If I put my notes here, I might be able to find them again later!
A Rant on Why Passive Preamps are Totally Stupid
Posted 16th July 2016 at 02:04 AM by rjm
Updated 20th July 2016 at 09:57 PM by rjm (corrected attenuator output impedance in attached diagram)
Updated 20th July 2016 at 09:57 PM by rjm (corrected attenuator output impedance in attached diagram)
A [just my opinion, bro] post...
I actually had occasion to try this the other week. I had a box with a volume control followed by the bboard unity gain buffer and in preparation for replacing it with a similar buffer with voltage gain (a power-derated Sapphire 3) I removed the buffer and briefly used the box as passive preamp, i.e. just the 47k stepped attuator, with 1 m interconnects to the amp and 2 m interconnects back to the phono stage. Sure enough the system noise increased, depending on the position of the volume control, with some nasty low level buzzing interference.
Why does this happen? It's pretty simple really. Noise is usually induced as a current, and the larger the resistance (impedance) this noise current is forced to flow through to reach circuit common, the larger the noise voltage since by Ohm's Law, V=IR. Noise induced between the volume control and the amp is faced with the high impedance of the amp (47k) or the output impedance of the volume control, which can be low, but at typical listening levels can be moderate to high.
If that volume control was backed by a buffer, the output impedance would be always low, and the induced currents would drain off cleanly through the buffer output without inducing any noise. While it's true that the buffer itself produces noise (and, yes, distortion) it was clearly the case that the system noise was far lower with the buffer in the circuit than without.
The noise/distortion from the amplifier and source components should always dominate unless the system gain is too high and the volume control is used at near the minimum setting.
There is a widespread perception in diyaudio that fewer gain stages are better than more gain stages.
This is dumb.
In an analog system, the signal begins as vibration of a needle in a groove and ends with the vibration of a voice coil. The signal gain is something in the order of 10,000. For a given speaker efficiency and cartridge sensitivity, the gain is fixed. If you don't have a line preamp with gain, the gain must be added to the phono stage or the power amplifier. Bragging about the "purity" of your passive preamp setup is silly: all you've done is moved the gain to a different box!
There are a very good arguments meanwhile for not adding this gain before the volume control, and several other very good arguments for not bolting it onto the power amplifier. The traditional approach, splitting the system gain between a phono stage getting the signal to a nominal line level, an attenuator typically operating at -20~-40 dB, followed immediately by a line stage / driver with 6-12 dB, the interconnects, and finally the input-driver-output stages of the power amplifier is pretty much the optimum situation from a noise and distortion point of view. The critical pain point where the signal is weakest is the output of the volume control. From an impedance standpoint it is more susceptible to noise pickup than even the cartridge output! You really don't want to extend that weak signal over the length of the interconnect and through to the power amplifier.
****
Note : the problem with a passive preamp is not just noise pickup, or at least not as I described it above as some sort of coincidental interference. The amplifier's input stage - if its a discrete bipolar design like mine is anyway - tends to have a fairly large current noise which will also develop a noise voltage across Zeff. While not such a big deal for vacuum tube or jfet input stages, bipolars are a poor choice to pair up with a passive preamp.
I actually had occasion to try this the other week. I had a box with a volume control followed by the bboard unity gain buffer and in preparation for replacing it with a similar buffer with voltage gain (a power-derated Sapphire 3) I removed the buffer and briefly used the box as passive preamp, i.e. just the 47k stepped attuator, with 1 m interconnects to the amp and 2 m interconnects back to the phono stage. Sure enough the system noise increased, depending on the position of the volume control, with some nasty low level buzzing interference.
Why does this happen? It's pretty simple really. Noise is usually induced as a current, and the larger the resistance (impedance) this noise current is forced to flow through to reach circuit common, the larger the noise voltage since by Ohm's Law, V=IR. Noise induced between the volume control and the amp is faced with the high impedance of the amp (47k) or the output impedance of the volume control, which can be low, but at typical listening levels can be moderate to high.
If that volume control was backed by a buffer, the output impedance would be always low, and the induced currents would drain off cleanly through the buffer output without inducing any noise. While it's true that the buffer itself produces noise (and, yes, distortion) it was clearly the case that the system noise was far lower with the buffer in the circuit than without.
The noise/distortion from the amplifier and source components should always dominate unless the system gain is too high and the volume control is used at near the minimum setting.
There is a widespread perception in diyaudio that fewer gain stages are better than more gain stages.
This is dumb.
In an analog system, the signal begins as vibration of a needle in a groove and ends with the vibration of a voice coil. The signal gain is something in the order of 10,000. For a given speaker efficiency and cartridge sensitivity, the gain is fixed. If you don't have a line preamp with gain, the gain must be added to the phono stage or the power amplifier. Bragging about the "purity" of your passive preamp setup is silly: all you've done is moved the gain to a different box!
There are a very good arguments meanwhile for not adding this gain before the volume control, and several other very good arguments for not bolting it onto the power amplifier. The traditional approach, splitting the system gain between a phono stage getting the signal to a nominal line level, an attenuator typically operating at -20~-40 dB, followed immediately by a line stage / driver with 6-12 dB, the interconnects, and finally the input-driver-output stages of the power amplifier is pretty much the optimum situation from a noise and distortion point of view. The critical pain point where the signal is weakest is the output of the volume control. From an impedance standpoint it is more susceptible to noise pickup than even the cartridge output! You really don't want to extend that weak signal over the length of the interconnect and through to the power amplifier.
****
Note : the problem with a passive preamp is not just noise pickup, or at least not as I described it above as some sort of coincidental interference. The amplifier's input stage - if its a discrete bipolar design like mine is anyway - tends to have a fairly large current noise which will also develop a noise voltage across Zeff. While not such a big deal for vacuum tube or jfet input stages, bipolars are a poor choice to pair up with a passive preamp.
Total Comments 14
Comments
-
A transformer (TVC) to me is the only sensible passive pre-.Posted 16th July 2016 at 07:19 AM by abraxalito
-
Yes, at least that way you harness the voltage usually thrown away in the attenuator and put it towards lowering the output impedance...Posted 16th July 2016 at 08:12 AM by rjm
-
Even the transformer is sub-optimal compared to the normal active buffer to drive the cables and input section of the amplifier. So while you are quite right in that the TVC is the best option if you are forced down that road, there are better ways to perform that function.
Even the name "passive preamplifier" is utter nonsense.Posted 18th July 2016 at 01:59 AM by anatech
-
Do you have any evidence/reasoning to support the notion that the trafo is sub-optimal compared to the buffer? Surely it would depend on the quality of the buffer, no?Posted 18th July 2016 at 10:49 PM by abraxalito
-
Sure. A buffer is flat in impedance and frequency response. The transformer type does matter in quality, but so does everything else. My experience has been that a TVC can be good, but a nice buffer stage can be absolutely brilliant! Interactions at higher frequencies can be difficult to control using a TVC, and depends greatly on the source impedance. A real buffer is high Z input and normally low Z output no matter the source impedance. The buffer is usually less expensive too. Gimmie another reason to go electronic - I'm there already!
-ChrisPosted 19th July 2016 at 10:01 PM by anatech
-
You have so far omitted considerations related to the need for a power supply for the buffer. How did you power your 'absolutely brilliant' buffer?Posted 19th July 2016 at 10:24 PM by abraxalito
-
Hi,
"A Rant on ..." is just the right phrase for this little treatise.
It just highlights possible flaws, thwarfs possible advantages, adds a failure at a decisive but rather inconspicuous point and is based on a single briefly observed circuit design, just after the fashion of īDonīt disturb my opinions with factsī or īI donīt care if my test procedure is unsuitable as long as it can support my prejudice.ī
So letīs see if thereīs some flesh to munch ...
· "sure enough the system noise increased ..."
- The omittance of the Buffer didīt change the gain structure of the system, hence it didnīt change the noise gain and the noise gain structure at all.
The only change been the cable connecting point to the power amp moving from ībehindī (or with) the buffer to ībeforeī (or without).
Seen from the VC the input impedance of the Buffer was only replaced by the input impedance of the power amp plus cable.
If we now assume a 1m long cable run as rather electrically or noise-wise negligible -or model it as bunch of stray parameters- and assume same impedances for the buffer, resp. the power amp input, than theresīs no reason to assume a higher noise value for the buffer-less mode.
In fact just the opposite applies as the bufferīs internal noise is omitted with.
· "Noise is usually induced as a current..."
- This is correct so far. As it is mainly related to the base/grid/gate-current of the input devices itīs obvious that bipolars with their much higher input currents suffer more from current noise than FETs or Tubes.
When in the discussed system the noise actually increased it could only be due to different input stage impedances and topologies of the buffer, resp the power amp, or radio interference due to indecent shielding of the cable.
In other words, if the VC īseesī a say 47k impedance it doesnīt care at all noise-wise if that impedance is directly at its output pin or at the end of a 1m cable run (it may give bandwidth issues ... but thatīs not the point here)
· "For a given speaker efficiency and cartridge sensitivity, the gain is fixed. If you don't have a line preamp with gain, the gain must be added to the phono stage or the power amplifier."
Though theoretically true it doesnīt apply so in praxis.
Ideally the gain structure is as such that You apply only so much gain as required.
Now allmost all modern sources supply for output voltage levels of ~6dB to 10dB more than needed to drive the power amplifier into clipping.
Hence attenuation of the signal is required already.
Due to increasing tolerances in Potis towards higher attenuation values, lower attenuation values would be preferrable.
rem: Probabely the only true reason for high priced Potis is their lower inter-channel tolerance at high attenuation values.
Adding another gain stage/Buffer only makes matters worse.
Itīd add fully unneccessary gain, noise and THD.
· "From an impedance standpoint it is more susceptible to noise pickup than even the cartridge output!"
- Correct, but thatīs the reason we use decent shielded cables ... to not pick up radio noise.
And what impedances are we talking about anyway?
The output impedance of a classical Poti varies between 0 and 1/4th of its nominal value - not 1/2 as falsely suggested in the Sketch. (47k-Pot --> 0R-11k75 ... which brings Zeff down to 9k4 instead of 15k6, already fulfilling the rule of thumb of > 1:5 sender-output-impedance : receiver-input-impedance ratio).
Almost all commercial preamps feature output resistors between ~47R and 470R to reduce capacitive sensitivity and/or to apply for muting etc. etc.
So, whats the big differnce anyway, compared to the 0R-2k5 of a more appropriate 10K-Poti??
- From a noise standpoint it is always best to put the highest gain to the front.
Only if radio interference is to be expected a bufferīs low output impedance may be helpful, but certainly not another gainstage.
A gain-stage only made sense if weīd like to drive the cable with higher voltage levels to increase the signal-to-(interference)noise-ratio on the cable-run itself.
But that required a dedicated attenuator at the power amps input or a power amp with correspondingly lower gain on the other hand.
Use decent cabling and You donīt have to cure the system with questionable measures in first place.
-- The only real reason that could speak against passive drive is impedance mismatch.
This could lead to bandwidth issues.
If the load impedance the VC sees is sufficiently high and(!) the upper bandwidth limit remains well above the audible range, than thereīs no reason -technically or acoustically- speaking against passive VC.
This claim asks for VCs with low nominal resistance value (10k or less for Potis).
The effect of varying output impedance of simple Potis -and as such varying bandwith and possibly too high output impedance can be omitted with VC-circuits featuring constant and lowish output impedance.
Iīve done this in my Preamp and it works perfectly (switched resistor network, bit-weighted, ~1k).
A optional cable-buffer can be switched unaudible into and out of the signal path on-the-fly via IR-control, allowing for reliable comparison.
Over more than 10 years of continuous use it showed that only in severe cases, like cable runs of considerably more than 10m (unbalanced!) an/or input impedances considerably lower than 5kOhm, that the buffer could make a positive difference.
With more conventional setups the passive mode was generally the preferred choice ... not because of better sound (only the soundstage seemed a tiny bit smaller with the buffer included), but simply to keep unneccessary parts/stages out of the signal path).
jauu
CalvinPosted 20th July 2016 at 08:03 AM by Calvin
-
I am just wondering how many beers that's after...
The diagram has been corrected to show the correct output impedance of the attenuator.
Also, I agree that there is no need to worry as long as the driving impedance is below 220 ohms, 47-100 ohms being pretty typical in line stages anyway. If you have a 10k or smaller pot., and never turn up the volume very high it's likely almost always within that safe range. The only downside you face then is the low input impedance.Posted 20th July 2016 at 12:45 PM by rjm
Updated 20th July 2016 at 10:07 PM by rjm -
If you turn the volume all the way up...?
It looks like another resistor in parallel with the input resistor on the "amp", no? This would have what sort of negative effect... and I'm a bit unclear how the input to a buffer looks much different than the input of a power amp, unless you are saying the interconnect is going to generate that noise, is that what's being said?
The problem with a "TVC" is that it is a transformer, and with that comes all the issues of using transformers... having looked at the waveforms coming off some of the highly regarded ones, I've not opted to go that way.Posted 25th July 2016 at 11:37 PM by bear
-
A 47k volume potentiometer at maximum is just a 47k resistance to ground. There is no series element, so the output impedance is defined by what the stage before it consists of (in parallel with 47k). It is essentially removed from the circuit - no negative influence - but at the same time it's function (adjusting the volume) is also defeated.Posted 26th July 2016 at 02:48 AM by rjm
-
so ur saying midrange of the theoretical pot is the most problematic for this noise issue...?Posted 26th July 2016 at 03:40 AM by bear
-
Absolutely. Worst case is when the wiper perfectly divides the resistance in half. This is usually around the 2-3 o'clock position, so it is rarely encountered in practice... but above 11-12 o'clock the impact can often be heard.Posted 26th July 2016 at 07:50 AM by rjm
-
ok, so moving the buffer next to the "pot" eliminates the cable and then drives the cable with a very low Z, removing the impedance that is generating the noise? But is the cable the offending element, or would it still generate noise if the "pot" were inside the amp, AT the input stage??
And, also, isn't the input to the buffer the same (in essence) high Z input??
Seems confusing to me.Posted 26th July 2016 at 01:07 PM by bear
-
The high impedance section is susceptible to noise pickup, so the longer and more exposed it is, the more likely it is to inject significant noise into the signal.
An amplifier input also has a noise current which will induce noise in proportion to the impedance seen by its inputs. This noise will not depend on the length of cable between the volume potentiometer and the amplifier.
Putting a lowZ buffer after the volume control eliminates both issues.
Yes, the buffer input now faces the volume control output. So if the current noise of the buffer is higher than the current noise of the amplifier it's true you are worse off in that respect even if the noise pickup from the cable is eliminated.
So obviously you design your buffer to have low current noise, which is not a difficult task given the low signal voltages involved.Posted 26th July 2016 at 11:39 PM by rjm
