KRS Japan modded 5235

[martin2395]

Thought you guys might find it interesting, I got a Kenrick Sound modded JBL/Urei 5235.
Not sure if it's the standard work the guys carry but this one seems to have been done this year.

https://ibb.co/Rv06R2d
https://ibb.co/5GxwwZg
https://ibb.co/3BRww4F
https://ibb.co/Q6rGW1q
https://ibb.co/1mrRxvQ
https://ibb.co/bLRm0j3
https://ibb.co/HB4x4sy

As you can see, lots of $ gone into the unit: recapped, PSU caps uprated to 2x 4700uF, Cree schottky diodes, custom transformer, Amtrans resistors, TKD attentuators, modded to SE, custom KRS frequency cards.
What left me puzzled though, is the use of NE5532 instead of something boutique or even discrete in a DIP socket. The Signetics 5532N is regarded as the best 5532 ever made by some folks but it's still 'just' a 5532.

Haven't used it yet so just enjoy the pictures for now but I suspect it will be replaced soon by an XVR-1 or something similar, thought don't understand why the 4350 has different high/lowpass frequencies, assuming 5235 accounts for that?

Kinda tempting to try different op-amps but considering the unit sits between the pre and power amps, oscillation might fry the speakers or the amps...or both.

[dn92]

Thought you guys might find it interesting, I got a Kenrick Sound modded JBL/Urei 5235.Not sure if it's the standard work the guys carry but this one seems to have been done this year.https://ibb.co/Rv06R2dhttps://ibb.co/5GxwwZghttps://ibb.co/3BRww4Fhttps://ibb.co/Q6rGW1qhttps://ibb.co/1mrRxvQhttps://ibb.co/bLRm0j3https://ibb.co/HB4x4syAs you can see, lots of $ gone into the unit: recapped, PSU caps uprated to 2x 4700uF, Cree schottky diodes, custom transformer, Amtrans resistors, TKD attentuators, modded to SE, custom KRS frequency cards.What left me puzzled though, is the use of NE5532 instead of something boutique or even discrete in a DIP socket. The Signetics 5532N is regarded as the best 5532 ever made by some folks but it's still 'just' a 5532.Haven't used it yet so just enjoy the pictures for now but I suspect it will be replaced soon by an XVR-1 or something similar, thought don't understand why the 4350 has different high/lowpass frequencies, assuming 5235 accounts for that?Kinda tempting to try different op-amps but considering the unit sits between the pre and power amps, oscillation might fry the speakers or the amps...or both.

In the past I did a thread on a French forum where you can find lots of information about JBL UREI 5235, including values and frequency responses for JBL 4350 dedicated boards.https://www.jbl-vintage.com/t3066-fi...bl-urei-5235Be carefull if replacing 5532 op amp as power supply voltage can be above common tolerances of most of the op amps. In JBL UREI 5235 they can be replaced by LME49860, but also by very modern op amps like OPA1612 or OPA1656 (can be found already mounted on smd/dip adaptor). Low noise linear voltage regulators such as sigma78/sigma79 from amb lab is also an improvement.

[martin2395]

In the past I did a thread on a French forum where you can find lots of information about JBL UREI 5235, including values and frequency responses for JBL 4350 dedicated boards.https://www.jbl-vintage.com/t3066-fi...bl-urei-5235Be carefull if replacing 5532 op amp as power supply voltage can be above common tolerances of most of the op amps. In JBL UREI 5235 they can be replaced by LME49860, but also by very modern op amps like OPA1612 or OPA1656 (can be found already mounted on smd/dip adaptor). Low noise linear voltage regulators such as sigma78/sigma79 from amb lab is also an improvement.

I've enjoyed your thread through google translate yesterday evening already
Regarding replacing the op-amps - I'm actually more concerned about high frequency oscillations when using high speed op-amps. I've already seen a BSS FDS360 oscillate when used with IC's like the 4562 (IIRC) and considering there are no more safeguards between the crossover and the power amps, the crossover must be rock stable.


I found a dual OPA627 in a DIP adapter at audiophonics, haven't checked the voltage requirements yet. AD713 is borderline within spec at 18V max voltage? Also - there seem to be many variations on the AD713, the AD713JNZ seems to be the current production.

[dn92]

I've enjoyed your thread through google translate yesterday evening already Regarding replacing the op-amps - I'm actually more concerned about high frequency oscillations when using high speed op-amps. I've already seen a BSS FDS360 oscillate when used with IC's like the 4562 (IIRC) and considering there are no more safeguards between the crossover and the power amps, the crossover must be rock stable.I found a dual OPA627 in a DIP adapter at audiophonics, haven't checked the voltage requirements yet. AD713 is borderline within spec at 18V max voltage? Also - there seem to be many variations on the AD713, the AD713JNZ seems to be the current production.

As replacement of AD713 you can use OPA1644 on a DIP-SMD adaptor. One friend of mine tried also OPA4134. Instead of 5532, OPA627 has to be on a dual op am DIP-SMD adaptor, as there are single op amps.Do you already have the cards for JBL 4350? If not you can get them from Restaur Audio located in France (as I read that you are located in Europe). For 4350 you can also use cards for 4355. LF and HF crossover frequencies are tuned to include the response of the loudspeakers driver.

[martin2395]

Yes, it also came with Kenrick Sound 4350 cards.

[Ian Mackenzie]

I recommend you try it as is. It’s been updated with modern parts which is all that can be done.

A lot of the stability and more importantly noise comes from the pcb layout which on the JBL crossover design was only engineered for the TL072 or similar op amp. To perform significantly better the residual noise needs to be lowered. The main source of noise is from the crossover card high pass RC values which would need scaling so R = from 600 to 1000 ohms. The NE5532 will do this done properly.

The point is that better sound comes from transparency which requires an order of magnitude lower noise floor. The NE5532 already has low distortion. It’s the noise from using resistor values that generate noise.

The other culprit is noise generated by the opamps operating in class B. This causes commutation noise on the power supply pcb tracks and the ground tracks. This cross feeds through all the opamps. Local power supply de coupling is well advised with all opamps. Sometimes RC decoupling with 220 uF capacitors and R 100 ohm resistors helps. A separate power supply ground should always be used. This is adds complexity and cost which often is not viable in this product category.

Alternatively the opamps can be biased into class A with a current source to 2-3 Ma. This will cause heating of the opamps and more current load on the power supply. This is why mixing consoles run hot and the electrolytic coupling capacitors need replacing. The technical staff at Abby Road Studio say it’s a never ending problem.

I would recommend removing the transformer from the chassis as a primary cause of noise and locating it in an external chassis.

The NE5532 is notorious for being blamed for so so subjective performance when in fact it’s the poor execution of the Opamp that is at fault. Modern recordings flow through hundreds of NE5532 op amps and no body complains. Discrete op amps are the ultimate but at a price which makes mass production not viable.

Modern op amps have up to -20 db less noise than the TL072 but the surrounding circuits must be optimised to take advantage of the low noise performance.

It’s not uncommon to see resister values between 10K and 47K in analogue active crossovers. This is done because A) the designer hadn’t followed proper design practices or B) high value resisters enable small low cost capacitors to be used. Often Mylar or polyester capacitors are used in favour of superior polypropylene capacitors because polyester capacitors are of much lower cost. However polyester capacitors introduce measurable THD according to testing completed by Douglas Self. Self also found that large electrolytic signal coupling capacitors of 220uF or more have much lower distortion than small values of 22 uf.

Chip rolling is generally nothing more than entertainment.

In summary to achieve the above the entire pcb would need re designing and an external power supply.

https://www.booktopia.com.au/the-des...SAAEgK2ofD_BwE


https://www.ti.com/lit/ds/symlink/op...Fproducts.html


https://www.ti.com/lit/ds/symlink/op...ct%252FOPA1602

[martin2395]

So pretty much better to sell it as it is because it cannot be brought to modern levels, sound wise?

[dn92]

So pretty much better to sell it as it is because it cannot be brought to modern levels, sound wise?

Mine has -10dB both for noise and THD than the stock model. I used modern low noise linear voltage regulators, with increased values for capacitors in this section, changed the input resistors to a lower value, changed TL074 to OPA1644, used lower resistor values on crossover cards, lower resistor values of HF output, changed 5532 by OPA1612.

[martin2395]

Just curious, have you compared it sound-wise to more hi-end gear like Bryston / Pass / Accuphase crossovers?

Also, Sigma 78/79 series seem to be discontinued, do you know any good alternatives? I'm looking at Sparkos 7818/7918 at Audiophonics right now.
Did you have to install heatsinks on yours, do you run them on 18V output?

[Ian Mackenzie]

Mine has -10dB both for noise and THD than the stock model. I used modern low noise linear voltage regulators, with increased values for capacitors in this section, changed the input resistors to a lower value, changed TL074 to OPA1644, used lower resistor values on crossover cards, lower resistor values of HF output, changed 5532 by OPA1612.

Good outcome.

How did you measure this?

Btw the JBL SN measurements are at the full rated output of 6.2 volts. Because the residual noise is a constant voltage the SN is actually the ratio of output signal voltage to the noise voltage. This noise measurement will be different for the low and the high pass outputs depending of the crossover frequency.

To determine the actual SN in your operating conditions simply normal the JBL measure to your normal output voltage from the active crossover.

For example to obtain a reference level at 85 Db at a listening distance of 10 ft the signal voltage out of the crossover is only 0.14 volts assuming a power amp gain of 26 db and 1 watt output.

Recalculation of the JBL 5234A / 5235 SN is 90 db - 33 db = 57 db. The dynamic range of a vinyl recording is generally quoted as up to 70 db.

Even if the spl at 10 ft was increase to 95 db with a power output of 15 watts the voltage output of the JBL 5234A is only 0.6 volts. Normalising the JBL SN measurement of 90 db ref 6.2 volts gives a SN 90 - 20.3 db = 69.7 db. What this means is that at 95 db spl the theoretical noise level is 25.3 db.

The topic of psycho acoustics is beyond the scope of discussion here but it should be noted that the SN under this operating condition falls within the dynamic range of a vinyl recording and well inside the dynamic range of digital signal sources.

These levels are regard as a quiet library. This is the thing. Once you adjust to the library environment any noise like someone tapping a pen or folding a page or a fart is really annoying. I think that’s a good analogy with what l am describing with SN. You can’t ignore it.


The astute individual or trained audiophile listener will know that the cochlear starts compression in a non linear manner with frequency dependent on on age and listening damage. The ear can readily discriminate particular sounds out of a range of other sounds very acutely such as a buzz, a cracking sound ect that would otherwise be regarded as not noticeable.

I offer no conclusions here other than the scientific calculations based on actual facts.

My point is do not be mislead by manufacturers SN. Think it through and look at how it fits with your scenario.

In my situation the SN of the active crossover must be equal or greater then -105 db at a reference level of 1.4 volts. When normalised the JBL 5234A has a SN of -77 db @ 1.4 volts. Thats a gap of 28 db in performance. My top line design reach better than -115 with an implementation that improves the SN of the whole signal path.

Unfortunately you simply can’t make a silk purse out of a sours ear. It doesn’t work out because the performance is the total sum of the entire design and the implements. Things like an iron core transformer inside the chassis are a clue. The length of the pcb tracks from the Opamp to the plug in crossover cards. These are just the obvious issues.

Of course you can do diy tweaks but don’t wish for the world

[dn92]

Good outcome.How did you measure this?Btw the JBL SN measurements are at the full rated output of 6.2 volts. Because the residual noise is a constant voltage the SN is actually the ratio of output signal voltage to the noise voltage. This noise measurement will be different for the low and the high pass outputs depending of the crossover frequency.To determine the actual SN in your operating conditions simply normal the JBL measure to your normal output voltage from the active crossover.For example to obtain a reference level at 85 Db at a listening distance of 10 ft the signal voltage out of the crossover is only 0.14 volts assuming a power amp gain of 26 db and 1 watt output.Recalculation of the JBL 5234A / 5235 SN is 90 db - 33 db = 57 db. The dynamic range of a vinyl recording is generally quoted as up to 70 db. Even if the spl at 10 ft was increase to 95 db with a power output of 15 watts the voltage output of the JBL 5234A is only 0.6 volts. Normalising the JBL SN measurement of 90 db ref 6.2 volts gives a SN 90 - 20.3 db = 69.7 db. What this means is that at 95 db spl the theoretical noise level is 25.3 db. The topic of psycho acoustics is beyond the scope of discussion here but it should be noted that the SN under this operating condition falls within the dynamic range of a vinyl recording and well inside the dynamic range of digital signal sources.These levels are regard as a quite library. The astute individual or trained audiophile listener will know that the cochlear starts compression in a non linear manner with frequency dependent on on age and listening damage. The ear can readily discriminate particular sounds out of a range of other sounds very acutely such as a buzz, a cracking sound ect that would otherwise be regarded as not noticeable.I offer no conclusions here other than the scientific calculations based on actual facts.My point is do not be mislead by manufacturers SN. Think it through and look at how it fits with your scenario.In my situation the SN of the active crossover must be equal or greater then -105 db at a reference level of 1.4 volts. When normalised the JBL 5234A has a SN of -77 db @ 1.4 volts. Thats a gap of 28 db in performance. My top line design reach better than -115 with an implementation that improves the SN of the whole signal path. Unfortunately you simply can’t make a silk purse out of a sours ear. It doesn’t work out because the performance is the total sum of the entire design and the implements. Things like an iron core transformer inside the chassis are a clue. The length of the pcb tracks from the Opamp to the plug in crossover cards. These are just the obvious issues. Of course you can do dot tweaks but don’t wish for the world

Ian, the SNR of the 5235 is not given for full output level and is -90dBV, with its output stage which is different from the one of the 5234A. With my mods I’m around your spec of -105dB.SNR depends anyway from the installed cards, and can be better with custom cards instead of JBL card by lowering resistors values (and changing capacitors values accordingly).Where is your design described?

[Ian Mackenzie]

Ian, the SNR of the 5235 is not given for full output level and is -90dBV, with its output stage which is different from the one of the 5234A. With my mods I’m around your spec of -105dB.SNR depends anyway from the installed cards, and can be better with custom cards instead of JBL card by lowering resistors values (and changing capacitors values accordingly).Where is your design described?

Okay. Firstly l will clear any confusion on the JBL 5234A and the 5235 SN.

Putting aside the product differences I’m referring to the JBL spec attached below at + 18 db which is the same on both the 5234A and the 5235 tech sheets.

My statement are correct if you read the architectural specifications correctly and cross reference to the rated output of 6.2 volts stated in the Specifications. What’s really happening is JBL have fibbed as they do to make their electrics look good on paper. Their SN spec is way over stated as l have stated above. It always was and always will be.

My recommendation is to reference your SN at a realistic voltage level such as the input sensitivity of your power amp. This is typically 1.4 or 1.7 volts.

[Ian Mackenzie]

The other less obvious thing to appreciate is that residual noises sum together. Unfortunately this means that even if your source component, preamplifier, active crossover and power amplifier all have a SN of 100 the final SN is not 100 db.

It’s 88 db. Thats right.

If your preamp SN is 80 db which is typically of a hifi preamplifier then your SN is never going to be 100 db overall. With the stock 5235 un use your preamp with a SN of 80 @ 1.4 volts and a power amplifier and sources with a SN of 100 db @ 1.4 volts the overall system SN is 71.7 db. Thats right.

This is a realistic scenario where your source is a CD player with a SN of 100 @ 2 volts output. What’s more concerning noise sits on top of the THD. This is seen in amplifier distortion measurements versus signal magnitude where the distortion appears to rise quickly at low signal levels. This is because the noise is greater than the THD.

So your THD + system SN is a limitation.

Subjectivist will speculate that such low levels are in audible. But the audiophile can readily perceive 2nd order HD -80 db of an amplifier below the signal level. Nelson Pass First Watt amplifier range exhibits individual distortion profiles that people like. If your noise level is above your signal path THD then the effect this has on your listen experience could well be negative. You’re certainly not hearing your system at its best.

This is why it’s important to be aware of SN across of all your audio components and how it was measured? The component(s) with the lowest SN are going to limit your dynamic range.

[Ian Mackenzie]

I enjoy the enthusiasm in these diy forums and l provide some useful technical insights to keep the discussions constructive and productive. The rest is up to you.

On my designs l don’t publish that on an open public forum and nor do l disclose details of the designs. I appreciate people are interested but I’m unfortunately there are information thieves lurking on cyber space and those that want to proffer from others without doing the real work.

I commission equipment and designs for specific needs and for those who are invested in the hobby but don’t have the time or the skills to realise their ambitions. Next year in 2024 l plan to start making a range of equipment to suit diy enthusiasts and audiophiles alike. The aim is to make it easier for the user to achieve their expectations. One of the designs was referred to as a monster DX1.

If you do search on the forums l have described the design principles of an analogue active crossover in some other threads a few years ago. It is a fairly involved subject.

[martin2395]

OTOH, why do we have to run the HF output through the active xover, just for the sake of the highpass for the 3107? Curious what would happen if we skipped the active xover for the MF/HF section and ran it directly from the amp.