High transconductance surface mount MOSFETs around logic level

[Sabrosa]

I'm working on a schematic for a variation on OPC's NTD1 (itself an update to the D1) that will allow it to function as an I/V stage for the new ES9038PRO DAC.

The quick gist of the circuit is that the MOSFETs are set up like a common-gate amplifier, with the source voltage set at 3.3V (1/2 AVCC). The MOSFET runs in linear mode and the higher the transconductance, the lower the source impedance, which gets the input closer to the virtual ground the DAC chip likes to see, and thus lowers distortion.

The drain resistor controls the gain and the source resistor the bias. Increasing the bias current ups transconductance as well as dissipation, but lowers gain. So you have to find the equilibrium you can accept between output voltage, power dissipation/heat you can handle, and distortion.

The increased current output of the ES9038PRO, approximately 4x that of the ES9018, necessitates lowering both the gain and voltage of the circuit to reach manageable dissipation and maintain 1VRMS output.

With the power supply rails at +/-6V the circuit can no longer reach the gate voltage threshold of the original FDB52N20 BOM MOSFET, so I've been looking for other high transconductance surface mount MOSFETs with a max threshold of ~4V.

The IRL40S212 HEXFET has come out on top in the simulations I've run. However, it's hard to know whether the models will translate to reality, so I'm putting together a list of MOSFETs to test for transconductance under the operating conditions of the circuit.

Here's what I've come up with so far:

IRL40S212
IRL40SC228
IRL60S216
IRFS7430PbF
IRFS7534PbF
IRF2204S
PSMN1R5-30BLE
PSMN3R4-30BLE
PSMNR90-30BL
CSD19535KTT
CSD19536KTT
CSD18510KTT
CSD18542KCS

I've avoided trench MOSFETs, except for the Nexperia PSMN series which claims an enhanced FBSOA, in favor of HEXFETs and a few of TIs NexFETs. Most of these are logic-level for operating with a low gate threshold voltage.

It's hard to judge transconductance, especially at the operating point in this circuit, from just scouring through data sheets. I'm sure I've missed some good candidates and so if anyone is aware of other high transconductance MOSFETs that I should add to this list for testing, please let me know!

I'm not set on the D2PAK footprint, they just seemed to strike the best balance of transconductance and ease of soldering. I wouldn't have any problem using a SOT-223 MOSFET, for example.

Operating conditions of the circuit above with IRL40S212:

Rail voltage: +/-6V
Voltage at gate: 4.4V
Voltage at source: 1.65V
Voltage at drain: 3V
Operating current (per rail): 1.09A
MOSFET power dissipation: 1.41W

 

[abraxalito]

Looks very interesting - I'll take a peek at your suggestions for suitable FETs and respond later. In the meantime the highest gm MOSFETs I've found (however they're in the main SOT-23 and smaller packages) is Toshiba's range - https://toshiba.semicon-storage.com...d,15d,16d,17d,18d,19d,20d,21d,22d,23h,24d,25d

If you need higher dissipation you can always cascode a small package FET with a TO220 or D2PAK.

 

[jan.didden]

The quick gist of the circuit is that the MOSFETs are set up like a common-gate amplifier, with the source voltage set at 3.3V (1/2 AVCC). The MOSFET runs in linear mode and the higher the transconductance, the lower the source impedance, which gets the input closer to the virtual ground the DAC chip likes to see, and thus lowers distortion.

Have you considered using a servo that keeps the gate at whatever DC required to keep the source at zero DC? That would be an optimum circuit and you could select FETs for the best audio performance.

Jan

 

[MarcelvdG]

For a given bias current, bipolar transistors always give you more transconductance than MOSFETs, so why MOSFETs and not bipolars? Of course bipolar transistors would need a low-impedance base biasing circuit.

 

[abraxalito]

Its an interesting question - purely subjectively I prefer the MOSFET's bass. This might be related to their not drawing any gate current - as a current source the output impedance of the MOSFET is higher at LF, compared to a bipolar and it looks to me due to the lack of base (gate) current.

 

[Sabrosa]

Looks very interesting - I'll take a peek at your suggestions for suitable FETs and respond later. In the meantime the highest gm MOSFETs I've found (however they're in the main SOT-23 and smaller packages) is Toshiba's range - https://toshiba.semicon-storage.com...d,15d,16d,17d,18d,19d,20d,21d,22d,23h,24d,25d

If you need higher dissipation you can always cascode a small package FET with a TO220 or D2PAK.

Thanks for the Toshiba suggestion, I'll definitely check them out. I hadn't looked at their line yet.

I was focused mainly on transconductance when compiling my initial list, as that'd directly correlated with a lower I/V stage input impedance, but I suppose I should also ask if there are additional MOSFET characteristics beneficial to audio that I should be considering in my selection.

Have you considered using a servo that keeps the gate at whatever DC required to keep the source at zero DC? That would be an optimum circuit and you could select FETs for the best audio performance.

Jan

I agree that a DC servo could be ideal in this situation. However, it would add quite a bit of complexity to a simple circuit, and, more so, I don't have experience with designing servos and their filters.

Nothing that couldn't be learned though, I'm sure. I'm up for playing around with some simulations and seeing what happens. Do you have any recommended reading material or suggested circuits to look at?

For a given bias current, bipolar transistors always give you more transconductance than MOSFETs, so why MOSFETs and not bipolars? Of course bipolar transistors would need a low-impedance base biasing circuit.

I've built OPC's ES9018 NTD1 I/V stage and was happy with the sound from the MOSFET, so I've been looking at a similar solution for the ES9038PRO.

 

[abraxalito]

I was focused mainly on transconductance when compiling my initial list, as that'd directly correlated with a lower I/V stage input impedance, but I suppose I should also ask if there are additional MOSFET characteristics beneficial to audio that I should be considering in my selection.

Its a good question and one I don't have an answer to yet. I've been playing with MOSFETs in I/V circuits for a few weeks now and have more questions myself than answers. Last night I incorporated one of Toshiba's PSPICE models into my circuit in LTSpice - all I can say at this stage is the model does not look at all close to the real-world behaviour.

One interesting discovery I've made is that while gm is important its not the sole determinant of the input impedance in a grounded gate stage in cases where the I/V resistor is high (to get best voltage gain). I've been trying to understand this from the theory, slow progress...

<edit> Forgot to mention, I read Owens long NTD1 thread with great interest. The FET he eventually chose I believe had gm of 0.42S at 50mA. According to the DSs from Toshiba there are offerings which comfortably exceed this figure though only in SMT relatively low (1W) dissipation packages.

 

[DPH]

For a given bias current, bipolar transistors always give you more transconductance than MOSFETs, so why MOSFETs and not bipolars? Of course bipolar transistors would need a low-impedance base biasing circuit.

There's some concern, depending on input filtering, about using BJT's (hf hash getting modulated). Definitely some appeal to using jfet/mosfet for current conveying.

 

[Itsmee]

I realise it's been a while since the last post...
Just bought a ES9038pro (ES9038PRO Ulitmate DSD DAC OELD Display Successor of ES9028 DAC / XMOS XU208 USB | eBay), I'm scrapping the I/V.

Definitely some appeal to using jfet/mosfet for current conveying.

Was thinking the same thing, maybe one of these made with ZVN3306A/ZVP3306A, possibly followed by a tweaked jfet buffer to manipulate the H2/H3 response, opinions are appreciated

 

[Markw4]

... the source voltage set at 3.3V (1/2 AVCC).

3.3v = AVCC
1.65v = 1/2 AVCC
Is that what you meant?