I can turn up the preamp/headphone amp and eliminate the issues with the pot at such a low level.
12-16 dB with an L-pad at consumer line level is super easy.
Maybe 6k8 : 2k2, 8k2 : 2k2 or 10k : 2k2 (or 6k8 : 1k5). Or even a bit lower if the D10 is fine driving that (4k7:1k5 to 4k7 : 1k). L-pad input impedance 5-10+ kOhms, output impedance <2 kOhms, these are values that should often work perfectly fine. Few modern devices should be bothered driving around 10 kOhms, which is a typical input impedance as well.
No fancy resistors required, ordinary 1/4 W 1% metal film jobs will be plenty.
Maybe 6k8 : 2k2, 8k2 : 2k2 or 10k : 2k2 (or 6k8 : 1k5). Or even a bit lower if the D10 is fine driving that (4k7:1k5 to 4k7 : 1k). L-pad input impedance 5-10+ kOhms, output impedance <2 kOhms, these are values that should often work perfectly fine. Few modern devices should be bothered driving around 10 kOhms, which is a typical input impedance as well.
No fancy resistors required, ordinary 1/4 W 1% metal film jobs will be plenty.
I found out that the input impedance of the preamp/headphone amp is > 10k.
Is there any benefit of L pad vs T pad? I'm guessing to keep the ratios better between input and output.
Ordered a resistor kit off Amazon to play with.
My OCD will probably make me buy .01% vishay foil resistors though for the finalized implementation.
Thanks!
Is there any benefit of L pad vs T pad? I'm guessing to keep the ratios better between input and output.
Ordered a resistor kit off Amazon to play with.
My OCD will probably make me buy .01% vishay foil resistors though for the finalized implementation.
Thanks!
BTW, you are already using the Atom's low gain setting (1x), right? The difference between low and high gain is about 13 dB...
Resistor values on the edge of 5% tolerance could obviously upset channel balance by a fraction of a dB if combined non-ideally, but if in doubt that's what you've got a multimeter for.
T-pads are used in controlled impedance systems. In audio you are just doing voltage transfer (short into open), so minimum source impedance it is, hence L-pad.
Honestly, you could probably use carbon film and it wouldn't matter. Who cares if they have voltage and temperature coefficient - as long it's the same for both, most of is is going to cancel out in a voltage divider. Not perfectly as we are talking a loaded voltage divider, but only about (R1||R2)/Rin is going to remain, so maybe 10-20%? Nobody would be using volume pots with carbon tracks if this were such an issue.
Resistor values on the edge of 5% tolerance could obviously upset channel balance by a fraction of a dB if combined non-ideally, but if in doubt that's what you've got a multimeter for.
Last edited:
Yep, low gain on the Atom.
I saw a pair of 12db attenuators for $25, though that's no fun.
https://www.amazon.com/gp/product/B0006N41B0/ref=ppx_yo_dt_b_asin_title_o01_s00?ie=UTF8&psc=1
I'll build some L-pads with the cheap resistors and see how it sounds.
For final installation what's the best way to install them?
Make custom (directional) cables with the attenuator built in or a shielded box with inputs and outputs, which would require purchase of yet another set of rca cables.
Thanks
Finally got around to trying this, how does the following look?
8dB attenuation
Series: 3K3
Shunt: 2K2
Would the output impedance be too high?
Thanks
Output impedance would actually be OK, input impedance is getting a tad low but should still be acceptable. I wouldn't say I'm thrilled but it should work.
If you need that little attenuation, I would be considering going for the high gain setting on the Atom and ~13 dB more on the attenuator. 10k : 1k should give you about 21-22 dB, with a more favorable ratio of input and output impedance. (Or 12k : 1k2 or even 15k : 1k5.)
(If the Atom is anything near the O2 it is based on, it has its gain stage ahead of the volume control, which means any difference in noise between gain settings should only be apparent with volume turned up near max, which does not seem a likely scenario in your case. Also, gain stage noise should be quite low either way.)
BTW, what are your levels on the digital side like? Are you using ReplayGain or similar at all? As I need to accommodate a collection of material spanning all the way from the early days of the CD to today, modern material often loses a total of 10-13 dB here (I have an extra -3.2 dB dialed in). An argument for 24 bit output if nothing else. The D10 seems to offer a dynamic range of around 110 dB, so CD material is not likely to be in danger any time soon.
If you need that little attenuation, I would be considering going for the high gain setting on the Atom and ~13 dB more on the attenuator. 10k : 1k should give you about 21-22 dB, with a more favorable ratio of input and output impedance. (Or 12k : 1k2 or even 15k : 1k5.)
(If the Atom is anything near the O2 it is based on, it has its gain stage ahead of the volume control, which means any difference in noise between gain settings should only be apparent with volume turned up near max, which does not seem a likely scenario in your case. Also, gain stage noise should be quite low either way.)
BTW, what are your levels on the digital side like? Are you using ReplayGain or similar at all? As I need to accommodate a collection of material spanning all the way from the early days of the CD to today, modern material often loses a total of 10-13 dB here (I have an extra -3.2 dB dialed in). An argument for 24 bit output if nothing else. The D10 seems to offer a dynamic range of around 110 dB, so CD material is not likely to be in danger any time soon.
Last edited:
Hmm, that's interesting. One of the setups I was testing with, for music, 12db was fine, and I could even go with more attenuation. When playing games though, the volume was lower and with 12db I could max it and still not be overly loud, though a bit higher than I'd normally use anyway.
I'll try the higher level of attenuation and see how it is for music. For games if necessary I could set the atom on high gain, or figure out why they are at a lower volume to begin with.
Thanks again! Will report back soon 🙂
That sounds like a good reason to introduce some digital attenuation in your music player (one of the most benign operations if done at a sufficient resolution, and the vast majority of players is likely to be using float32 samples these days). See the edit to my last post.
I went back and tried one of your original examples, 4k7 : 1k.
About 15db of attenuation, and with music I am at about 12 o'clock on the Atom using speakers and 1 o'clock using headphones.
With games, youtube, etc I am at about 2 o'clock.
Originally I was somewhere between 9 and 10 o'clock.
I think we have a winner.
I'll play around with turning down the music app, I just don't trust a software update or something resetting it back to 100% to rely on it.
Attaching a pic of the test setup 🙂
I have a second setup with the Topping D50s and Atom where I am using the D50s digital volume control (needed remote volume control) but it's too low for my liking so I'll likely do the same thing there after testing this config for a bit. It's a higher end setup though so I'll need to do some serious listening to see how it sounds.
Thanks again!
Attachments
Further testing and investigating....
15.1dB 2K2 : 470
Dac output impedance: 100, a factor of 22
Preamp input impedance: 10k, a factor of about 22
Does my math check out?
I think I am done, time to order parts for the final build...
15.1dB 2K2 : 470
Dac output impedance: 100, a factor of 22
Preamp input impedance: 10k, a factor of about 22
Does my math check out?
I think I am done, time to order parts for the final build...
Last edited:
In modern world, computer uses you.
Look, I grew up in a world where you had to update your software by hand, and it generally didn't spontaneously nuke your config. I've been dragging along the same Foobar2000 install for nigh on a decade now (in portable mode since v1.0.x, which keeps everything in its directory), updated as needed. It never lost my settings or playlists, and that's exactly how it should be. This particular player admittedly may be a bit of an extreme, but since it's so flexible you may be tweaking things for months or years and might end up looking at a considerable amount of work to restore a tricked-out setup from scratch. This is basically an "investment" approach, not a "throwaway consumer goods" one.
Find a good music player that makes it easy to access and back up its config, and you should not have to worry about such issues.
Kinda - I don't think you've got your impedances calculated quite correctly but it's in the ballpark.
The DAC sees a load impedance of 2k2 + (470 || 10k) ~= 2.65 kOhm.
The preamp sees a source impedance of 470 || (2k2 + 100) ~= 390 ohms.
You may still be missing a few other things to consider.
a) Do you know the size of output coupling capacitors in the D10, if any?
b) Do you know whether the output amplifier will still be happy driving less than 2.7 kOhms?
Re: (a), I looked at pictures again and possibly this DAC has a DC-coupled output amplifier, as I failed to spot any good candidates for output coupling caps. Probably uses a split power supply internally.
Re: (b), the OPA2134 has been used in small headphone amplifiers and makes a fairly decent load driver, roughly on the level of a NE5532 or so. It seemed basically unfazed driving 2 Vrms into ~1.5 kOhms at a gain of 7 with 0.001% of distortion or less (tested here). At lower-impedance loads, high-frequency performance progressively worsens.
So it looks like the DAC should drive this quite happily. The little 2-transistor preeamp circuit that I was simulating earlier might break a bit more of a sweat. It would be up to 0.005% of THD at 1 kHz, from 0.0004% with a 50 kOhm load. Still nothing audible by far, mind you. 600 ohms would be a different story though, with close to 0.03% (if dominant 2nd, so probably still OK). I imagine an OPA2134 would be a good bit less fussed.
If I only used foobar or a dedicated music app, maybe I'd trust it more. I mostly stream and seem to be changing services as I've yet to find the "best" one. Tidal, Qobuz, Deezer, Amazon Music HD, and even Spotify.
As a random test I set the computer volume at 50%, and changed usb ports for my dac and it reset the volume to 100% as it saw it as a different device. Strange, but true.
I'm a software developer by trade and know too well how easy it is for things to go awry. Things that "should never happen" somehow find a way.
You lost me a bit here.
The output/power amplifier should have an input impedance of 10K, while the atom has a rated output impedance of 0.1 ohm.
I thought the rule of thumb was low output impedance --> high input impedance. I've seen the "rule of thumb" as anywhere from 8x to 20x between devices.
This is the formula I was looking for to compute input/output impendance.
Can you break this down, as a formula? I don't understand the (470 || 10k) part.
Thanks!!