[FlukeII]

I forgot to mention, the VISHAY Bulk Metal® Foil Technology Industrial Precision Resistors, VSR Series resistors are very expensive. $$$$
 
Personally I favour the VISAHY Vishay Beyschlag MBA 0207, Professional Leaded Resistors for all the positions in the circuit:
 

 
I used the VISHAY MBB 0414, 1W 10E Resistors as the Emitter Resistors in the output stage. 
Latter I cahnged these to three (3) 30 Ohm MBB 0207 resistors wired in parallel (10 Ohm) to replace each of the 10 Ohm Resistors; to my ears and mind this improved the slew-rate of the amplifier - making a change in the speed and impact of the sound to be faster and a little more punchy. Putting a 1kHz square wave into the Amp and looking at the slope of the waveform on a Oscilloscope, there was about a 14% improvement in the Slew Rate, steeper angle of dV/dt, after making this mod. Kostas Metaxas takes this to the extreme, check out his website: http://www.metaxas.com  This is an example of what I mean:
 
Excerpt from IKARUS User Manual (downloaded off website):
 

 
 
My experience with the sound of VISHAY BC MBB 0207 resistors in the Lehman Amp circuit has been:

• Fast and rythmic
• Clean mids that are a little warm,
• Detailed imagery
• Sound that was alive, but to me natural
• I found every song I listened to to be independent in timbre and tonal characteristics

 
I was using the Beyerdynamic T1's and using a varity of Jazz and Alternate, such as Zero 7 - I really like their music and the variations on a single CD or Album!
 
I would be really interested to know what others have tried. One of my friends has used all 5% Carbon Resistors and likes the sound a lot - It wasn't to my taste.

 

[Knusperfisch]

 
Quote:

-2 x BC550 (stock LC) - 11 (Whatever that means...)
-2 x BC560 (stock LC) - 20 (Whatever that means...)
 
-10 x BC550 (bought in Germany) - 8 (Whatever that means...)
-10 x BC560 (bought in Germany) - 9 (Whatever that means...)

 
Thats really interesting..
Have you also measured the BD139 and BD140 from the stock LC compared to the BD139/140 you've bought ?

 

[ringer]

Quote:

 
Quote:

-2 x BC550 (stock LC) - 11 (Whatever that means...)
-2 x BC560 (stock LC) - 20 (Whatever that means...)
 
-10 x BC550 (bought in Germany) - 8 (Whatever that means...)
-10 x BC560 (bought in Germany) - 9 (Whatever that means...)

 
Thats really interesting..
Have you also measured the BD139 and BD140 from the stock LC compared to the BD139/140 you've bought ?

As I know you can not measure hFE of low power transistors by average multimeter. Only BD139/140 could be properly measured in Lovely Cube KIT.

 

[ringer]

My ST BD140-16 transistors has hFE from 192 till 205.

 

[BlaBlaBla]

Quote:

 
Quote:

-2 x BC550 (stock LC) - 11 (Whatever that means...)
-2 x BC560 (stock LC) - 20 (Whatever that means...)
 
-10 x BC550 (bought in Germany) - 8 (Whatever that means...)
-10 x BC560 (bought in Germany) - 9 (Whatever that means...)

 
Thats really interesting..
Have you also measured the BD139 and BD140 from the stock LC compared to the BD139/140 you've bought ?

 
Unfortunately, I have not measured stock LC BDs...
 
My intention was not to get right values. I wanted only to compare transistors. If I get wrong result for LC BC, the same error I will get for another BC (because they are the same PNP and NPN). But difference between those two errors will represent exact value.
 
Vishay is excellent brand. I use their products as much as I can. Their products are better and more reliable tham most hifi oriented companies.

 

[francisdemarte]

Default Vout voltages were +14.8V and -15.13V. Then I used 1k resistor and trimpot (200R) to tune Vout. Now Vouts are +14.16V and -14.2V. I lowered values on Vout to achieve minimal and symmetrical DC offset on both channels, and also I took care about difference between + and - (ideal is +14.2 - 14.2 but that does not give me satisfied results - DC offset 0.2mV and 4.5mV).
 
Without trimpots I had 0 and 8.5 mV offset. But I do not like such numbers. I like keeping the balance.
 
Lower Vout (compared to Vin) means bigger heat disipation on regulators, but they are enough tough to sustain 0.5A each. When I do something I always try to make my most expensive device satisfied. That means - headphones. Good one are expensive, need 50> hours to burn in, and are hard to find here in Croatia. If you are able to handle DC signal in a good way, then your long time expenses are kept to a minimum. If someone talks about amplifier, my first technical question is:"How high is DC offset in your amplifier?" And then goes rest.
 

 
I believe the M-Stage uses the same method. You can see the trim pots in this image. I wonder how the LCB can avoid the use of these trimpots?
 

 

[BlaBlaBla]

Quote:

Default Vout voltages were +14.8V and -15.13V. Then I used 1k resistor and trimpot (200R) to tune Vout. Now Vouts are +14.16V and -14.2V. I lowered values on Vout to achieve minimal and symmetrical DC offset on both channels, and also I took care about difference between + and - (ideal is +14.2 - 14.2 but that does not give me satisfied results - DC offset 0.2mV and 4.5mV).
 
Without trimpots I had 0 and 8.5 mV offset. But I do not like such numbers. I like keeping the balance.
 
Lower Vout (compared to Vin) means bigger heat disipation on regulators, but they are enough tough to sustain 0.5A each. When I do something I always try to make my most expensive device satisfied. That means - headphones. Good one are expensive, need 50> hours to burn in, and are hard to find here in Croatia. If you are able to handle DC signal in a good way, then your long time expenses are kept to a minimum. If someone talks about amplifier, my first technical question is:"How high is DC offset in your amplifier?" And then goes rest.
 

 
I believe the M-Stage uses the same method. You can see the trim pots in this image. I wonder how the LCB can avoid the use of these trimpots?
 

Yes. I did the same thing like in Martix M Stage.
 
The main fact is that Vout imbalance of almost 1V is equal to 18-20mV DC offset on headphone output(I have measured that). And second fact is that you will never hear big difference if there is not balance between + and -, because the same imbalance is equal on both channels (left and right). That topology has also bad point that is "low channel separation" (but also previous mentioned strong point that is in equal supply less prone to imbalance). That is reason why is more important to match left PNP (or NPN) with right PNP (or NPN) than trying to find "left PNP" - "left NPN" combination with equal hFE.

 

[ringer]

Guys, what do you think about quality of resistors in LC? Which do you think will be better to replace?

 

[BlaBlaBla]

Quote:

Guys, what do you think about quality of resistors in LC? Which do you think will be better to replace?

I think that they are (at least) 30% better than those in Matrix M Stage.
 
If you want to replace resistors then start with resistors in signal path (around Opamp and BCs). Next are output 10R resistors (1W). 1W is too small. 2W will be better. All others (before and after regulator)  I will not replace. They do not affect the signal too much.

 

[ringer]

What do you think about Epcos LongLife electrolythic caps? Is it good enough for replacing 4700uF capacitor?

 

[FlukeII]

 
Quote:

What do you think about Epcos LongLife electrolythic caps? Is it good enough for replacing 4700uF capacitor?

It our experiences, it's kind of like renovating a house - it's not complete until you've done everything. Yes you do get some gains replacing some parts, changing out the ELNA 4,700uF cap with a Low ESR type from VISHAY BC COMPONENTS - MAL205857472E3 - CAPACITOR, 4700UF, 40V we reckon you get quite an improvement in the bass and speed of the sound. The Low ESR caps seem to work best over the so-called audio caps from ELNA and NICHICON. (The Lovely Cube board we got had ELNA 4,700uF fitted)
 
Replacing the 3,3E CF resistors between the diode with a 1 Ohm (BI Technologies, BPC 5 1,0E 5W 1%) Thick Film resistor that has very low ESR ESL we think added to improving the pace and timing, along with the 4,700uF upgrade to the VISHAY MAL205857472E3 (This resistor is expensive and the audible gain wasn't worth it - sorry this is something just pointed out to me). Taking the resistors out all together seems to give the best sound - they seem necessary to prevent the 1A fuses on the original Lehmann Board from Blowing when you first turn on the power - the inrush current to charge the 4,700uF is quite high. (Update)
 
Probably the single biggest impact was going from a 30VA 15V-0-15V toroid to a 60VA 50VA or 100VA 120VA 15V-0-15V Toroid added a real increase in the presence of the mids, especially noticeable on vocals like Diana Krall; we found that one of biggest changes came with changing the Power Transformer:
Step 1: MULTICOMP - MCFE030/15 - TRANSFORMER, 30VA, 2 X 15V (Farnell Order Code: 9531726) - Sound's like a Lehmann

Step 2: MULTICOMP - MCFE050/15 - TRANSFORMER, 50VA, 2 X 15V (Farnell Order Code: 9531785) - Get more mids presence, timing seems to pick-up

Step 3: MULTICOMP - MCFE120/15 - TRANSFORMER, 120VA, 2 X 15V  (Farnell Order Code: 9531874) - Faster paced sound with real dynamic range, emotional impact of the music.
 
These are the best transformers we could find locally, you can get Nuvotem-Talema, these are really good but expensive...another guy uses: Nuvotem Transformer,toroidal,80VA,2x15V,2.67A (RS Stock No. 422-5269) (Nuvotem Manufacturer's part number 92422-P2S2).
 
The 300VA Transformer did not seem to make any audible improvement over the 120VA model...at least to me. (300VA Transformer mentioned in previous post)
 
One of my friends prefers the sound with 100nF rectifier diode capacitors removed from the board - says it gives a more in your face sound with rich PRaT; he used the AD823N Op Amp over the OPA and also uses a Super Better Transistor circuit to feed the V+ and V- to the Op Amp. I'll post a diagram of the Super Beta Filter if anyone is interested?

 

[FlukeII]

 
Quote:

I think that they are (at least) 30% better than those in Matrix M Stage.
 
If you want to replace resistors then start with resistors in signal path (around Opamp and BCs). Next are output 10R resistors (1W). 1W is too small. 2W will be better. All others (before and after regulator)  I will not replace. They do not affect the signal too much.

I've found that 1W Resistors work OK in the output stage as the 10 Ohm Emitter Resistors, it's just they look kind of small on the board. Getting the ESL (Effective Series Resistance Inductance) down by putting smaller resistors in parallel I found helps to improve the slew rate.

 
 
 
Those Emitter resistors are only dissipating around 20mW in heat, 2W is really over kill from an Engineering standpoint. The BC MBB 0207 0,6W in parallel are fine.
 
Changing the value of the 10E Emitter resistance changes the DC Operating point of the circuit, it also helps to degenerate some of the gain; Don't change 10E value - it screws up the operating point of the circuit...
 
To get the offset as low as possible, I've found that BlaBlaBla is correct, by setting the supply rail voltages to have the same magnitude |-15.4V| = |+15.4V| - However, the other half of the story is to match the Beta and VBE of the BC560C and BC550C - we're pretty sure this is what Lehmann does and would explain the hand-soldering on their Wave Soldered boards.
 
We used a METEX Instruments H-5040D DMM (belonging to a friend) with Transistor Gain Measurement to do this. Don't pick up the transistors with you bare hands, the heat dissipated will  change affect the readings. We ended up buying 250pcs of the FAIRCHILD BC560C and BC550C @NZD$0.04 each initially, before I switched to using the PHILIPS parts, we only managed to get hold of 100pcs of these.
The other guys are quite happy with Fairchild parts. We found that there was no noticeable difference with using the PHILIPS / NXP BC546C / BC556C or BC557C / BC547C.
 
We also found that the cheap no name brand DMM we had initially to measure Hfe wasn't very good, results drifted around, the meter only cost NZD$20 - the METEX gave much better results...
 
Please let me know if you try the Emitter resistor MOD above and how you find the sound?

 

[FlukeII]

If you're interested, someone sent me this link on transistors matching:
 
http://www.musicfromouterspace.com/analogsynth/TRANSISTORMATCHER/TRANSISTORMATCHER.php
 
Happy experimenting...
 
If someone is looking for a book that ( I think) gives a good introduction to Electronics, check out: Tab Electronics Guide to Understanding Electricity and Electronics [Paperback]
G. Randy Slone. For more advanced ideas around troubleshooting and stuff that isn't taught in Engineering School or in the text books, check out: Bob Pease's books:

• Troubleshooting Analog Circuits (EDN Series for Design Engineers) [# ISBN-10: 0750694998 or # ISBN-13: 978-0750694995]
• Analog Circuits (World Class Designs) [# ISBN-10: 0750686278 or # ISBN-13: 978-0750686273]

 
These aren't the only books on Earth, there are a lot of very good texts, the above is just a suggestion; I learned a lot from Pease's books.

 

[FlukeII]

 
Quote:

Originally Posted by FlukeII /img/forum/go_quote.gif
 
 
Changing the value of the 10E Emitter resistance changes the DC Operating point of the circuit, it also helps to degenerate some of the gain; Don't change 10E value - it screws up the operating point of the circuit...
 
To get the offset as low as possible, I've found that BlaBlaBla is correct, by setting the supply rail voltages to have the same magnitude |-15.4V| = |+15.4V| - However, the other half of the story is to match the Beta and VBE of the BC560C and BC550C - we're pretty sure this is what Lehmann does and would explain the hand-soldering on their Wave Soldered boards.
 

Two extra clarifications to make here:
 
1. The total resistance of the resistors in each BD139 / BD140 Emitter Resistor should remain 10 Ohms: You can use 3x 30E in parallel, or even 10x 100E --> The total resistance is 10E
 
2. The Metex meters only measure Hfe (at a preset current, i think it is around a few milliamps(?)) - you can't measure VBE. You need another circuit for that - see the link above.
 
In an ideal world it would be best to match the transistors under the identical conditions used in the circuit, e.g. same Ib and Ic, Ie etc. So even using the Metex doesn't mean that the transistors are matched as well as they could be - I guess it's a matter of how far do you go. This is to do with the shape of the Ib vs Ic vs V graphs for the transistors, you may get a match at one point of Ib, but get a gap at another Ib test current or operating value: Have a look at the data sheets for the BC560C and BC550C and compare the graphs for Transistor gain etc they vary a lot...
 
Certainly, using the Metex meter and going through to get about 60-70 matched PNP and NPN transistors measuring and noting the values displayed, we managed to get pairs that where within 2% 20% of each other - this gave a DC offset on the output that was around +/-0.01V - in one case we managed to get the DC Offset to about -0.002V, almost a factor of 10 better -> took a lot transistors to get to, we ended up measuring all 250pcs - You could say this was a bit extreme...
 
The reason for doing 60-70 pairs, was we wanted to see if the gain ranges of the BC560C and BC550C transistors had much affect on the sound - we couldn't hear any difference in the transistors we tried - we didn't test all the devices, we picked 10 pairs across the sample we made: Probably it was a waste of time; but we wanted to see - why not?
 
We do think that the higher gain C Hfe ranking of the BC550 and BC560 sound better than the lower A or B Hfe ranked devices. (These parts are available from the manufacturer in different gain ranges)
 
UPDATE: We found that we could get consistently close values of Beta (2 -5% for transistors) for NPN Transistors, but there was always a gap of at least 10 - 20% between the gains (Hfe) of the PNPs - I guess with the ON SEMI devices and the more wider spread of gain values, with enough transistors tested, you could probably get very close matches between the PNP and NPN devices. With the 100 PHILIPS devices we ended up seeing very close gain values between the NPN or PNP devices, but a quite a gap when comparing the PNP and NPN gains - nevertheless we still ended up with a DC off-set better then 0.01V or 10mV  - I think this is good enough(?). I checked with a high-end manufacturer of one brand of headphone amplifiers and he said that they would let their kit go out of the factory provided the DC Offset was under 20mV or 0.02V.
 
We set out an A3 piece of paper divided up into 1" x 1" squares (1" is easy to measure with an imperial ruler) and went through the NPN transistors writing down in pencil the value of the Metex meter, after we filled all the squares, we got out the PNP and started measuring the devices until we got readings that were identical to the NPN devices. This process took us about 3 days - you can only do so much in a day and you have to have beer and chill-out, it's boring as hell!
 
A friend works for a local company who makes domestic audio equipment and this is similar to the process they follow when matching transistors for the differential input pair of their amplifiers. (For what it's worth, he mentioned you can get the ON Semi versions of the BC550C and BC560C cheaper than the Fairchild ones, but he made note that the processes ON Semi use seem to mean that they have seen a much wider range of spreads of transistor gains, in some cases they have needed as many as 1000 pcs to match up 300 or so pairs for the PNP diff pair.
(They also match the current mirror NPN transistors to the same gain as the PNP, and <-- This was incorrect). It was a similar story with the number of transistors to go through for the PNP NPN pair matching. In no way are the ON Semi devices inferior, they are just different from a production spread point of view. He reckons they'll spend up to two - three working days before a production run to sort this out.) These guys were trying to match up NPN pairs and PNP devices. Thinking about it, if DC Offset is your thing, the ON Semi devices may be the way to go?
 
Either way you end with a lot of transistors that aren't a lot of use in less you're going into production...
 
I still feel that the PHILIPS BC550C and BC560C have a better tonal characteristic than the Fairchild devices  - it's personal taste...
 
The following gives an example; the NPNs under the Diff Pair form a Current Mirror and the PNPs are the Diff Pair - for more info see Doug Self or Bob Cordell's books:

 
 
 
We're playing around with a Bob Moog circuit to see if we can make it work to be able to measure PNP and NPN transistors...so far the results aren't great. Here's the circuit construct:

 
UPDATE: The issue with the Moog Circuit for matching transistors is that it's fine for matching NPN or PNP transistors to one another, but requires trimming to make it so that you can use it to match the PNP to NPN transistors. The reasons for this are the component tolerances in each of the testing circuits. The reason for matching the gains in the first place is to get as best as possible symmetric amplification of the +ive and -ive cycles of the audio signal, as well as getting the DC offset as low as possible. However, the 47E resistors in the legs of Base connections in the BC560C and BC550C provide some degeneration so that the characteristics of the stage aren't completely dependent on the Hfe of the transistors: If you want to know more about this read Doug Self's book Audio Power Amplifier Design Handbook, 5th Edition for example.
 
While the buffers are a unity gain design, e.g. Av=1, if you inspect the output waveform with 1kHz sinewave going in to the input, you see that the +ive amplitude of the sinewave cycle will be slightly different than the -ive cycle (this is called Waveform Asymmetry) - the matching of transistors helps to deal with this. How far you go is up to you - But, you'll never get it perfect that's impossible there are too many other variables at play to achieve that. Even laser trimmed OPAMPs such as the OP627 still exhibit some waveform asymmetry in the few devices I have played with.
 
Having a DC offset in the first place isn't all that bad, but too much offset, say more than 100mV, will disturb the dynamic characteristics of the headphone drivers (speakers) and affect they way the music sounds; low impedance headphones seem to be more affected - e.g. less than 100 Ohm. Some manufacturers will, I understand, compensate for this in the magnetic design of their drivers; others don't - it's a cost thing.Personally, the lower the DC Offset, the better, if you can get it right down then go for it.(Too much DC Offset and you'll damage the voice coil windings of the driver - not a good look, my Grado's smelt real good after that circuit I was playing with failed - see previous posting about the 2SD669As).
 
one of the reasons I prefer AKG, Beyerdynamic and Sennheiser headphones is because they engineer their drivers from the ground-up e.g. raw materials to finished product. I understand that some other manufacturers buy-in their drivers and I personally don't think they have the level of control over the design quality and performance that AKG, Beyer and Senn can achieve. The trouble is, with these brands you get what you pay for, I own a Senn HD-25 II and HD-218 headphones, to me the HD-25s eat the HD-218. (I own 8 pairs of headphones, but I don't have a flash stereo system - why? They all sound different and I like to switch between them; there are things that sound better in the AKG K-240s I have than the T1']s for example...it's just life!)

 
I recently bought a pair of Chinese made AKG K-271 MkII, the other AKG headphones I own, K-141 and K-240 were Austrian made. I really feel that HARMAN have gone too far with their cost cutting and the new K-271MkII is a compromised design IMO. To my mind the older K-271, Austrian made sounds better - a friend owns these.
Trouble is, Harman is run by lawyers and accountants (IMO) who seem to have one loyalty - the share holder IMO. I think this drives a culture of profit before quality while exploiting the reputation of a very good brand such as AKG. I understand that AKG no longer manufacturer in Austria their range of headphones. Shame, I think the Austrian build and quality was much superior to the Chinese cost improved models. This isn't a beat up on Chinese made - in China you get what you pay for...

 

[BlaBlaBla]

I think that we will have the best results if we make custom Lovely Cube. So far, we analized (almost) every part in the circuit. Design is simple enough to be very cheap but also good sounding if good parts are used.
 
Maybe someone who has a small factory reads this, and after few weeks he will start to make modified PCB that includes free space for all modifications and upgrades. It would be great. I will buy one or two.
 
I am too lazy to do the PCB alone.