Is there anybody interested in this kind of contribution to the constant ongoing discussion, verticals vs laterals?
Once upon a time, maybe Anno Domini 1977, Hitachi announced its S series stating the lateral structure was used in order to achieve enhanced high frequency and gain performance and low gate-drain capacitance.
The predominant vertical MOSFET topology has due to huge industrial demand been subject of massive continuous research and development. However, the lateral topology encompasses features which can not be replicated by any means.
Linear transfer means, most importantly, low intermodulation distortion . High bandwidth, in subjective terms, signifies a soft, musical and less distorted sound, characterizing a large spectrum of the audible frequency range.
Parameters of MOSFETs regarding transfer and high frequency performance:
Idss = saturated drain (cut-off) current. High Idss values have the penalty of increased Crss and Ciss.
Rdson = static drain-source on resistance. Low Rdson brings about high Coss. Rdson increases with increasing temperature.
W = channel width.
L = channel length. A crucial structural difference. L is very short and more precisely defined for laterals offering several advantages like low Crss.
In the case of verticals, the lack of close control over the device characteristics in the vertical dimension results in longer channel lengths, increased leakage current, and increased parasitic capacitances. A large W/L ratio is necessary to achieve high Idss and low Rdson values being important for switching but that is not possible without increasing certain capacitances.
VGSth = the minimum gate bias voltage which enables the formation of the channel between drain and source.
Gm = forward transconductance. The ratio of a change in AC output current to a change in AC input voltage at other parameters fixed. Gm is proportional to W/L. Gm decreases with increasing temperature.
ft = maximum frequency of operation. ft is directly proportional to Gm and inversely proportional to Ciss and to L.
Capacitances supplied by manufactures in data sheets:
Ciss = input capacitance, the sum of Cgs and Cgd. Cgd is small compared to Cds.
Coss = output capacitance, the sum of Cgd and Cds. It is mainly a junction capacitance having its highest value at Vgs = 0, decreasing with increasing Vds, its minimum value specified at a given Vds. Coss is directly proportional to W/L.
Crss = reverse transfer or feedback Miller capacitance, same as Cdg, suffers from voltage gain multiplication. Despite being the lowest value compared to Ciss and Coss, it has the most devastating effect on high frequency performance, thus on the sound. Low Crss values reduce Idss.
For good high frequency performance a low ratio of Crss/Coss and even more importantly a high ratio of Gm/Crss is essential.
High Gm is required also for high power gain and linearity, obtained at low gate voltages and having flat characteristics over a wide range of gate voltage.
In conclusion: large power handling capability has an inevitable price. Because of its construction vertical MOSFETs can not challenge the linear transfer characteristics and high frequency performance of laterals.
Once upon a time, maybe Anno Domini 1977, Hitachi announced its S series stating the lateral structure was used in order to achieve enhanced high frequency and gain performance and low gate-drain capacitance.
The predominant vertical MOSFET topology has due to huge industrial demand been subject of massive continuous research and development. However, the lateral topology encompasses features which can not be replicated by any means.
Linear transfer means, most importantly, low intermodulation distortion . High bandwidth, in subjective terms, signifies a soft, musical and less distorted sound, characterizing a large spectrum of the audible frequency range.
Parameters of MOSFETs regarding transfer and high frequency performance:
Idss = saturated drain (cut-off) current. High Idss values have the penalty of increased Crss and Ciss.
Rdson = static drain-source on resistance. Low Rdson brings about high Coss. Rdson increases with increasing temperature.
W = channel width.
L = channel length. A crucial structural difference. L is very short and more precisely defined for laterals offering several advantages like low Crss.
In the case of verticals, the lack of close control over the device characteristics in the vertical dimension results in longer channel lengths, increased leakage current, and increased parasitic capacitances. A large W/L ratio is necessary to achieve high Idss and low Rdson values being important for switching but that is not possible without increasing certain capacitances.
VGSth = the minimum gate bias voltage which enables the formation of the channel between drain and source.
Gm = forward transconductance. The ratio of a change in AC output current to a change in AC input voltage at other parameters fixed. Gm is proportional to W/L. Gm decreases with increasing temperature.
ft = maximum frequency of operation. ft is directly proportional to Gm and inversely proportional to Ciss and to L.
Capacitances supplied by manufactures in data sheets:
Ciss = input capacitance, the sum of Cgs and Cgd. Cgd is small compared to Cds.
Coss = output capacitance, the sum of Cgd and Cds. It is mainly a junction capacitance having its highest value at Vgs = 0, decreasing with increasing Vds, its minimum value specified at a given Vds. Coss is directly proportional to W/L.
Crss = reverse transfer or feedback Miller capacitance, same as Cdg, suffers from voltage gain multiplication. Despite being the lowest value compared to Ciss and Coss, it has the most devastating effect on high frequency performance, thus on the sound. Low Crss values reduce Idss.
For good high frequency performance a low ratio of Crss/Coss and even more importantly a high ratio of Gm/Crss is essential.
High Gm is required also for high power gain and linearity, obtained at low gate voltages and having flat characteristics over a wide range of gate voltage.
In conclusion: large power handling capability has an inevitable price. Because of its construction vertical MOSFETs can not challenge the linear transfer characteristics and high frequency performance of laterals.
Both types have good and bad characteristics. On balance I favour the lateral, the negative tempco is a major plus point in my book, and they also seem much better matched when used as complementary pairs. The downside is the relatively low current capability per device and the higher on resistance and cost of course.
jacco vermeulen said:
I would guess not, but high Rdson does have some disadvantages in audio, two to be precise.
1) It takes up power dissipation, resulting in lower efficiency for the design. Typical lateral has Rdson of about 1 ohm, which at it's full current of 7A drops 7V. A net 49W dissipation you can't do away with, and 7V of headroom lost from power rail voltage that you can't avoid.
2) Using the MOSFET at high currents and high amp output voltages gets into the nonlinear capacitance problem, because Vdg not only goes into the low values where Cds is nonlinear, but it can also become negative, as in gate voltage higher than drain voltage. This sort of thing generates an inflection type characteristic in the Cgd graph, resulting in aded 4th and higher order harmonics.
As a consequence, in order to use your power supply as efficiently as possible, you need to provide a higher voltage for the driver stages (laterals typically need about 9-14V to fully saturate, but a large Rdson still remains even then), but then you have the added penalty of nonlinear Cdg. Granted, it's less nonlinear than with verticals, but still. Of course, i am assuming output follower topology. In an aoutput stage with gain, rail voltage utilisation is better but the nonlinear Cdg problem remains, and is actually rendered more signifficant because Cdg is prone to the miller effect in these configurations.
This means that they are the more efficient the higher the rail voltages are, in those configurations there will also be many in parallel, mitigating the high Rdson problem. In other words, simple single pair output amps with laterals are not very efficient, and may have other problems one would not expect.
peranders said:
That is true... although, if you take as an example, say, an amp able to drive a 4 ohm fully resistive load, your power supply needs about 35V instead of 29V (typical for a vertical MOS) for comparative current of a single lateral pair, 7A. The output power is about 96W.
The problem is the power dissipated in the transistors, and that you do not design for on a logarithmic scale (unfortunately), additional watts will be very well noticed. What I am trying to point out is that some seemingly unexpected problems may arise with the usage of laterals in simple amps (say a single output pair), which the designer should be aware of.
Mooly said:
I agree. It is really a matter of picking your poison vs picking your advantages. Either laterals or verticals, in the right designer's hands, can be made to perform exceptionally well. It is certainly true that the negative tempco for laterals is nice for bias design and stability, and also contributes to ruggedness under abuse.
The drain-gate capacitance of verticals is not a big problem if the right kind of drivers are used. Verticals are fundamentally much faster than laterals in terms of equivalant ft.
The transconductance of both verticals and laterals decreases at lower currents, leading to distortion. That is the so-called transconductance droop. That is why I applied error correction to the application of verticals in audio power amplifiers and came up with the lowest distortion reported at the time (0.0006% THD at 20 kHz full power in 1983). The application of error correction would have been just as appropriate to laterals. The transconductance of verticals just keeps on going higher as the current goes higher, at least further before pooping out. The transconductance of the laterals as predicted by the square law poops out sooner on the laterals. Many good amplifiers have been made with laterals, but they do have their own set of limitations and I think it would be wrong to generalize that they make better amplifiers. It really depends on the designer.
I do think that it is fair to say that a designer who does not know what he is doing can get into a lot more trouble with verticals than laterals. For one thing, their inherent high speed (equivalanet ft of hundreds of MHz) makes them more prone to oscillate in inexperienced hands. If asked to deliver 100 amps into a short circuit, the verticals will happily try, and may fall on their sword doing so if they are not properly protected.
Cheers,
Bob
Neither, i favor straight balanced operation and all Class A from source to speaker out.(as in: including the inherent balanced operation of MC cartridges and fully balanced DACs)
My comment refers to the Rds(on) part, in the late '80s i built sort of a Hitachi TO3 MOSFET copycat of a Burmester 850 in full Class A.
(the 850 is a balanced power amp, 4 pairs of Toshiba 2SA1302/2SC3281 per channel)
I got the idea from a Japanese power amp, this one.
Turned out that the LB-4 used vertical MOSFETs for the balanced output stage, something i discovered years later when i saw the insides.
Very expensive amp for a mere 25W/ch btw, and the interior looks like any average Japanese consumer spaghetti amplifier from the 1980s.
My comment refers to the Rds(on) part, in the late '80s i built sort of a Hitachi TO3 MOSFET copycat of a Burmester 850 in full Class A.
(the 850 is a balanced power amp, 4 pairs of Toshiba 2SA1302/2SC3281 per channel)
I got the idea from a Japanese power amp, this one.
Turned out that the LB-4 used vertical MOSFETs for the balanced output stage, something i discovered years later when i saw the insides.
Very expensive amp for a mere 25W/ch btw, and the interior looks like any average Japanese consumer spaghetti amplifier from the 1980s.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.