Schiit KARA Preamp and Headphone Amp Review

[T.M.Noble]

Linear in terms of dB, which is a logarithmic measure (see formulas). Let me try another explanation: Consider, for example, a fixed angular change of the pot, say by 10 degrees out of the total possible 270 degrees. As you turn a truly linear pot, for every such angular change the signal changes by the same constant amount (in this case 1/27th of the maximum). But as you turn a truly exponential pot (incorrectly called "logarithmic" in audio), for every such angular change the ratio of the next signal to the previous signal is constant. You would only perceive the latter as a linear change in volume.

Got it. Always happy to learn more about this stuff. Glad I have access to people who can explain the nuances.

 

[MRC01]

Often, these relay actuated volume controls are controlled by an analog pot connected to a sensor that detects the resistance and triggers the relays. In this case, the analog pot used to control it can be linear, while the relays it triggers have logarithmic steps. Whether the pot is linear or log is just an implementation detail.

Put differently, the net effect of the volume control can be logarithmic, same dB change for each click, even if the pot used to control it is linear.

 

[solderdude]

Pretty sure this it what I said if you forgive the spelling mistake. The pot on Kara is linear, not an audio taper.

There may well be a pot (single track) that may well be linear but that probably drives an ADC input with a voltage which drives the relays and those steps are in dB so log type in attenuation and not linear even though the used pot could be.

 

[Stoutblock]

Do we know if these relays are setup in a ladder or shunt circuit?

 

[MRC01]

Do we know if these relays are setup in a ladder or shunt circuit?

I don't think it can be a ladder since @T.M.Noble said the input impedance is always 10 kOhm while output varies from 0 to 5 kOhm, with max 5 kOhm at -40 dB. If it were a 10 kOhm ladder, it should be max 2.5 kOhm at -6 dB.

 

[T.M.Noble]

Please be a pedant, is the Kara's RCA channel distortion difference of 7.8db within Schiit's spec tolerance? The 9.1db RCA channel distortion difference on the Freya S was considered out of spec and a 2nd unit was sent for retesting.

For reference, posting regarding Freya S 9.1db difference:



If within tolerance, where between 7.8db and 9.1db (within that 1.3db window) is the crossover point. Many Thanks in advance!

Looks like Amir got a super matched pair of JFETs on one channel. Attached is what you would normally see from Kara.

Nexus requires matched JFETs. On one of the channels they were super matched and showed a bit higher performance there. There will be always a bit a variability here. A quirk of NEXUS.

Hope that helps.

 

[Stoutblock]

I don't think it can be a ladder since @T.M.Noble said the input impedance is always 10 kOhm while output varies from 0 to 5 kOhm, with max 5 kOhm at -40 dB. If it were a 10 kOhm ladder, it should be max 2.5 kOhm at -6 dB.

The comments about the impedance variation made me assume it was a shunt circuit. Looking inside it does not look like nearly enough relays for 128 step ladder.

 

[MRC01]

The comments about the impedance variation made me assume it was a shunt circuit. Looking inside it does not look like nearly enough relays for 128 step ladder.

Especially not for a balanced output (4 channels instead of 2)!
But if it were a shunt, the input impedance wouldn't be constant.
Constant input impedance with variable output impedance, suggests a ladder. But 5 kOhm at -40 dB doesn't sound like any ladder I know of. But then I am no expert, I've built attenuators but only simple ones.

 

[T.M.Noble]

The comments about the impedance variation made me assume it was a shunt circuit. Looking inside it does not look like nearly enough relays for 128 step ladder.

My response was just parroting Jason's response to the question posed. If there is something missing or context that was missing I know there are ways you can reach him and he could get you an answer. However, I do know it is a 128 step ladder.

 

[MRC01]

My response was just parroting Jason's response to the question posed. If there is something missing or context that was missing I know there are ways you can reach him and he could get you an answer. However, I do know it is a 128 step ladder.

OK so if it's a ladder, then constant 10 kOhm input impedance makes sense, as does variable output impedance. Though the output impedance should max out at 2.5 kOhm at -6 dB. That's where the series & parallel resistors are equal value, 5 kOhm each, so the voltage is cut in half (-6 dB), and output impedance is the two 5 k resistors in parallel, which is 2.5 kOhm.

Which means impedance matching would be easier, since the amp downstream from the Kara only needs 25 kOhm or higher input impedance to give you a safe 10:1 ratio.

 

[Salt]

This gear would be the perfect one for my needs with 2 XLR inputs for phono and streamer and high output for purify amp, if only the HP section would made fit my low impedance Ananda ....

 

[solderdude]

The comments about the impedance variation made me assume it was a shunt circuit. Looking inside it does not look like nearly enough relays for 128 step ladder.

7 relays = 128 step

Relay Resistor Switch Networks for Audio Volume Control Operation and StuFF.

We are often seeing and commenting on relay controlled volume controls and some peeps including myself have no idea how they operate so here's a thread on them. Here is a audio volume relay attenuator with IR control for more reference. This circuit is similar to what is implemented in...

www.audiosciencereview.com

Similar to Freya

 

[lc6]

Often, these relay actuated volume controls are controlled by an analog pot connected to a sensor that detects the resistance and triggers the relays. In this case, the analog pot used to control it can be linear, while the relays it triggers have logarithmic steps. Whether the pot is linear or log is just an implementation detail.

Put differently, the net effect of the volume control can be logarithmic, same dB change for each click, even if the pot used to control it is linear.

The volume control does not need to be an actual pot. It could simply be a rotary pulse generator: Turn it clockwise and it generates pulse(s) incrementing a 7-bit counter (with a limit of 0..127) that drives the relays that control a discrete resistive attenuator. Turn it counterclockwise, and the opposite happens. Much easier to implement, more reliable (no gradual wear) and accurate (a "linear" pot may not be very linear) than having an actual resistive pot read by an ADC.

To make this whole discussion more fact-based, it would be helpful to see @amirm 's graph of the dB volume linearity and channel mismatch of this DUT.

 

[MRC01]

7 relays = 128 step ... Similar to Freya ...

So a series ladder instead of a parallel ladder? I thought ladder implied parallel. But in either case, not a shunt. A series attenuator looks simpler and cheaper, but some volume positions have more resistors than others, and wouldn't the resistor noise and tolerances be additive?

 

[T.M.Noble]

OK so if it's a ladder, then constant 10 kOhm input impedance makes sense, as does variable output impedance. Though the output impedance should max out at 2.5 kOhm at -6 dB. That's where the series & parallel resistors are equal value, 5 kOhm each, so the voltage is cut in half (-6 dB), and output impedance is the two 5 k resistors in parallel, which is 2.5 kOhm.

Which means impedance matching would be easier, since the amp downstream from the Kara only needs 25 kOhm or higher input impedance to give you a safe 10:1 ratio.

Jason forwarded these for more clarification:

Logarithmic resistor ladder - Wikipedia

en.wikipedia.org

Logarithmic Attenuator Calculator

I have some reading to do.

 

[solderdude]

The volume control does not need to be an actual pot. It could simply be a rotary pulse generator: Turn it clockwise and it generates pulse(s) incrementing a 7-bit counter (with a limit of 0..127) that drives the relays that control a discrete resistive attenuator. Turn it counterclockwise, and the opposite happens. Much easier to implement, more reliable (no gradual wear) and accurate (a "linear" pot may not be very linear) than having an actual resistive pot read by an ADC.

To make this whole discussion more fact-based, it would be helpful to see @amirm 's graph of the dB volume linearity and channel mismatch of this DUT.

As this is a volume control (with motor) and 270 degrees rotation it is most likely a linear pot. Linear pots this size can be of good quality and only needs to be one track.
The advantage is that on switch-on the volume level is known (its a voltage).

You can use a rotary encoder and do the same thing but would be optic and at least 7 bits. A simple up down would require remembering the last setting and would require a memory. The volpot would be much easier to implement. The onboard chip for the control is very likely to have a few analog inputs as well. 8 bit at least.

Given that T Noble said a linear pot, the thing is 270 degrees and motorized (and no display showing the attenuation) makes it a safe bet it is a volpot.

 

[respice finem]

This is a review, listening tests and detailed measurements of the Schiit KARA balanced preamplifier and headphone amp with stepped relay control. It was sent to me by the company and costs US $699.
View attachment 313439
Oh wait... that is the picture of our dog, Kara! Here is the Schiit version:
View attachment 313440

Took me a day to decode the subliminal message: Amir is dog tired testing all this stuff
Nice dog BTW.

 

[CleanSound]

Took me a day to decode the subliminal message: Amir is dog tired testing all this stuff
Nice dog BTW.

I thought the dog was a new rating trumping the golfing panther!

 

[respice finem]

But then he would need a white dog, a black dog, a red dog... and for Class A amps, a hot dog

 

[SCG]

Looks like Amir got a super matched pair of JFETs on one channel. Attached is what you would normally see from Kara. Hope that helps.

Got it, it's not that channel one was low and out of spec, it's that channel two was higher than normal. Was wondering if that might be the case but after reviewing the 81 page APx555 report on Kara didn't see any RCA to RCA measurements for a baseline so asked the question.

https://www.schiit.com/public/upload/PDF/APx555%20Measurements--Schiit%20Kara.pdf

Thanks for the quick response.