| Shaun |
Hi
What are the implications of substituting a video amplifier transistor such as BF871 in place of a MJE340 as a Vas transistor? Would the Cdom have to be changed?
Regards
Shaun |
|
|
| AKSA |
Hi Shaun,
Video transistors are good in this roll.
Cdom need not be changed, BUT it may be gainfully reduced.
A very good one to use is the Sanyo 2SC3955, a 300MHz beast with impeccable specs and collector capacitance of just 1.8pF.
Cheers,
Hugh |
|
|
| janneman |
Hugh,
Those Sanyo's look pretty good indeed! Normally these types have relatively low Hfe, but these have groupings going up to 160 min.
Thanks for the tip.
Jan Didden |
|
|
| Shaun |
Thanks for the info, Hugh.
Regards
Shaun |
|
|
| capslock |
Dear Hugh,
I also like this Sanyo series of video transistors. The problem is it seems to be discontinued, I even got that info in Japan.
While there is still a limited supply of the C3955, its counterpart 2SA1540 seems to be gone for good.
I can still get a pair from the same series that has a lower current rating, but it's no longer possible to order them from the same gain group.
Do you have a special source or know about adequate replacements?
Regards,
Eric |
|
|
| Shaun |
In a brief search for high-voltage video transistors, I got the impression that most of these devices are now being offfered as surface-mount parts. I hope that there will still be leaded parts in circulation in the future...
Shaun |
|
|
| AKSA |
Hi Eric,
Well! I did not realize it was discontinued!! Bummer!
But it is available at www.pacificsemi.com (US46c), and www.bdent.com (US92c) in one off and 100+ quantities.
I made a list of transistors suitable for this work some time back but knowing me I've lost the damn thing.......
Sanyo has a proprietary manufacturing process which reduces collector capacitance, and frankly this makes them very desirable as VA devices.
I will seek out more; I buy from Peacock Industrial in Singapore and yesterday sent out an order. I'll let you know.
Cheers,
Hugh |
|
|
| capslock |
Dear Hugh,
thanks for your link. Pacific Semi has good prices, but you'll notice their stock is 0 for all of these parts. I have no idea whether that means the do special order only or if it's simply no longer available.
BD do not say whether they have stock.
Do you have any way of ordering parts form the same h_FE group?
Greetings,
Eric |
|
|
| capslock |
| Just found out they tell you if a part is out of stock! Anbody have experience with bdent? |
|
|
| mlloyd1 |
Just to share some info....
I just got some 2SA1540 and 2SC3955 from MCM (in USA).
And no, I haven't checked to see if they are fake or not. :(
Both NPN and PNP were "D" grade. That means HFE @10mA range of 60-120.
FYI
mlloyd1
oops, I forgot: price was $1.15 and $1.07 USD (>15 pieces each) respectively |
|
|
| jcarr |
If the 2SA1540/2SC3955 meet your requirements, similar Sanyo TO-126 devices include the 2SA1478/2SC3788, and 2SA1209/2SC2911. Perhaps also the 2SA1538/2SC3953.
In a molded TO-220 package, the 2SA1697/2SC4474. Definitely.
Then there's also Toshiba's 2SA1360/2SC3423...
hope this helps, jonathan carr |
|
|
| AKSA |
Jonathan,
My sincere thanks. I will look into all of them, although I note that 2SC3423 is rated at 200MHz, though it does have collector capacitance of only 1.8pF. (The others are 400MHz I believe).
What's your feeling about the Toshiba 2SC4439 (400MHz, 150V, 8W, 4pF)
Cheers,
Hugh |
|
|
| jcarr |
Hugh: The Sanyos that I listed are for the most part 150~200MHz devices, although I think that the 2SA1697/2SC4474 are 300MHz.
OTOH, for linear audio applications, once you clear the 100MHz point, IMO factors like Cob, Cre, Vaf, and linearity matter much more than finding the ft champion. In the Sanyo range, some devices sport both higher ft and higher Cob than other, similar devices, but personally I would prioritize low capacitance over high ft. Besides, as you know, the maximum ft rating is only realized at a specific Ic, usually with a clear drop-off on either side of that IC, the value of which may or may not coincide with the current that the rest of your circuit wants to run at.
>What's your feeling about the Toshiba 2SC4439 (400MHz, 150V, 8W, 4pF)<
I haven't used it yet. No idea of how it performs and sounds in the real world. But the data sheet looks reasonable, at least. Let's run a very rough comparison between it and the Sanyo 2SC3788.
Both the Sanyo and Toshiba clear 100MHz ft by around 3mA Ic, at 10mA, the Toshiba is kissing 200MHz while the Sanyo is still at about 150Mhz. However, the ft of the Sanyo appears pretty uniform at any Ic between 15~40mA, while the Toshiba's ft curve stabilizes between 30~100mA.
The Cob of the Sanyo starts out at perhaps 4pF at 1V Vcb, and drops down to 1.5pF or so at 100V. For the same voltage swing, the Toshiba goes from 10pF~3pF, so I would expect distortion to be higher on the Toshiba.
The hfe-Ic curve is clearly more linear on the Sanyo. The Early voltages are probably somewhat lower with the Toshiba, but both devices look to be pretty mediocre in this parameter. OTOH, the Vbe-Ic stability appears to be superior with the Toshiba.
If I was only allowed to judge from the data sheets, I would probably choose the Sanyo (unless I had specific requirements in mind, which _is_ a common occurance), but the Toshiba does look attractive enough to consider ordering samples and breadboarding some circuits.
My two pence... jonathan carr |
|
|
| AKSA |
Hi Jonathan,
My thanks again! This is invaluable information, and I cannot find a single point of disagreement. :nod: I too give preference to collector capacitance and uniformity of beta; these to me are the big ones. I feel that since the application calls for a stage current around 8mA, any more Ic than 100mA is not required, enabling TO-126 packages and generally more nimble performance.
I have continued searching and settled on the Sanyo complementary pair 2SC3600/2SA1406. I do not know if they are obsolete; I hope not. The 3600 (my app is npn) looks very attractive, with 200V, Cob at 1.5pF/70V and dead linear, saturation voltages at around 150mV (very good for negative swing linearity), and hfe as straight as a die from 5 to 45mA. I would suggest this would be a killer Voltage Amplifier.
I have enjoyed your website. True high end; I doffs me cap. And clearly you are to be admired for mastering the only language jokingly described as an '...impediment to communication'!!
Cheers,
Hugh |
|
|
| capslock |
In designs with a single-ended VAS (rather than complementary), common wisdom has it that the constant current source (CCS) transistor may be much slower than the VAS transistor.
Sanyo has some superfast video transistors which don't come in complentary pairs, so one would be forced to use a slower part for the CCS.
Now, it's true that Miller capacitance is not that important in the CCS because the base can be buffered with a large cap. On the other hand, I am not sure that f_T is really unimportant. I think it is still needed for the transistor to react to HF changes in the load conditions. True?
If so, a slow CCS would introduce signicant phase lag. One could put a .5 to 5k resistor in series with the CCS collector, which would probably eliminate HF lag issues, but this would limit voltage swing. |
|
|
| Christer |
| quote: | Originally posted by capslock
Sanyo has some superfast video transistors which don't come in complentary pairs, so one would be forced to use a slower part for the CCS.
Now, it's true that Miller capacitance is not that important in the CCS because the base can be buffered with a large cap. On the other hand, I am not sure that f_T is really unimportant. I think it is still needed for the transistor to react to HF changes in the load conditions. True?
If so, a slow CCS would introduce signicant phase lag. One could put a .5 to 5k resistor in series with the CCS collector, which would probably eliminate HF lag issues, but this would limit voltage swing. |
You can add a cascode transistor to the CCS to speed it up. |
|
|
| capslock |
| Limits voltage swing, too, and does not do a thing dynamically (f_T remains the same if it's the same type). |
|
|
| Christer |
It will lower the voltage swing a bit, yes, but it does speed up
the CCS since it keeps Vce almost constant for the CCS
transistors, thus avoiding the Miller effect in it.
This was discussed a while ago, by the way, in another thread,
but I can't remember which one right now. |
|
|
| capslock |
| The Miller effect should be irrelevant in a CCS because the AC impedance of the control voltage at the base is essentially 0. The cascode transistor will increase (LF) output impedance further by getting rid of the I_C vs. V_CE variation. There is nothing it can do about the limited f_T, though. |
|
|
| Christer |
I will have to think about this. I got the cascoding idea myself
once (it is a pretty obvioous idea) and if my suggestion was
only based on that I would probably yield and assume I was
wrong. However, this idea was later suggested on this forum by
other people. I think it was an EE who suggested it, although I
may be wrong there. I haven't succeded to find the thread yet.
Does anybody else have an opinion on cascoding CCSes? |
|
|
| mlloyd1 |
My experience is that cascoding the CCS primarily boosts the output impedance (obviously over a limited range); I agree with capslock.
The question then becomes: of what benefit is this characteristic in your particular application? :scratch:
I think most often, not vey much ;) (Oops, did I mean to say that)
jcarr, hugh: very useful comments! thanks!
How rugged (as in damage resistant) are these VERY low capacitance devices in real world amplifier applications? Any historical experience to share? Generally speaking, low capacitance implies small size which doesn't always generate warm fuzzy feelings of robustness. I've even seen a significant number of mfr schematics that parallel devices in the VAS, which brings the capacitance back up. Yes, I know Sanyo has a fancy proprietary process for fabricating these devices. :)
mlloyd1 |
|
|
| traderbam |
"Does anybody else have an opinion on cascoding CCSes?"
A constant current source is slow by definition. It is meant to provide a constant dc current. Ft is not directly relevant.
An ideal CCS has infinite Z. A single BJT source will have a dc resistance caused by changes in beta with Vce (because Ie = Ib+Ic). This can be reduced by increasing the bias voltage across the emitter resistor. The ac Z is this dc resistance in parallel with Ccb and Cce and the miller effect of Ccb. If the base voltage bias circuit has zero Z there is no miller effect. The capacitances can be mitigated with a series R if you have the voltage headroom.
The dc resistance effect and the capacitive effects can be mitigated by keeping Vce constant. This is what a cascode tries to do. However, you get the effect of the Ccb and Cce of the cascode transistor. However, the miller effect is virtually eliminated. The cascode also maintains a constant power dissipation of the CCS transistor so thermal drift is mitigated too.
To realize all these benefits, each transistor must have an independent base bias circuit. |
|
|
| capslock |
These Sanyo devices have a respectable current and power rating, so the chip would not be very small.
My concern about the slow CCS is twofold:
- Parasitic C_cb and C_ce would cause lag, limiting the bandwidth available from the VAS and maybe even causing oscillation.
- Above f_T the CCS certainly no longer is a CCS, so impedance of the VAS stage drops, causing additional droop above f_T, which causes another pole in the open loop gain which will force us to use a much more severe frequency compensation. Just as an example, consider using one of the Sanyo NPN devices with about 1 GHz f_T as a VAS. As a CCS transistor, we use the PNP BF470. This is an old video transistor that has very low output capacitance (1p8 pF) but also a low 60 MHz f_T. Gain is pretty pointy, so at the current chosen, f_T may well be only 10 MHz. This will force us to compensate the whole amp for a unity gain bandwidth of 5 MHz (or 250 kHz total bandwidth at a gain of 19) where using the fast VAS transistor and fast output devices, a unity gain bandwidth of 50 to 100 MHz should have been easily achievable. This is admittedly an extreme example, but it serves to contradict the notion that the speed of CCS transistors is completely irrelevant. |
|
|
| capslock |
| quote: | Originally posted by traderbam
To realize all these benefits, each transistor must have an independent base bias circuit. |
The base circuit is just a low impedance voltage source. So if you use heavy capacitive buffering, one should be sufficient for all CCS, shouldn't it? |
|
|
| traderbam |
| Yes, if the bias circuits are extremely good then you can use one. But then you wouldn't need to cascade in this case (other than for thermal stability reasons). So I would use two just to make the most of the cascade: it doesn't take many parts. That's my opinion. |
|
|
| kimschips |
Hello all.
All the Sanyo transistors you are talking about can be optained from these:
www.heldt-electronic.de
Push "Aktive Bauteile" in left menu to see the catalogue.
JørgenP in Denmark Phonenumber is +4544945244
They only have a giant catalogue in paper but they get you all types.
By the way I am ordering Sanyo transistors now from Heldt electronic so let´s see how it gets!
Best regards
Kim |
|
|
| amplifierguru |
Ha chasing that elusive 1pF Vas transistor with 500MHz and Beta of 500 all peaking at 10mA for a Vas! Sifting through obsolete devices. Think of the PSRR when I slug it to ground instead of miller. Brings back memories. Got all that extra GBW from the dominant pole stage, hope the other stages are up to it.
Commercially you need to spec devices that support you're design in 10 years.
Of course output stage distortions under reactive loads are likely to swamp the early distortion and the sum of the squares is?
Feels good though. |
|
|
| kimschips |
Hey
Amplifierguru could you please explain what you mean with your first lines in your message? an exsample?
What kind off transistor are you talking about?
In which way will PSSR be affected?
Best regards
Kim |
|
|
| Workhorse |
Hi everyone,
how about using a Mosfet for VAS[drawback it has huge Cgs & Cds]. Since Nelson Pass has also used it in its designs.
regards,
kanwar |
|
|
| darkfenriz |
mosfets have terribly low trnsconduction compared to bjts.
I would rather think about RF bjt. Have you ever used RF bjt? |
|
|
| amplifierguru |
Hey
kimschips - just as added miller Cdom capacitance is used to slug down the GBW of the Vas to establish the dominant pole which also degrades PSRR to the supply that the Vas input is referenced to,... so too with the intrinsic Ccb of the device which sets the upper limit. intuitivey a device with 1pF is better than one with 2pF. |
|
|
| cunningham |
If the rest of the circuit is made for it, what is wrong with designing the VAS Iq around 1-2mA? Then small, Lo Noise devices like FJV992 (& FJV1845) can come into play, particularly the 'E' gain series, Beta 400-800! Fairly linear gain up to 6 or 7mA. I am playing with a circuit that uses reg. +/- 60V with this device as the VAS sourced with the FJV1845 mostly out of curiosity, well when I am at home...I work out of state. :xeye: I'm in Utah right now...nice scenery. A bit different from east Tennessee.
Anyway for most Applications under 10mA, also current sources, I think these could be useful devices and they are cheap and small. (SOT-23) As a VAS, I'm not totally convinced, but it does sound pretty good. I have yet to do "experiments" though to see what it will do in this application. Any thoughts?
http://www.fairchildsemi.com/ds/FJ/FJV992.pdf |
|
|
| AndrewT |
Hi,
that FJV992 has adequate Vce0, mediocre capacitance, poor current, poor power and poor fT specs. It appears that fairchild have sacrificed most to achieve very high gain. There must be a better way? |
|
|
| kimschips |
Hey
Now if we are in for smd:
www.Infinion.com
BFN 18-19
BFN 24-26
BFN 38-39
Rohm 2SC3906 2SA1514
BCP53-56
Best regards |
|
|
| LineSource |
I would appreciate recommendations for a 1 watt complementary pair of VAS transistors for a +/- 15 V power supply. I was hoping that the lower voltage would yield some better specs.
I am currently using BD549C and BD559C in a VAS and Diamond Buffer, but they are only 500mw. |
|
|
| janneman |
... BD135/136 or 137/138?
Jan Didden |
|
|
| kimschips |
Hey Linesource
Here is some off my favorite transistors:
2SA1540E
2SC3955E
2SA1405E
2SC3599E
2SC3601E
2SA1407E
I think you like them if low capitance, very linear hfe and 1.2W disspation without heatsink is desired:)
Anyone got a schematic or idea for a preamplifier with nonfeedback, 6dB gain and symmetrical circuit?
And off course very god soundquality.
Best regards
Kim |
|
|
| kimschips |
Hello
As I said erlier I write when I got new transistors
2SA1540E
2SC3955E
From http://www.heldt-electronic.de/
Just mail them your order and send them money by Iban and switchcode.
2SA1405E
2SC3599E
From Jørgen P in denmark TLF +4544945244
Don´t know about paying them from another country, should be not be any problem.
Heldt send me MJE350 instead off 2SA1407E yuk!
Could not get 2SC3601E
Jørgen P could only get 2SC3601E and not 2SA1407E
2SA1407 can optains in another Hfe ranking
Best regards
Kim |
|
|
|
