NAIM use Tantalums for coupling caps in their IXO Active Crossover - good or bad?

I recently won an IXO crossover and decided to check on what I thought was a loose Din socket.
On opening it, I noticed that :-

  • The 18db/oct filters used Zetex (formerly Ferranti) ZTX transistors - not an IC in sight (a refreshing change not to see NE5523/34 opamps)
  • The LF & HF 50k level controls are wired unusually - wiper to the input sockets with a 4k7 series resistor to prevent a dead short of the preamp output on minimum. (20k on max)
  • The single rail power supply uses a LM317 set to 24v - one for each channel.
  • The input, output and interstage coupling capacitors are Tantalum. 10uf/35v

Now I don't remember ever seeing Tants used for this purpose. Usually a standard polarised electrolytic would be used on single rail designs.

I have to say the crossover has made a significant improvement to my 35yr old SBLs. (Quad 306 driving the bass/mids and derated Krell KSA50 clone for the Scanspeak tweeters.)
I found myself listening for 4hours yesterday evening, about twice as long as I usually listen so I am not planning any "upgrades" to the IXO.
 

Attachments

  • ixoinside.png
    ixoinside.png
    149.3 KB · Views: 296
As far as I recall, in Cyril Bateman's measurements, they distort a bit more than aluminium electrolytics.
https://linearaudio.nl/cyril-batemans-capacitor-sound-articles

They can handle less reverse voltage, and if I'm not mistaken, they have less effective series resistance (hence less thermal noise) and less leakage than aluminium electrolytics. The distortion can still be negligible if the signal voltage across them is kept small enough.

I used to avoid tantalum like the plague because of it being a conflict mineral. That's less of an issue now, as nowadays there are many manufacturers stating that they did their best to obtain the tantalum from legitimate sources.
 
Disabled Account
Joined 2002
Problem with tantalum caps in audio is that they are unknown to many and still suffer from their tainted past. The drop type is exactly the type that caused tantalum caps bad reputation due to their unreliable character and unreliable sources in war torn countries. The Bateman results are of drop type or "bead" type tantalum caps. Recent tantalum polymer caps are of very high quality also in audio (but not popular as they are simply very expensive). Pro/medical/military equipment is full of them as today they are excellent* parts.

Replacing old drop types is advisable certainly when they are used directly on power rails. They have a habit of burning/exploding during/after a useful life. If you want to avoid that risk then only replace for tantalum polymer or niobium types by the known brands. Avoid the drop type, there exist "normal" versions.

*Open your smart phone and see what is used there.
 
Last edited:
  • Like
Reactions: 1 user
The 18db/oct filters used Zetex (formerly Ferranti) ZTX transistors - not an IC in sight (a refreshing change not to see NE5523/34 opamps)
Not at all, 5532 and 5534A's are a fine engineering choice, cheap and very performant, trying to get the same performace from discrete rather than opamps is indicative of failing to practice rational design, why make extra work for yourself to arrive at a weaker design? Would you design a steam-powered hedge trimmer for instance? Same logic involved.

Data point - I build a circuit with 6 NE5532 opamps in cascade, distortion way low even after passing through 12 stages, which is more than needed for any active crossover, and about £3 for the lot (in bulk pricing a few years ago). Totally the way to go.
 
  • Like
Reactions: 4 users
Disabled Account
Joined 2002
Batteryman seems to have a device without NE5534 so any debate on in his device non existing NE5534 seems useless :) Discrete designs require more skills than designing with opamps and obviously Naim went that route in a then already dominantly opamp world with a particular (in their eyes) logic reason. Naim made/makes more often unusual choices that make them a unique brand in a way -> DIN/BNC connectors, single supply designs, discrete designs, overdimensioned low noise PSUs, tantalum caps, made in England.
 
Last edited:
Disabled Account
Joined 2002
Todays polymer tantalum caps against yesterdays (think many decades) drop type tantalum caps may show an unexpected positive picture if one is prepared to go out of ones comfort zone. Adoring is a doubtful way of expressing in technical matters, we are not 13 year old girls adoring Justin Bieber.

BTW Mister Bieber made some albums like Justice that are quite OK so no offense.
 
Last edited:
  • Like
Reactions: 1 user
Most of what has been said above echoes what I have been able to find online.

I have concluded that:-

The rating should be well above the design voltage - so in the IXO, 35v Tants with 12v across them.
Limited to low signal levels i.e line level - about 1v RMS max.
Expect possibly slightly higher distortion, but lower noise than electrolytics.
Very little aging.
Low leakage.

One assumes Naim spent a considerable amount of time trying various types and brands of coupling caps (as well as the filter ones which look like polystyrene.)
 
Disabled Account
Joined 2002
With the difference (again) that todays polymer tantalums don't have inherited the negative aspects of their ancestors. Every repair guy/girl has replaced burnt out drop type tantalum caps and loves to hate them but apparently manufacturers like Kemet and Kyocera AVX instructed their employees to develop better products in the last 40 years or so.

Expect better properties, not worse. Try out, measure, verify, listen, write µF and Volt. There will be a different sound character though as the old ones had eh... quite specific shortcomings.
 
Last edited:
Not at all, 5532 and 5534A's are a fine engineering choice, cheap and very performant, trying to get the same performace from discrete rather than opamps is indicative of failing to practice rational design, why make extra work for yourself to arrive at a weaker design? Would you design a steam-powered hedge trimmer for instance? Same logic involved.

Data point - I build a circuit with 6 NE5532 opamps in cascade, distortion way low even after passing through 12 stages, which is more than needed for any active crossover, and about £3 for the lot (in bulk pricing a few years ago). Totally the way to go.
The cost is immaterial to Naim (& myself) and maybe Naim did try an opamp version and decided an all-discrete design sounded better.

Sometimes there is no need for opamps when the design is simple. Maybe we should be using the humble transistor more often.
 
As has been said Naim have been using tantalums as coupling caps ever since they began and due to their design philosophy there are a lot of them in the
signal path. Years ago when I had a small business modfying Naim pre amps I used to change them for Philips solid aluminium ones type 156 I think. I saw
improvements in distortion and the noise floor with this change and I thought they sounded a lot better. However a few customers said " I dont like it, its not the
Naim sound anymore ". I later found that the tantalums used were a bit microphonic, tapping them lightly with a pencil produced a signal on the scope and a dull
thud out of the system. Covering them with blutack damped this with similar sonic results.

I concluded that the use of tantalums by Naim contributes to their particular sound signature which some customers (quite a lot actually) like.
 
  • Like
Reactions: 2 users
Audio is chockfull of cliches going up in scale equally with age. You and I are both guilty too :D


I accept full responsibility :) But where is the cliche in this case? The tants in this device are the exact same types you described as having a "tainted past". Nothing super dooper polymer about them. It's a miracle they are not leaking right there on the floor. :p
 
Not at all, 5532 and 5534A's are a fine engineering choice, cheap and very performant, trying to get the same performace from discrete rather than opamps is indicative of failing to practice rational design, why make extra work for yourself to arrive at a weaker design?

Perhaps from Naim's perspective 'performance' is related to noise rather than distortion? I don't know the details of the design but ISTM that a discrete design may well be capable of lower noise, at the expense of higher distortion than a 5532-based design.
 
  • Like
Reactions: 1 user