Schiit KARA Preamp and Headphone Amp Review

[solderdude]

Except ... it isn't an unknown.

The Saga (+) is a tube buffer with 2.2uF output caps. These form a first order high pass filter. So does the Freya +.
These will affect the LF response with lower impedance loads and do so predictively and can be accurately calculated.

The Saga S is solid state (so was the Freya S) and has no output caps, nor does the Kara which is also solid state.
These pre-amps lack the output capacitor as it is not needed, thus do NOT roll-off the lows like the tube versions do with lower load impedances.
There is a DC servo in there but at a very low frequency which will cause a roll-off (independent of the load) below a few Hz.

The passive stages all are DC to MHz region and are purely resistive so will have an Ohmic output R and will not vary with frequency because there are no capacitors in the path.
There is no reason to measure it either in that aspect.

 

[SCG]

It would similarly effect all passive volume controls.

It is not the pre-amp itself that is the issue but the output resistance in combination with the cable capacitance that forms a low pass in passive mode.
For this to reach audible levels you would need certain conditions which could occur with passive volume controls. A 10k pot, driven by a low output R source would have an output R of 2.5k ohm.

Let's say that a really poor cable capacitance would be 500pF/m you would get a 5nF capacitance (10m). This will give you a -3dB point at 13kHz which would be really audible. Most cables, however, will be in the 100-150pF range, so around 1nF for 10m = 60kHz -3dB so at 20kHz -0.5dB or so.

Great info you posted in this thread to explain to people!

You used long cable runs and high capacitance cables to illustrate a point but it's amazing the difference 4nf can make. Point being no one should use a passive volume control with long run, high-capacitance cables with a lower input impedance amp.

You used a R output of 2.5k ohm max in Passive mode for your examples but for Schiit preamps with the resistor attenuation models wouldn't the max be 5k as stated below by TM and validated by MCR? Maybe 2.5k is typical (or only an example) and 5k is actual for Schiit, just want to make sure I'm following the thread correctly.

10K input impedance, 0-5K output impedance, worst case at midpoint (-40dB)

That's what I've seen called a "series attenuator", not a ladder. And your original reply is then correct. It's simpler than a ladder, requiring fewer resistors. The drawback is that max output impedance is twice as high as a ladder, 5k ohm instead of 2.5k ohm.

 

[MRC01]

... You used a R output of 2.5k ohm max in Passive mode for your examples but for Schiit preamps with the resistor attenuation models wouldn't the max be 5k as stated below by TM and validated by MCR?

Correct

Maybe 2.5k is typical (or only an example) ... and 5k it actual for Schiit,

2.5k would be the output impedance for a 10k ladder attenuator layout, but Schiit uses the series attenuator layout which doubles the output impedance for the same input impedance.

just want to make sure I'm following the thread correctly.

You are. The increase in output impedance from 2.5k to 5k increases the HF attenuation, or conversely, you need an amp with a higher impedance to avoid the attenuation.

If one is going to use a series attenuator layout, I'd recommend halving the impedance. 5k input impedance is still high enough to keep most solid state devices happy, and the lower 2.5k output impedance gives more flexibility in choosing the downstream power amp. It should also have slightly lower noise, too.

PS: to avoid impedance matching issues, 10:1 is a general rule of thumb. That is the attenuator's input impedance should be 10x higher than the upstream device's output impedance, and the attenuator's output impedance should be 10x lower than the downstream device's input impedance.

 

[solderdude]

Yes, I was kind of surprised by the 5k at -40dB but it is because of the construction to create a constant input R ladder:
It is explained quite simply here:

Logarithmic Attenuator Calculator

I calculated for a regular 10k potmeter and -6dB attenuation which would be worse case and be 2.5k.
If it is 5k then you could half the frequency response as calculated.
So a 'regular' or 'stepped attenuator' would be better in this particular aspect than a ladder.

If one were to measure the output R of a 10k attenuator which is connected to a low impedance source then between source and wiper there would be 5k and between wiper and ground there would be 5k. But for AC those two 5k resistors would be in parallel and thus be 2.5k and attenuation would be 0.5x (-6dB).

In the ladder this will be different due to a bunch of resistors being present.

This is only in passive mode though. When used as designed (with the buffer in place) the output R would be 75ohm in SE. One would expect 150ohm in balanced but it looks like some extra resistors are used in balanced outputs (2 x 300ohm it seems).
No idea why, maybe it is a standard (75ohm for SE is) and 600ohm is common in balanced circuits so they may have decided to go for that.
In any case that won't be differ with frequency (like the tube buffers do).

And yes, long cable runs using a passive attenuator, certainly 50k and 100k that seem to sell as well, is not recommended at all.
Short, not exotic cables (think 2 meters/6 feet) should not be an issue at all.

 

[MRC01]

Yes, I was kind of surprised by the 5k at -40dB but it is because of the construction to create a constant input R ladder:
It is explained quite simply here:

Logarithmic Attenuator Calculator

I calculated for a regular 10k potmeter and -6dB attenuation which would be worse case and be 2.5k.

That's correct for the parallel-series "ladder" layout. I've always known the term "ladder" to refer specifically to this layout. This Schiit uses a series attenuator which is a different layout having higher output impedance for the same input impedance. Calling it a "ladder" seems confusing.

This doc shows 3 different layouts: ladder, series, and shunt: http://diyaudio.co.kr/wwwboard1/data/board1/compare.pdf
According to that document and the links & conversation here, I believe the Schiit uses "series".

 

[solderdude]

I think Schiit uses a constant input resistance logarithmic resistor ladder.

Logarithmic resistor ladder - Wikipedia

en.wikipedia.org

The downside is that when the relays are switching over this takes a short time and can give short pulses (very short duration cracking noise) during volume adjustment.
A bit like a scratchy pot.

 

[SCG]

Thanks to both SD & MRC for the info!

I'm (mostly) using a balanced DAC with 75Ω output Z connected via balanced XLR to Kara's 10K Z input (133:1 ratio) and using the active buffer mode, and from there outputting via Kara's balanced XLR outputs (600Ω Z) into dual mono block amps each with 50KΩ input Z (83:1 ratio) and all with short low-capacitance cables, so I'm all good and happy with Kara's 117 SINAD in this mode. Damping factor, speaker wire length, and capacitance are all good too, but whole other subject.

Kara has the passive mode so I'm always interested in the pros and cons of using that mode and being able to explain it clearly and concisely to others. If newer people want some more empirical data on passive volume controls (which has been talked about alot in this thread) below is a link where someones tests a SA2X 10K ohm stepped attenuator, 4 gang switch with precision low noise resistors vs a NS-05P 50K ohm potentiometer, ALPS 4 gang.

His Post #4 of that thread is the most relevant to this passive conversation where he compares both units by measuring frequency response vs cable length and capacitance:

Passive Volume Control Measurements:

I think Schiit uses a constant input resistance logarithmic resistor ladder. The downside is that when the relays are switching over this takes a short time and can give short pulses (very short duration cracking noise) during volume adjustment. A bit like a scratchy pot.

My only unfavorite thing about the Kara, I posted a video earlier in the thread that gives some inidcation of what it sounds like.

 

[MRC01]

I think Schiit uses a constant input resistance logarithmic resistor ladder.

Logarithmic resistor ladder - Wikipedia

en.wikipedia.org

The downside is that when the relays are switching over this takes a short time and can give short pulses (very short duration cracking noise) during volume adjustment.
A bit like a scratchy pot.

Yeah, on third glance, that diagram does have the same layout as the traditional ladder attenuator, a voltage divider with 2 resistance values, one in series with the downstream load, one in parallel with it. But the Schiit design minimizes the number of resistors, which forces it to have a higher output impedance than the traditional ladder attentuators that I have built, using a separate pair of resistors for each volume setting.

That is, the traditional approach with a separate pair of resistors for each volume setting is as you and I both described earlier. But the layout Schiit uses, to minimize the resistor count, shares & combines resistors across the steps, which constrains you to have higher output impedances. It's optimized for simplicity and cost, not for output impedance.

IMO, the most important downside to this approach, the key take-away, is that the output impedance is twice as high as the traditional approach. This means you need to be careful about selecting a compatible downstream device. For a safe 10:1 ratio, a traditional 10k ladder attenuator has can drive a 25 kOhm input impedance device which covers most solid state amps on the market. Yet for the Schiit your amp should have a 50 kOhm input impedance, which narrows the field.

I'm (mostly) using a balanced DAC with 75Ω output Z connected via balanced XLR to Kara's 10K Z input (133:1 ratio) and using the active buffer mode, and from there outputting via Kara's balanced XLR outputs (600Ω Z) into dual mono block amps each with 50KΩ input Z (83:1 ratio) and all with short low-capacitance cables, so I'm all good and happy with Kara's 117 SINAD in this mode. Damping factor, speaker wire length, and capacitance are all good too, but whole other subject.

Kara has the passive mode so I'm always interested in the pros and cons of using that mode and being able to explain it clearly and concisely to others. ...

You're lucky and should be able to use Kara passive no problem. On the input side, as you noted, the Kara's 10k impedance is 133x your source output impedance. On the output side, your amps have 50 k input impedance, so you are 10x in the worse case scenario, which passes the rule (just barely). Use decent cables and you should have no attenuation in the audible spectrum.

You can see that if the Kara were 5k input / 2.5k output instead of 10k input / 5k output it would be easier to impedance match. You'd have 66:1 on the input side and 1:20 on the output side. It's unusual to see source devices with output impedances > 500 ohms, yet it is common to see amps with input impedances < 50 kOhm.

To explain it clearly and concisely to others: just follow the 10:1 rule as above. Every time upstream device A connects to downstream device B, B's input impedance should be at least 10 times A's output impedance. The trick with passive attenuators is that since it's in between 2 devices, to get 10x on both sides you need 100:1 total from source to target.

 

[SCG]

To explain it clearly and concisely to others: just follow the 10:1 rule as above. The trick with passive attenuators is that since it's in between 2 devices, to get 10x on both sides you need 100:1 total from source to target.

Thanks again, the min 10:1 rule is something I fully understand, though as you, SD, and others have effectively demonstrated with passive attenuators (min 100:1 both sides) how you get there can be variable, interdependent, and with reduced equipment options.

 

[MRC01]

Thanks again, the min 10:1 rule is something I fully understand, though as you, SD, and others have shown with passive attenuators (min 100:1 both sides) how you get there can be variable, interdependent, with reduced equipment options.

Yep. Yet really, most solid state devices give you that 100:1 easily, and you don't even need 100:1 because the attenuator's output impedance is lower than its input impedance. With a traditional ladder attenuator, worst-case highest output is 1/4 its input, so from source to target you only really need 1/4 of 100:1 or 25:1. With the Kara, worse-case is 1/2 the input, so from source to target you need 1/2 of 100:1 or 50:1.

Most solid state devices easily meet and exceed these requirements. Typically, output impedances of CD players and similar devices are under 200 ohms, and input impedances of amps are at least 20 kOhm. That's 100:1 right there. Kara or any other well designed passive works just fine if you pick the right impedance (typically 5k or 10k).

PS: now that I said that, everyone feel free to post all kinds of counter examples. Yes, they do exist.

 

[SCG]

Ok my understanding is if the preamp passive max Z output is 5kΩ then you still need a amp with 50kΩ Z input to avoid any potential issues at that level (10:1), and the preamp/input source Z ratio only comes into play when the passive preamp's min Z is low or near zero, when matching with the amp's Z ratio.

 

[MRC01]

Ok my understanding is if the preamp passive max Z output is 5kΩ then you still need a amp with 50kΩ Z input to avoid any potential issues at that level (10:1)

True.

, and the preamp/input source Z ratio only comes into play when the passive preamp's min Z is low or near zero, when matching with the amp's Z ratio.

Not sure what you are asking here. The passive input impedance is the same 10 kOhm at every setting - it does not vary like the output impedance does.

Match the passive input impedance to your upstream sources, and match its output impedance against your downstream amp.

 

[solderdude]

Ok my understanding is if the preamp passive max Z output is 5kΩ then you still need a amp with 50kΩ Z input to avoid any potential issues at that level (10:1), and the preamp/input source Z ratio only comes into play when the passive preamp's min Z is low or near zero, when matching with the amp's Z ratio.

No, you can even use it with an amp presenting a 10k load.
That load, however, should be resistive which it very likely is.
The ONLY thing that happens (when the output R is 5k, in passive mode) is that the attenuation is not -40dB but -43.5dB (for amp with 10k input) which really does not matter at all because volume is set by ear and not by dB numbers.
In case of an unbuffered R2R ladder DAC the 1:10 rule would certainly apply.

The 1:10 rule is sometimes valid for certain dynamic headphones. Also some sources with under-dimensioned coupling caps (tube circuits mostly) will best be loaded with at least a 10x higher Z. Another example could be a device that has a complex input (transformers) should be driven properly.

So the 1:10th rule exists but rarely is of importance, only with certain specific gear (tube circuits or very old gear or some really exclusive and weird designs).

All passive pre's (especially the ones such as LDR and transformer based) should best not be used with long cables or into 'complex' inputs.
In all other circumstances it can be used. Even in low resistance (resistive) loads. In the latter case attenuation will just be a little higher than when a buffer would be used.

 

[Noske]

This is off the topic of the product being reviewed.

After about eleven pages and a couple hundred comments, I managed to figure out there are fundamental differences between a solid state and a valve/tube preamp. That is perhaps a source of my confusion.

So, valve/tube preamps versus solid state appears to be significant issue. Is it as simple as that?

What of a valve/tube preamp that is described as two 6SN7 driving output transformers (no capacitors in the signal path). I suspect that may be an unusual topology, one which I can find no information as to ohm/frequency.

I have seen the term transformer coupled somewhere, but specific information is elusive.

I'm fishing, of course, so mods please delete this comment ought it be out of bounds. I am learning a steep curve.

edit - in hindsight, I could have perhaps posted this as a general question in an new appropriate thread, but oops, the damage is done. My apologies.

 

[solderdude]

So, valve/tube preamps versus solid state appears to be significant issue. Is it as simple as that?

Not really, one can design tube pre-amps which can be loaded with lower impedance loads.
The Saga and Freya just aren't amongst those because of the relatively small output capacitors they use.
The load impedance for these amps should be above 5kohm when one does not want to have an audible roll-off in the subbass range when the buffer is engaged.
In passive mode this is not an issue, cable capacitance (long cables) will be the limiting factor there... but not for the tube buffer.

in hindsight, I could have perhaps posted this as a general question in an new appropriate thread, but oops, the damage is done. My apologies.

It is a good thing that differences between Schiit tube buffers/pre-amps have been discussed in this thread and that the output resistance came up for discussion.

This is only clear for people with (audio) electronics knowledge so a good thing this has been discussed.

We all are hear to learn and have conversations about audio products in the broadest sense and most of it was about Kara and differences between other products as well as about technical aspects of the Kara.

 

[MRC01]

No, you can even use it with an amp presenting a 10k load.
That load, however, should be resistive which it very likely is.
The ONLY thing that happens (when the output R is 5k, in passive mode) is that the attenuation is not -40dB but -43.5dB (for amp with 10k input) which really does not matter at all because volume is set by ear and not by dB numbers.
In case of an unbuffered R2R ladder DAC the 1:10 rule would certainly apply.
...

This is an interesting point. So most solid state amp input impedances are purely resistive? I did not know that.

...
It is a good thing that differences between Schiit tube buffers/pre-amps have been discussed in this thread and that the output resistance came up for discussion.
This is only clear for people with (audio) electronics knowledge so a good thing this has been discussed.
We all are hear to learn and have conversations about audio products in the broadest sense and most of it was about Kara and differences between other products as well as about technical aspects of the Kara.

Indeed.

 

[solderdude]

So most solid state amp input impedances are purely resistive? I did not know that.

Most are resistive but do have a small capacitance to ground (often after an input resistor).

 

[MRC01]

Most are resistive but do have a small capacitance to ground (often after an input resistor).

Even if the downstream device (power amp) is a purely resistive load, don't we still have an RC circuit? The capacitance comes from the cables and the resistance is the output impedance. For example at 5 kOhm output impedance and and 6' cable with 20 pF per foot (120 pF total) we have a bandwidth of 1 / (2 * pi * R * C) = 265 kHz.

That's not a problem but if you had longer higher capacitance cables it could be. For example I've seen some cables having 150 pF per foot, and if you used 3 meters / 10 feet you'd have -3 dB @ 21 kHz which attenuates in the passband. In this example, if the output impedance was 2500 Ohm the bandwidth would be 42 kHz so no passband attenuation.

Understood, that is an extreme example just to illustrate a point.

 

[solderdude]

Yes, for all passive attenuators and also for source that have an unusually high output Z sources this is a potential reason for concern.
A high capacitance cable could lead to audible high frequency roll-off depending on conditions.
Long or unusually high capacitive interlinks will limit the upper frequency response at certain attenuation levels more than at other attenuation levels so may only become somewhat of an issue at specific attenuation levels.

Lengths up to 6 to 10 feet with normal interlinks gives no audible issues at all.
One can even use longer runs but would need to ensure they are low capacitance.

The rule is simple... with longer cable runs > 10 feet (3m) it is best to use the active out as it is more than excellent.
In case of the Kara I would not even bother to use the passive option anyway.

 

[nagster]

7-5 o'clock rotation with hard stops, volume also changeable via remote, no infinite rotation encoder.

thank you.
it's good. I don't like the infinite rotary encoder for volume control. (despite owning several encoder amplifiers)

My request for the future is to display the volume step value.
It may not be necessary if you are just enjoying music, but it is very convenient for verification work to be able to accurately reproduce the gain without a measuring device.