sound quality of different transistors?

Hi all,

I'm curious if anyone has quantified the differences in sound quality between devices, small signal and output? There has been a lot of work done over the years on the differences between tubes; ie 2A3/300B, 12AX7/12AT7, etc, but not much in the solid state field as far as I know and my cursory search of the net has found apart from some obscure mention here and there.

This question crossed my mind after the huge difference I noticed replacing the ouput devices in my Digi-125 amp with the Toshiba 2SC5200/2SA1943 used in the AKSA amp.
At the time I was emboldened by the success of this experiment enough to consider replacing the small signal devices with MJ15030/1531's but in the end I decided on Toshiba's 2SC4793/2SA1832 which were alittle closer in Hfe/fT to the original BC546/556 devices.

After a lot of fiddling around removing/replacing the devices I plugged the amp back in and sat back for a listen. To my utter dismay and disapointment the music now sounded flat and totally uninvolving. Quickly I pulled the amp apart and reinstalled the original small signal devices, took about half hour and bingo the music was back again!

Obviously the small signal devices were responsible for this remarkable and very obvious change for there was nothing subtle about it...

Anyone have similar experiences/comments/opinions they would like to share?

tomcat:)
 
You might stop and think a little bit about the difference between the contribution of the circuit as well as the transistor. Also, ponder the concept of "voicing". I don't think you'd expect to take a speaker, replace one tweeter with a different tweeter, and expect only an improvement in the sound. The tweeter's performance in a given speaker must be "voiced" to the speaker, including the characterisitcs of the tweeter (impedance, frequency response, dispersion, etc), as well as the designer's goal for the speaker.

A well designed circuit is built around the performance and the characteristics of the parts used- regardless of the merits of one transistor or another, if you modify a circuit with a different device, and the cicrcuit wasn't designed around that device, you'll probably get different, or rather, indifferent results, regardless of the relative tradeoffs and quality merits of the devices.

Regards,

Jon
 
Tomcat, have you tried doing a search on this topic? Yes I believe the different devices make a huge difference to the sound produced. Good on you for giving it a go and remaining objective enough to realise that just because you went to the effort of changing something it was going to sound better.
 
I have wondered the same thing. There has always been plenty of discussion about XYZ brand caps sounding better than ABC brand caps.

I have never heard of any body "auditioning" brand ABC of bi-polar transistors against brand XYZ bi-polar.

Nelson Pass has, on more than one occasion, said he doesn't understand Audio Amature's neurosis about cap types, when "active devices introduce much more distortion."

tomcat: Did you allow for any break in or warm up period for your new components? Maybe that is an interesting idea for a project. A transistor burn in / tester / stresser.

Regards,

Aud_Mot
 
I tried different transistors on a standard-design power amp:
- input: bip (BC546b),mosfet(BS107),j-fet(BF245b); sound differce was obvious: ref bip, the fet had finer hi-range, but less resolution on low level( less room), the mosfet bad / sharp highs, maybe by their bigger inp capacity;
- output: power bip, mosfet; ref bip, the mosfet had more room and a little finer hi;
since that time i made my power amps with bip input and mosfet output, this was the optimum configuration for me; but dont forget, this is only one part of the amp-sound, the design begins at the power supply and ends at the kind of wires you use, all parts make a change in sound quality...

alf
 

tomcat

Member
2001-09-30 10:36 pm
Sth. Oz
Thanks to all who have replied thus far.

It's given that there are a certain number of circuit topolgies, be they tube or solid state that can use any number of output devices.

For instance had I asked the same question about tubes, there would, I am sure been a flood of correspondence on the sound quality of various output tubes in a given circuit using say, a PP60.

Some would insist that the KT66 is the only tube to use, others the EL37. But what of the humble 6BG6 or the venerable run off the mill old workhorse, everybodies ex military surplus 807? This tube is so plentifull they are almost dime a dozen. Yet nobody except a handfull take it seriously for audio because of it's industrial background, yet it is one of the sweetest sounding tubes around.

The same principle applies to solid state circuits. The Digi-125 is general purpose circuit similar in concept the tube one mentioned above and originally designed to use either the 2N3055/2955 pair or the MJ802/4502 o/p pair.
I can vouch that the latter devices sound much nicer/sweeter than the former.

Since then I have tried several other o/p devices including the MJ1004/1005 pair (hard sounding)and the 2SC2987A/2SA1227A (smooth, but oh so,,, ho-hum) devices before I stumbled on to the 2SC5200/2SA1943 as used in the AKSA, and these really made the Digi sing!

So there are differences, much greater than many imagine or, those who are in the know are prepared to concede.
Because if some can (as many claim) hear the difference that resistors, caps and cables can make in their systems, then surely they could tell the difference in sound quality between output devices as it would be just as audible.

Surely the onus is on those of us who have the ability to build our own (ss) amps and discover the sonic differences between components to share our discoveries/knowledge/experiences with others of the same ilk.

Logical, No? Or is that too much to ask. Hmmm???

tomcat:D
 
different transistors?

You are obviously ignoring Jon Marsh's post above. I think he made it clear that if you just change some active device in an amp, you really built yourself a different amp, unless you make sure that the new device is propperly 'integrated' in the amp as far as biasing, compensation, etc is concerned.
If you don't do that, any results and this whole exercise is meaningless.
Sorry, I don't want to spoil your fun, but if you just want to play around, then that's all you are doing.

Cheers, Jan Didden
 

rdarnall

Member
2002-01-24 2:14 pm
I'd like to make some comments about a few statements previously made...

"A well designed circuit is built around the performance and the characteristics of the parts used- regardless of the merits of one transistor or another, if you modify a circuit with a different device, and the cicrcuit wasn't designed around that device, you'll probably get different, or rather, indifferent results, regardless of the relative tradeoffs and quality merits of the devices."

I believe this statement is true and false.

True, in that any particular device has a certain set of general operating parameters and limits (voltage, current, power diss., etc...) that the circuit is to accomodate or follow.

But I disagree with this statement in that the device parameters can vary way too much even for the same "part number." I'm speaking particularly of transistors, although any device suffers from the same manufacturing variables, even something as simple as a resistor. These variances are more problematic in analog circuits than digital, I believe, simply because of the need for linearity.

Let's say you want to design a transistor amplifier (any type you want) and you want to use part 2Nxyz. Perhaps you would look at the spec sheets for its parameters, design, build, and test it. Then you decide that your design is the greatest thing since slice bread and go out and buy 100 more so you can build more and sell them to all your friends. Well, you put another 2Nxyz in, and the tests come back significantly different. That's because the first 2Nxyz is not the same as the others. You designed your circuit off of the spec sheets and that's only a guide.

Even worse is when you get out your curve tracer and design your circuit to that.

you cannot expect to replace components with "better" parts (or even same parts) and expect better overall performance. Unless, perhaps it's a circuit-dependent design rather than a device-dependent design. A device-dependent design relies mostly on the parameters of the device for its performance, but a circuit-dependent design allows for variations between parts while still maintaining its designed parameters. But alas, the real word is made up of compromises and both are always in effect.
 
Thanks rdarnall for clarifying what I was trying to say that obviously a couple of others overlooked.
If what they are saying were true, how is it that so many have been able to build both the early and the latter versions of the JLH SE 10w amp, as in both versions there are devices that are, either obsolete or hard to get so substitutes had to be found... And find them they did, with good results, even the small signal devices not just the outputs. In one example one chap changed from one device (I think it was 2n3055) to MJ 15003's and noticed an instant improvement! This is a good example of a circuit dependant design that you were referring to.
Another one I saw posted recently on this forum was the "Hiraga 20w" somebody built using the 2n3055 and BF149 (from memory) sounding quite nice... as most of the devices in this circuit are as rare now as hens teeth too, yet it worked (in this one example) with these common and easy to get devices.... So it is possible, in the tube fraternity they (I think) call it tube rolling, what can we call it with SS devices? Any Ideas?? Anyone??
tomcat :D
 
The Digi 125 Amplifier was designed for use with a number of different output transistors. Indeed I have built examples with
2N3055/MJ2955, TIP3055/TIP2955, MJE3055/MJE2955, MJ802/MJ4502 and MJ15003/MJ15004 output pairs. Probably the best sounding of these were the modules I built using the MJ802/MJ4502 combination. There are now more linear devices such as the Toshiba pair you have tried, and it seems that there may be adaptations of the original circuit that use these and other
modern output devices.
 
Thanks 900gripen for your input.
I was waiting for somebody else from Oz to speak up...
I agree the Digi is a very flexible circuit. You also hit the nail on the head when you say there are adaptations of it out there using more up to date devices...
I guess you have your suspicions just like I have mine:D
Like you I found the MJ802/4502 much bettere than the others until I tried the Toshiba 2SC5200/ 2SA1943. This is another step forward similar to the one over the 2n3055/2955 with the MJ802/4502.
You can get these devices from WES Components in Sydney and they aren't very expensive either.
I can wholeheartedly recommend changing to these devices.
I tried a few other mods but they weren't as significant and you have to be carefull what you change as the open nature of the sound disapears, though there are some mandatory mods that don't affect sound quality so much as increase reliabilty.
Upgrade R6 to a 2w as it gets hot and replace Q3 (BC546, which also gets quite hot and can go pop depending on the ps voltage you are using) with either the MJ340, BF469 or 2n3019 that I am currently trying in that spot. These only affect sound quality slightly, a bit like how many shades of grey are there. If you know what I mean. I would recommend heatsinking this device too. There are a couple of other worthwhile mods but that will do for the moment I think. Are you still using the Digi?, or are you into something else? Perhaps we could share experiences... Or start a Digi Forum, possible title: "How to Hot Rod your DIGI in 6 easy steps or less":D
Cheers
tomcat
 

Greg Erskine

Member
Paid Member
2002-01-05 11:56 pm
Sydney
Hi tomcat

I've been following your Digi125 thread, and have looked at the kit available from KITS-R-US. They don't go into much detail except its $39. Is this where you bought your's from?

If so, is this for one amp or two? What is the quality of the PCB?

I have downloaded some Digi125 articles from Paul Cambie's site and the circuit is extremely simple. From the original article it uses two diodes to set the bias but I think I can see a led in the middle of the very poor quality picture on the KITS-R-US site. Do you know if the circuit available from the KITS-R-US site has changed from the original articles?

By the sounds of it, you have replaced all the components, at least the transistors many times. Have you thought of putting in sockets?

Greg
 
Greg,

I think the price is for one board only plus parts. This would give you one channel. The boards are quite small, about two thirds of a business card lengthwise and a about a 1/4" narrower which is quite a feat initself, but awkward to work on because of the close proximity of tracks to each other, etc.
A better alternative is to get the larger boards from RCS Radio in Sydney, Ph: 02 9587 3491. email: [email protected]
They are located at: 651 Forest Rd, Bexley, NSW, 2207. I'm not sure of the price but I think they are around $8 ea. which is reasonable.
RCS Radio have boards for just about every project ever published in Australia by, ETI, AE and Silicon Chip.
The alternative boards they do for the Digi are quite good, much larger and therefore easier to use by an inexperienced constructor than the small ones. I think they are on the Mk6/7 versions of this board now, which should be the one you can use either TO92 (BC546)/TO5 (2N3019) or TO126 (BF469/MJ340) devices in the position of Q3.
On the question of sockets, I'm not sure they are available as these are discrete devices so you have to solder them in individually. I've seen Paul Cambie's site, but never saw the circuit diagram for the Digi there or anywhere else except the one on the earlier Digi-125 thread on this site posted by Palesha. In the original circuit there is no led amongst the diodes although my friend Mark had thought of the idea of installing a certain type of green led but gave up on it when he couldn't get the right ones here in Oz. In stead we settled on an extra diode making it 3. You are quite right, the circuit is very straight forward to build with no setting up for the novice to worry about. If you want a better copy of the circuit let me know and I'll send you a copy of the whole article, inc circuit diag by snail mail as I haven't a scanner.
Cheers,
tomcat
 
Humm>
My 2c worth. Small signal 2N5086 & it's comp,MPS-8500/8099,MPS-A18-A20, BC-546 and comp. High volt Predrive or secong gain stage TO-126. BF-469 & 470 as a Higher Ft MJE-340. the MJ15032-15033 TO-220 drivers work well in alot of Audio circuits and MJ-15024 &15025 as outputs as well as MJ-1294 &2195.2SC1302 is also great. my faveroite are the sanken RET transistors. as outputs. Fuji used to make some nice ones also
 
Hi Tomcat,

tomcat said:
Hi all,

I'm curious if anyone has quantified the differences in sound quality between devices, small signal and output? There has been a lot of work done over the years on the differences between tubes; ie 2A3/300B, 12AX7/12AT7, etc, but not much in the solid state field as far as I know and my cursory search of the net has found apart from some obscure mention here and there.

This question crossed my mind after the huge difference I noticed replacing the ouput devices in my Digi-125 amp with the Toshiba 2SC5200/2SA1943 used in the AKSA amp.
At the time I was emboldened by the success of this experiment enough to consider replacing the small signal devices with MJ15030/1531's but in the end I decided on Toshiba's 2SC4793/2SA1832 which were alittle closer in Hfe/fT to the original BC546/556 devices.

After a lot of fiddling around removing/replacing the devices I plugged the amp back in and sat back for a listen. To my utter dismay and disapointment the music now sounded flat and totally uninvolving. Quickly I pulled the amp apart and reinstalled the original small signal devices, took about half hour and bingo the music was back again!

Obviously the small signal devices were responsible for this remarkable and very obvious change for there was nothing subtle about it...

Anyone have similar experiences/comments/opinions they would like to share?

been there, done it ...
Have compared a lot of transistors in all places, for over 10 years now.

Yes, the 2SC5200 from the AKSA is very good. Just a month ago I repaired a Hirage Le Classe A (30w version of the classic 20w) with these, and the owner couldn't tell the difference. BTW, the Hiraga has been built around Toshiba output devices, and although L'Audiohile scratched away the numbers, pouring water on the plastic cases revealed them. I just don't remember where I wrote down these numbers ...

Re the MJ15030: I once tried it just before the output devices, but found it too slow, too. I prefer for many years now the BD139/140 from Philipps, biased at 6 .. 10mAmps.

Generally speaking: big transistors (get the Ic rating) sound powerful, but slow, especially when biased at current very small compared their max current rating.

regards,
Hartmut from Munich
 
Hi Alfsch,

alfsch said:
I tried different transistors on a standard-design power amp:
- input: bip (BC546b),mosfet(BS107),j-fet(BF245b); sound differce was obvious: ref bip, the fet had finer hi-range, but less resolution on low level( less room), the mosfet bad / sharp highs, maybe by their bigger inp capacity;
- output: power bip, mosfet; ref bip, the mosfet had more room and a little finer hi;
since that time i made my power amps with bip input and mosfet output, this was the optimum configuration for me; but dont forget, this is only one part of the amp-sound, the design begins at the power supply and ends at the kind of wires you use, all parts make a change in sound quality...

I did these comparisons, too, but came to different results: in the input, jfets are very good, if good types are used, like 2SK30 or siblings, otherwise bipolars are better. In the output stage, I tried several years to work with Mosfets, until I gave up, they make too much haze and spoil the flow of music, the ryhtm and pace.

There are big differences in sound between different JFETs, the European stuff is nearly unusable for hifi. Even US types are lacking. If you want to see a good sounding amp with Jfets in the input, look at the various Kaneda realizations (the preamp is on the Bonavolta page, don't know where the power amps are on the net).

regards,
Hartmut from Munich
 
Hi Tomcat,

There are, IMHO, several things one should consider.

I too have replaced transistors and have had both sides of the experience. Bad to worse and bad to better. Here are the things to consider:

Hfe. I have found that Hfe plays a big role in the sound quality one gets in both detail and warmth. In small signal applications one should try a higher beta transistor. One will get more detail and warmth owing to the fact that there is less base current and less Early effect (AKA base width modulation) making for a more stable gain stage. I had an amp once that used transistors of middle gain (100-200) and replaced them with high gain ones. The effect was astonishing. Warmer sound quality, a more tolerable high end, and more detail. The output transistors I retained the same values but a different brand. Yup, a different brand made a big difference in the sound quality also. Deeper bass and sweeter sound.

Basically, I went form Motorola to NTE. Another factor in the attempt to improve the sound. Manufacturer. I do not know if Motorola has a newer factory or not now, but refinement of the silicon, whether laser or (whatever the other type of process is used in cutting the pieces), the doping process, etc. and better "clean rooms" make for a better or worse product. NTE electronics started in the early 1980's with new laser equipped factories, and they make a very good product. I do not yet use any other brand.

As far as IC in audio, I have switched from NTE to Analog Devices. They are more costly but worth every penny.

Time. Yes time. Transistors also need a break in period. I purchased a Sony CD player for my car last year and did not like the sound. Sony components (transistors and amps) tend to sound very hard. This one was no exception. Yet after 8 months all of a sudden the darned thing started sounding so alive and warm. Evidently it needed a break in! But 8 months! It amounts to about 240-300 hours. So perhaps you didn't give your new components enough time to break in.


As regards FETs and MOSFETS, they depend, as with any other component, on the circuit they are in. I love the sound of MOSFETS (thee are ways to get arond the input capacitance problems, as with tubes etc). My first introduction to the FET was in a small phono player my father designed and built for me. It has such a kick and great sound quality for only ten watts (the pre was all FET). But at the time I knew very little about electronics and had not played with FETs until about 1982, ten years later.

Now 90% of my solid state designs are FET/Bipolar mix. I always replace any op-amp circuit with BIFET op-amps.

Anyway, that's my 2 cents worth.
 

Greg Erskine

Member
Paid Member
2002-01-05 11:56 pm
Sydney
Tomcat

Thanks for your pointers.

I followed up on RCS Radio @ http://www.cia.com.au/rcsradio/oldtonow.txt and will probably buy a pair of the mk6 PCBs for my next project, and just in case someone else is interested here's the latest prices.

et1430 $ 7.50 2068A digi-amp 25/ 5Owatt - mk1 1989May
et1430 $ 11.00 2068B DIGI-AMP 75/125watt - mk6 1989May

Their new location is 41 Arlewis St Chester Hill.

Paul Cambie's site had a pdf of the original ETI article in a yahoo briefcase.

As far as sockets go, I thought for experimenting you could do something like this:

http://peufeu.free.fr/audio/memory-9-photos.html

Thanks again, Greg
 
Gabevee said:
In small signal applications one should try a higher beta transistor. One will get more detail and warmth owing to the fact that there is less base current and less Early effect (AKA base width modulation) making for a more stable gain stage.

...

Basically, I went form Motorola to NTE.

The early effect is in fact stronger for higher h_FE transistors. In cascode configuration, however, this should not matter. Whether low h_FE is a problem depends mostly on the impedance of the previous stage .

Excuse my ignorance, but who is NTE?

Regards,

Eric