Hifisonix kx-Amplifier

www.hifisonix.com
Joined 2003
Paid Member
hifisonix kx-Amplifier: Discontinued

If you have purchased PCB’s and are still constructing your kx-Amplifier, I will continue to fully support your efforts as required, so do not worry.

114 PCB sets of the kx-Amplifier were sold, with as far as can be ascertained, a large proportion of them being built. Unfortunately 6 of the builders reported oscillation problems (20-80mV at 15-20 MHz). This was fixed via the December 2020 Update given below. Additionally, the offset adjustment was somewhat interactive between the two channels.

This necessitated a lot of hand-holding and support from my side to the builders, and quite some frustration on their side for which I apologize.

The decision has therefore been taken to pull the kx-Amplifier and Jim’s Audio will no longer stock the PCB’s.

In place of the kx-Amplifier, a new revised design will be launched in the next two weeks that addresses the kx-Amplifier shortcomings and offers dramatic hum/noise reductions and distortion improvements on the already low values featured on the kx-Amplifier.

The new amplifier that replaces the older design from 2018 is the

kx2 Amplifier

The kx2 PCB's will be available in about 2 weeks from Jim's Audio. I will open a new thread shortly.

______________________________________________________________________________________________________________________________________


December 2020 Update



As of December 2020, 107 kx-Amplifier PCB sets have been sold by Jim’s Audio. 6 builders have reported 15-20 MHz oscillation at between ~1.5mV and 80 mV. Since the oscillation is at very high frequency, it is not loop related, but more likely parasitic oscillation (loop related problems are usually <5 MHz, and normally below 2 MHz even in a CFA amp like this).
Builders reported that touching Q3 or Q4 collector stopped the oscillation, or, placing a scope probe on the collectors also seemed to stop the problem whilst monitoring the output.
The probable cause of this issue appears to be parasitic oscillation in Q2 and Q1. These are emitter followers (called ‘beta helpers’ when used in an amplifier in this way) and have a tendency to oscillate at HF due to the formation of Colpitts oscillator structures arising from the parasitic L and C components around the devices. Modelling showed it was easy to get oscillation in the 30-40 MHz range with the probable parasitic L and C on the PCB – not as low as that reported, but a sure indication that there was potential for problems.
If you can see this oscillation with a scope, or you can hear a clearly discernible hiss (‘Shhh’) from your amp with the inputs unplugged then do the following:- (note: a well-built kx-Amp must be absolutely silent on 90 dB sensitivity speaker with the inputs unplugged on both hum and hiss)
1. Change R4 and R5 from 10k to 1k (this will BTW improve the square wave response as well)*
2. Raise Q1 and Q2 up off the board and insert a 1k resistor in series with the collector (see the post #479 – thanks to Polsol for the technique)
3. Across each of the 1000uF capacitors under the board, solder an SMD 1206 1uF 50V X7R decoupling capacitor (example part from Mouser 187-CL31B105KBHNFNE)
4. As a general improvement with signal sources with 50 Ohm impedance (most CD players and modern preamps), replace R31 with a 1.8k resistor. This reduces the slight overshoot which is quite normal for a TPC compensated amplifier (See Cordell ‘Designing Power Amplifiers’ 1st Edition pages 177-180)
*if you leave the R4 and R5 10k resistors in situ, you must instead place a 68 Ohm in series with 330pF capacitor network across R4 and across R5. This has the same effect as the 1k resistor in that it dramatically lowers the impedance at the emitter of Q1 and Q2 at HF, suppressing the oscillation.

You can dramatically improve the LF distortion on the kx (and sx amplifiers) by adding a 220uF 4V capacitor to the bias controller circuit. Details in this post here: Hifisonix kx-Amplifier (PDF in post #758)

My thanks to diyAudio member Rallyfinnen for identifying this issue, and for the patience of the affected builders.

Commercial disclaimer: I make no money from any of the PCB's sold via Jim's Audio. I have an arrangement with him to supply the PCB's I design for interested DIY builders.
 

Attachments

  • IMG_5492 (2).jpg
    IMG_5492 (2).jpg
    323 KB · Views: 4,756
Last edited:
KEF LS50

http://hifisonix.com/wordpress/wp-content/uploads/2018/10/kx-Amplifier-Oct-2018.pdf

page 20

Most of my assessment has been done with a pair of Kef LS50’s and a B&W sub-bass to augment the LS50’s low end (down by 3 dB at 79 Hz), which in an small listening environment, make for a wonderfully immersive experience. The kx-amplifier is very smooth and euphonic, but above all, its also a very accurate amplifier. On B&W 703’s (nice easy load to drive), the kx-Amplifier has a wonderfully relaxed sound with plenty of beautifully articulated bass...
page 34

I listened to a variety of classical and jazz music (vinyl and CD) on my B&W 703’s and my Kef LS50’s – Ella Fitzgerald, Leonard Bernstein’s George Gershwin recordings, Steve Gadd, Fourplay and the Berlin Philharmonic’s Swan Lake and Sleeping Beauty in amongst others. The kx-Amplifier retains the warmth and ‘liquidity’ of the sx-Amplifier, but my impressions are that the top end is cleaner (cymbals, strings) and better bass performance – its almost as if it was slightly more boosted than the sx-Amp and the notes better articulated. The mid range to my ears is cleaner – and I think this has to do with the dramatically lower (measured) distortion. On 4 ohms and higher at full power, the kx-Amp produces no more than about 0.01% distortion, which is about 60x lower than the sx-Amplifier and it will happily drive a 2 Ohm load - obviously in class AB mode - with commendably low distortion.

Imaging remains outstanding, and particularly enjoyable on the KEF’s which really excel in this department. The left to right and sound stage depth are the best out of all the amplifiers I use. This characteristic seems to be a particular feature of class A amps – I have no explanation as to why – it’s a matter that needs some investigation in my view. I used a B&W sub-bass with the KEF’s on some material but as I remarked elsewhere, this was simply to augment the lower end of these speakers which drop off rapidly below about 70 Hz. On the 703’s the bass is superb and extends very low with fabulous articulation.

Used with the appropriate speakers (either 90 dB+ efficiency, or if lower, in a small listening space), the kx-Amplifier delivers an outstanding listening experience. The classic class A smoothness, lack of any glare and the effortless mid-range are available in bucketloads. The bass performance in my view is a significant improvement as well and I put that down to the more capable EF2 output stage.
KEF LS50 has a true sensitivity of only 83 dB. Minimum impedance of 4 Ohms.

How does orchestral music sound with them -without sub? Presumably they require more peak power. Many of my musics have a DR15.
 
Last edited:
[Polish] KEF LS50 - Zespoły głośnikowe | Testy w Audio.com.pl

to English: Google Translate

Almost 4 Ohms.

The jubilee product obliges ... But apparently not necessarily to reliable information. The nominal impedance, according to the bad practice of large British companies (a large slow more? British is allowed more?) Is presented as 8 ohms, here with quite an insolent note that the minimum is 3.3 ohms ... Maybe I am imprudent, writing such things? Well, we are a guest.

21988-max_kef_ls_lab.jpg


And we still like ... On the characteristics measured in our laboratory, I do not see such a low minimum, slightly lower than 4 ohms we are falling at the very bottom of the band, but I would rather take into account the minimum at 200 Hz, there it has a value of 4 ohms, and hence the "nominal impedance" should be taken into account.


21989-max_kef_ls_lab.jpg
The sensitivity is not high, but decent - 83 dB, which despite appearances is a standard result for a small substitute, although usually the value of this parameter overstates, even considerably. KEF gives 85 dB, and here I will not be defiled, because even though 2 dB is quite a bit, on the other hand, differences between catalog and measured data are usually larger...
 
Last edited:
Here is the newest Hifisonix class A amplifier.....

Andrew!

Welcome back! It's great to see you here again and hopefully more builders will try the Kx amplifier. I was saddened when you had left!

I have got all the parts, boards, etc...for mine and I just finished rebuilding/rewiring my room so the Kx should get into the build log hopefully by Spring 2019.

Your measurements on the Hifisonix web page look promising, and I hope my corrections weren't too forthcoming.

Thanks again for your contributions and particularly for the Ripple Eater PS which can be used on many other Class A and Class AB solid state designs as well. It saves a bunch space compared to having inductors and large caps! Much lighter chassis' are always a big plus!

Kind regards,
Anand.
 
www.hifisonix.com
Joined 2003
Paid Member
Maty
My sensitivity figures for the LS50’s were taken from the KEF specification. Indeed, they may well be much lower - you often see big discrepancies between manufacturers and independent tests. Nevertheless, the kx-Amp drives the LS50’s very well. They sound good on orchestral music.
 
Last edited:
Class AAB -> 50 watts at 8 Ohms, 100 watts at 4 Ohms

My questions is with kx-amplifier working only at class A.

I have a KEF Q100 (true sensitivity almost 85 dB, minimum 4.7 Ohms) cheap tweaked by me and really I LOVE the sound (after my last tweak -more mass- they sound much better than new!!!).

Your amplifier can works like class AAB -> 50 watts at 8 Ohms, 100 watts at 4 Ohms.

page 13

About the kx-Amplifier ‘mode selection’ option and Heatsink requirements

J8 in the circuit is a 2 pin link that increases the standing current from ~300 mA per output pair to around 600 mA per pair when closed. This allows the amplifier to be switched from class A to class AAB mode, retaining some of the characteristics of the class A sound but at much reduced power consumption.

Further, in the class AAB mode, you can run the kx-Amplifier at +- 35V rails which will provide 50W RMS into 8 Ωs, and close to 100W into 4 But do note, with the higher supply rails you will still need substantial heatsinks even in class AAB mode, where the standing dissipation will be around 50 Watts.
The kx-Amplifier requires a heatsink of at least 0.3 ̊C/Watt or lower per channel. Each channel heatsink will therefore weigh a minimum of 1.5 kG

So, either run on the +- 22-26V and have the option to switch between the two modes, or link out J8 permanently and run the amplifier on +- 35V rails in class AAB mode.

An externally hosted image should be here but it was not working when we last tested it.
 
Last edited:
It’s 50 W per channel with the original sx amp supplies - about 60 watts on the slightly higher rails for class A

page 21
Output power: >28 W peak class A into 8 Ωs = 15 W RMS into 8 Ωs class A; 50 W RMS into 4 Ωs class AB; 95W RMS class AB into 2 Ωs;

Standing current Class A mode: 1.2 amps (600 mA per output pair); standing current class AAB mode: 600 mA (300 mA per pair) +-20%

Distortion (measured): typically 0.006% at 15W into 8 Ωs; 0.015% at 50 W class AB into 4 Ωs; <0.1% at 90W into 2 Ωs

Slew Rate: 140 V/us (10%~90%, symmetrical)
If the standing current is the half in class AAB, the dissipated power should not be 30 watts by channel?
 
Last edited:
Founder of XSA-Labs
Joined 2012
Paid Member
Welcome back Bonsai! Nice amp design, and another one for us to try out for Class A designs. I am curious how this compares to Hugh’s Alpha 20 (or 35 Omega or 55w Alpha BigBoy). Also interested to find you listen to LS50’s as I had a chance to audition their self powered and DSP “w” variant the other day when I was in Germany. They sounded very nice for 5in bookshelf speakers.
 
www.hifisonix.com
Joined 2003
Paid Member
Maty, the kx-Amp delivers 50W into 4 ohms quite easily. Also, remember the peak class A power is about 25W (same for sx-Amp BTW). So, there is plenty of power to handle crescendo's. As mentioned on the website, of you want rock concert hall levels from a class A amp like this in a large listening space, you need to use more efficient speakers, or you need to constrain your listening space appropriately. If you shop around, you can find plenty of speakers at 93dB~96dB or more efficiency. I bought my LS50's to do an audio show after auditioning them - they were driven by a 100W amp on that show. But, they sound good even with a low power amp like this one.
 
Member
Joined 2011
Paid Member
I wish the square wave tests (page 25) weren't presented with oscilloscope bandwidth limiting turned on. There might be a bit of overshoot or underdamped ringing, which BWLIMIT prevents us from seeing.

Just barely visible is a smidgen of overshoot: the smoothed, BWLIMITed yellow trace just barely overshoots enough to kiss the upper dotted line. Then it recovers from overshoot and settles down to its final asymptotic value. Click on the white "X" at lower left to see the image full size and undistorted. The overshoot is also visible in the zoomed out (20usec/div) picture on the same page.

I also wish there were 3A or 5A "catch diodes" across the Zobel networks, to protect the output transistors when clipping into crazy reactive loads. And I wish the "OMG a rail is dead" safety diodes {Cordell section 15.3, p.325} D2 and D3 had a continuous DC current rating higher than fuses F2 and F3. I personally prefer to use the 10A04 thru hole diode in those positions; I don't mind paying 2 x $0.76 to buy them.

_
 

Attachments

  • BWLIMIT.png
    BWLIMIT.png
    78.9 KB · Views: 765

fab

Member
Joined 2004
Paid Member
Good project Bonsai
As usual you provide a very good informative manual and I appreciate that peoples can also learn in the process of diy.:)
I was impressed when I built your NX amplifier with « similar » topology as the SX or KX. The depth of the soundstage and the dynamic is special. It has motivated me into continuing experimentation of the “CFA” topology..

Fab
 
Bonsai, trust me, my KEF Q100 with my tweaks sound very well. They have a very similar frequency response than KEF LS50, front bass-reflex and 49 Hz at -3 dB. More sensitivity (2 dB) and minimum impedance (4.7 Ohms).

KEF saved on the furniture, with a thin DM. The 5" coaxial is very good (the 6.5", at least the old, NO).

They love power. I modified the bias of my AV Marantz SR4500, now is more high -> better sound. Maybe at low volume, in near field (with the PC), it works like class A, maybe.

My idea is the ALPHA Lender preamp + very clean power amp. Wich power amp? I do not know yet.