PWM controllers... shopping

I am driving a thermoelectric cooler with an H-bridge configuration
(inductor in series with the TEC). > So I need to drive 4 switches, but really only need one PWM signal and
its inverse.

If there's anything EMI sensitive on the cold plate, you'll want to filter better than that--the end plates of the TEC are made of alumina, which has an epsilon around 10. You can couple an astounding amount of hash through that. That's why I generally use a Class H drive with a buck switcher and low-sat BJT drive--almost as efficient and much much quieter. If you really sit on the SMPS supervision, you can keep V_CE down below 200 mV, so you can put a few amps through a SOT23 transistor.

The other thing to watch out for is ripple current heating. Dissipation goes like I_RMS but cooling goes like I_avg.

Cheers

Phil Hobbs

 

[sea moss]

If there's anything EMI sensitive on the cold plate, you'll want to filter better than that--the end plates of the TEC are made of alumina, which has an epsilon around 10. You can couple an astounding amount of hash through that. That's why I generally use a Class H drive with a buck switcher and low-sat BJT drive--almost as efficient and much much quieter. If you really sit on the SMPS supervision, you can keep V_CE down below 200 mV, so you can put a few amps through a SOT23 transistor.

The other thing to watch out for is ripple current heating. Dissipation goes like I_RMS but cooling goes like I_avg.

Cheers

Phil Hobbs

Thanks for sharing this. Yes, there is potentially EMI-sensitive stuff on the cold plate. I am now throwing the original H-bridge idea in the garbage. Cheers.

 

[sea moss]

Sloman A.W., Buggs P., Molloy J., and Stewart D. “A microcontroller-based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor” Measurement Science and Technology, 7 1653-64 (1996)

Describes an example of such a circuit. We got the the low frequency content of the PWM waveform down by playing around a small programmable logic device. The device is long obsolete, but there are plenty of more modern equivalents around.

--
Bill Sloman, Sydney

I found this paper online, but to read it I have to let academia.edu have access to my google contacts.. WTF?! I will email you for a copy. Thanks

 

[Bill Sloman]

On Sunday, April 12, 2020 at 10:11:45 PM UTC+10, pcdh...@gmail.com wrote:

I am driving a thermoelectric cooler with an H-bridge configuration
(inductor in series with the TEC). > So I need to drive 4 switches, but really only need one PWM signal and
its inverse.

If there's anything EMI sensitive on the cold plate, you'll want to filter better than that--the end plates of the TEC are made of alumina, which has an epsilon around 10. You can couple an astounding amount of hash through that. That's why I generally use a Class H drive with a buck switcher and low-sat BJT drive--almost as efficient and much much quieter. If you really sit on the SMPS supervision, you can keep V_CE down below 200 mV, so you can put a few amps through a SOT23 transistor.

If you use an inductor-capacitor filter, you can do a lot better, and if you mess around with PWM waveform you can do even better.

Sloman A.W., Buggs P., Molloy J., and Stewart D. “A microcontroller-based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor” Measurement Science and Technology, 7 1653-64 (1996)

> The other thing to watch out for is ripple current heating. Dissipation goes like I_RMS but cooling goes like I_avg.

That's not entirely correct. Ohmic heating goes up as I^2 (as it always does) while Peltier heating and cooling is directly proportional to the current. I went into the issue in more detail in the paper above.

Email me if you need a reprint at bill.sloman@ieee.org

--
Bill Sloman, Sydney

 

[Dimiter_Popoff]

On 4/11/2020 2:19, sea moss wrote:

Seems like every time I start a new power design, I waste a bunch of time looking for the ideal PWM controller.

At the end of the day I'd really like to have the simplest IC possible to handle just the PWM function, and I provide my own error amp, gate driver, and reference externally. The ideal part would look like this: oscillator set by resistor, comparator non-inverting input, and pulse out. Plus Vdd and ground equals 5 pins only! Does anything like this exist?

I've also been checking out the famous TL494, and MC34063 since I have never used them. Does anyone here have horror stories from these parts that I can't infer from the datasheets?

Another one I discovered is MCP1632, pretty simple but not quite what I'm looking for. Vdd=6V max for example, kind of sucks. I do like how they use a 50uA current source as the reference pin, so you set Vref with a resistor or drive from low-impedance source, nice.
http://ww1.microchip.com/downloads/en/DeviceDoc/20005254A.pdf

I am really puzzled why I am the only one on the planet doing it but
the perfect chip for that exists and I have been using it in countless
designs for decades now.It is the 74HC123, earlier the 74LS123 and even
earlier the 74123.
You have to provide an oscillator but you get two PWMs using a 123 part
(using half of it as oscillator is doable but very unreliable, sometimes
it just won't start). I typically use a 555 to get something like
500 kHz and it clocks all the 123-s.
A *significant* benefit over using the 123 instead of an MCU with
timers/pwm is that there is no clock granularity.

Dimiter

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com
======================================================
http://www.flickr.com/photos/didi_tgi/

 

[Bill Sloman]

On Monday, April 13, 2020 at 12:45:18 AM UTC+10, sea moss wrote:

On Sunday, April 12, 2020 at 7:39:25 AM UTC-7, sea moss wrote:
Sloman A.W., Buggs P., Molloy J., and Stewart D. “A microcontroller-based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor” Measurement Science and Technology, 7 1653-64 (1996)

Describes an example of such a circuit. We got the the low frequency content of the PWM waveform down by playing around a small programmable logic device. The device is long obsolete, but there are plenty of more modern equivalents around.

--
Bill Sloman, Sydney

I found this paper online, but to read it I have to let academia.edu have access to my google contacts.. WTF?! I will email you for a copy. Thanks

I tried bill.sloman@ieee.org, but it bounced back, what's the full address?

Add the extra full stop - you need two, not one.

--
Bill Sloman, Sydney (where it is bed-time)

 

[sea moss]

I once did a digital SMPS with what Lasse suggest here.

The error signal was fed into the ADC and DMA transferred the error voltage directly to the PWM timer (so not computation involved)

Around that was looped a peak current limit and soft start

In this can it only replaces the PWM part

You can do more by adding the loop, but then you get phase problems if you loop is not fast, wrt to the PWM frequency

Cheers

Klaus

Klaus, what was the advantage of using a microcontroller? Software-adjustable dead time?

 

[sea moss]

On Sunday, April 12, 2020 at 7:39:25 AM UTC-7, sea moss wrote:

Sloman A.W., Buggs P., Molloy J., and Stewart D. “A microcontroller-based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor” Measurement Science and Technology, 7 1653-64 (1996)

Describes an example of such a circuit. We got the the low frequency content of the PWM waveform down by playing around a small programmable logic device. The device is long obsolete, but there are plenty of more modern equivalents around.

--
Bill Sloman, Sydney

I found this paper online, but to read it I have to let academia.edu have access to my google contacts.. WTF?! I will email you for a copy. Thanks

I tried billsloman@ieee.org, but it bounced back, what's the full address?

 

[Phil Hobbs]

On 2020-04-12 11:09, Dimiter_Popoff wrote:

On 4/11/2020 2:19, sea moss wrote:
Seems like every time I start a new power design, I waste a bunch of
time looking for the ideal PWM controller.

At the end of the day I'd really like to have the simplest IC possible
to handle just the PWM function, and I provide my own error amp, gate
driver, and reference externally.  The ideal part would look like
this: oscillator set by resistor, comparator non-inverting input, and
pulse out.  Plus Vdd and ground equals 5 pins only!  Does anything
like this exist?

I've also been checking out the famous TL494, and MC34063 since I have
never used them.  Does anyone here have horror stories from these
parts that I can't infer from the datasheets?

Another one I discovered is MCP1632, pretty simple but not quite what
I'm looking for.  Vdd=6V max for example, kind of sucks.  I do like
how they use a 50uA current source as the reference pin, so you set
Vref with a resistor or drive from low-impedance source, nice.
http://ww1.microchip.com/downloads/en/DeviceDoc/20005254A.pdf




I am really puzzled why I am the only one on the planet doing it but
the perfect chip for that exists and I have been using it in countless
designs for decades now.It is the 74HC123, earlier the 74LS123 and even
earlier the 74123.
You have to provide an oscillator but you get two PWMs using a 123 part
(using half of it as oscillator is doable but very unreliable, sometimes
it just won't start). I typically use a 555 to get something like
500 kHz and it clocks all the 123-s.
A *significant* benefit over using the 123 instead of an MCU with
timers/pwm is that there is no clock granularity.

Plus all that phase noise smooths out the spectrum.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Dimiter_Popoff]

On 4/12/2020 19:15, Phil Hobbs wrote:

On 2020-04-12 11:09, Dimiter_Popoff wrote:
On 4/11/2020 2:19, sea moss wrote:
Seems like every time I start a new power design, I waste a bunch of
time looking for the ideal PWM controller.

At the end of the day I'd really like to have the simplest IC
possible to handle just the PWM function, and I provide my own error
amp, gate driver, and reference externally.  The ideal part would
look like this: oscillator set by resistor, comparator non-inverting
input, and pulse out.  Plus Vdd and ground equals 5 pins only!  Does
anything like this exist?

I've also been checking out the famous TL494, and MC34063 since I
have never used them.  Does anyone here have horror stories from
these parts that I can't infer from the datasheets?

Another one I discovered is MCP1632, pretty simple but not quite what
I'm looking for.  Vdd=6V max for example, kind of sucks.  I do like
how they use a 50uA current source as the reference pin, so you set
Vref with a resistor or drive from low-impedance source, nice.
http://ww1.microchip.com/downloads/en/DeviceDoc/20005254A.pdf




I am really puzzled why I am the only one on the planet doing it but
the perfect chip for that exists and I have been using it in countless
designs for decades now.It is the 74HC123, earlier the 74LS123 and even
earlier the 74123.
You have to provide an oscillator but you get two PWMs using a 123 part
(using half of it as oscillator is doable but very unreliable, sometimes
it just won't start). I typically use a 555 to get something like
500 kHz and it clocks all the 123-s.
A *significant* benefit over using the 123 instead of an MCU with
timers/pwm is that there is no clock granularity.

Plus all that phase noise smooths out the spectrum.

Cheers

Phil Hobbs

Not much phase noise if done right (it is sensitive to power supply
noise, especially via the feedback network), our (up to) 5kV HV bias
for HPGe detectors has really minimal noise, a few mV (this is just
noise, no ripple to speak of, the current load is what goes into the
1G feedback resistor).
But it can get noisy, many years ago I learned the "quiet power"
lesson the hard way .

Dimiter

 

[Phil Hobbs]

On 2020-04-12 11:09, Dimiter_Popoff wrote:

On 4/11/2020 2:19, sea moss wrote:
Seems like every time I start a new power design, I waste a bunch of
time looking for the ideal PWM controller.

At the end of the day I'd really like to have the simplest IC possible
to handle just the PWM function, and I provide my own error amp, gate
driver, and reference externally.  The ideal part would look like
this: oscillator set by resistor, comparator non-inverting input, and
pulse out.  Plus Vdd and ground equals 5 pins only!  Does anything
like this exist?

I've also been checking out the famous TL494, and MC34063 since I have
never used them.  Does anyone here have horror stories from these
parts that I can't infer from the datasheets?

Another one I discovered is MCP1632, pretty simple but not quite what
I'm looking for.  Vdd=6V max for example, kind of sucks.  I do like
how they use a 50uA current source as the reference pin, so you set
Vref with a resistor or drive from low-impedance source, nice.
http://ww1.microchip.com/downloads/en/DeviceDoc/20005254A.pdf




I am really puzzled why I am the only one on the planet doing it but
the perfect chip for that exists and I have been using it in countless
designs for decades now.It is the 74HC123, earlier the 74LS123 and even
earlier the 74123.
You have to provide an oscillator but you get two PWMs using a 123 part
(using half of it as oscillator is doable but very unreliable, sometimes
it just won't start). I typically use a 555 to get something like
500 kHz and it clocks all the 123-s.
A *significant* benefit over using the 123 instead of an MCU with
timers/pwm is that there is no clock granularity.

So you replace Rext with a variable current source or something like that?

I'd be okay using something like that closed-loop, but I don't see any
specified min/max pulse width specs, so I wouldn't be super confident
using it open loop. AFAICT it could be as bad as the HC4046, i.e. 3:1
or worse.

Cheers

Phil Hobbs


--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Dimiter_Popoff]

On 4/12/2020 20:03, Phil Hobbs wrote:

On 2020-04-12 11:09, Dimiter_Popoff wrote:
On 4/11/2020 2:19, sea moss wrote:
Seems like every time I start a new power design, I waste a bunch of
time looking for the ideal PWM controller.

At the end of the day I'd really like to have the simplest IC
possible to handle just the PWM function, and I provide my own error
amp, gate driver, and reference externally.  The ideal part would
look like this: oscillator set by resistor, comparator non-inverting
input, and pulse out.  Plus Vdd and ground equals 5 pins only!  Does
anything like this exist?

I've also been checking out the famous TL494, and MC34063 since I
have never used them.  Does anyone here have horror stories from
these parts that I can't infer from the datasheets?

Another one I discovered is MCP1632, pretty simple but not quite what
I'm looking for.  Vdd=6V max for example, kind of sucks.  I do like
how they use a 50uA current source as the reference pin, so you set
Vref with a resistor or drive from low-impedance source, nice.
http://ww1.microchip.com/downloads/en/DeviceDoc/20005254A.pdf




I am really puzzled why I am the only one on the planet doing it but
the perfect chip for that exists and I have been using it in countless
designs for decades now.It is the 74HC123, earlier the 74LS123 and even
earlier the 74123.
You have to provide an oscillator but you get two PWMs using a 123 part
(using half of it as oscillator is doable but very unreliable, sometimes
it just won't start). I typically use a 555 to get something like
500 kHz and it clocks all the 123-s.
A *significant* benefit over using the 123 instead of an MCU with
timers/pwm is that there is no clock granularity.


So you replace Rext with a variable current source or something like that?

Sort of. If I can power the error opamp off something like 10-12 volts
its output drives directly Rext (instead of having Rext to +5V). I
usually use 1/2 LMC6482 for that purpose, last 25+ years at least.
But if all I have is the (quiet) 5V supply the output of the 6482
drives the gate of a say 2N176 (SST176 really), source to +5V, drain
into some Rext (last time it was a 2k).

I'd be okay using something like that closed-loop, but I don't see any
specified min/max pulse width specs, so I wouldn't be super confident
using it open loop.  AFAICT it could be as bad as the HC4046, i.e. 3:1
or worse.

Oh I use it closed loop only of course. One can get 0 to say 80+% duty
(you can keep the 123 constantly on by pulling the RC pin to GND, handy
for protections, gating etc., obviously the power switch is off while
the 123 is on).
That said 123 parts are remarkably similar, even between vendors, have
been over the years.
The RC is specified for time and probably they all use the same
currents/thresholds or sort of. Or it is just the same design they
all use, no idea.

Cheers

Phil Hobbs

Dimiter

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com
======================================================
http://www.flickr.com/photos/didi_tgi/

 

[Phil Hobbs]

On 2020-04-12 14:09, Dimiter_Popoff wrote:

On 4/12/2020 20:03, Phil Hobbs wrote:
On 2020-04-12 11:09, Dimiter_Popoff wrote:
On 4/11/2020 2:19, sea moss wrote:
Seems like every time I start a new power design, I waste a bunch of
time looking for the ideal PWM controller.

At the end of the day I'd really like to have the simplest IC
possible to handle just the PWM function, and I provide my own error
amp, gate driver, and reference externally.  The ideal part would
look like this: oscillator set by resistor, comparator non-inverting
input, and pulse out.  Plus Vdd and ground equals 5 pins only!  Does
anything like this exist?

I've also been checking out the famous TL494, and MC34063 since I
have never used them.  Does anyone here have horror stories from
these parts that I can't infer from the datasheets?

Another one I discovered is MCP1632, pretty simple but not quite
what I'm looking for.  Vdd=6V max for example, kind of sucks.  I do
like how they use a 50uA current source as the reference pin, so you
set Vref with a resistor or drive from low-impedance source, nice.
http://ww1.microchip.com/downloads/en/DeviceDoc/20005254A.pdf




I am really puzzled why I am the only one on the planet doing it but
the perfect chip for that exists and I have been using it in countless
designs for decades now.It is the 74HC123, earlier the 74LS123 and even
earlier the 74123.
You have to provide an oscillator but you get two PWMs using a 123 part
(using half of it as oscillator is doable but very unreliable, sometimes
it just won't start). I typically use a 555 to get something like
500 kHz and it clocks all the 123-s.
A *significant* benefit over using the 123 instead of an MCU with
timers/pwm is that there is no clock granularity.


So you replace Rext with a variable current source or something like
that?

Sort of. If I can power the error opamp off something like 10-12 volts
its output drives directly Rext (instead of having Rext to +5V). I
usually use 1/2 LMC6482 for that purpose, last 25+ years at least.
But if all I have is the (quiet) 5V supply the output of the 6482
drives the gate of a say 2N176 (SST176 really), source to +5V, drain
into some Rext (last time it was a 2k).

I'd be okay using something like that closed-loop, but I don't see any
specified min/max pulse width specs, so I wouldn't be super confident
using it open loop.  AFAICT it could be as bad as the HC4046, i.e. 3:1
or worse.

Oh I use it closed loop only of course. One can get 0 to say 80+% duty
(you can keep the 123 constantly on by pulling the RC pin to GND, handy
for protections, gating etc., obviously the power switch is off while
the 123 is on).
That said 123 parts are remarkably similar, even between vendors, have
been over the years.
The RC is specified for time and probably they all use the same
currents/thresholds or sort of. Or it is just the same design they
all use, no idea.

Good to know, thanks!

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com

 

[Ricky C]

On Sunday, April 12, 2020 at 10:39:25 AM UTC-4, sea moss wrote:

Sloman A.W., Buggs P., Molloy J., and Stewart D. “A microcontroller-based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor” Measurement Science and Technology, 7 1653-64 (1996)

Describes an example of such a circuit. We got the the low frequency content of the PWM waveform down by playing around a small programmable logic device. The device is long obsolete, but there are plenty of more modern equivalents around.

--
Bill Sloman, Sydney

I found this paper online, but to read it I have to let academia.edu have access to my google contacts.. WTF?! I will email you for a copy. Thanks

What you don't have a crap google account for stuff like this?

--

Rick C.

--- Get 1,000 miles of free Supercharging
--- Tesla referral code - https://ts.la/richard11209

 

[Hul Tytus]

The attiny25 allows use of the comparator and a pwm timer/counter at the same time. A few pennies more, though.

Hul

bitrex <user@example.net> wrote:

On 4/10/2020 7:59 PM, Ricky C wrote:
On Friday, April 10, 2020 at 7:19:36 PM UTC-4, sea moss wrote:
Seems like every time I start a new power design, I waste a bunch of time looking for the ideal PWM controller.

At the end of the day I'd really like to have the simplest IC possible to handle just the PWM function, and I provide my own error amp, gate driver, and reference externally. The ideal part would look like this: oscillator set by resistor, comparator non-inverting input, and pulse out. Plus Vdd and ground equals 5 pins only! Does anything like this exist?

I've also been checking out the famous TL494, and MC34063 since I have never used them. Does anyone here have horror stories from these parts that I can't infer from the datasheets?

Another one I discovered is MCP1632, pretty simple but not quite what I'm looking for. Vdd=6V max for example, kind of sucks. I do like how they use a 50uA current source as the reference pin, so you set Vref with a resistor or drive from low-impedance source, nice.
http://ww1.microchip.com/downloads/en/DeviceDoc/20005254A.pdf

I can't get to a 5 pin device, but I bet I can do that in a 48 pin QFN using an FPGA. ICE5LP1K-SG48ITR50, under $3.00 qty 100.

There's not much in the control world that can't be done primarily digitally these days.


ATTiny13A, 1k program memory, two PWM channels, 10 bit ADC, 20MHz clock.
can do all sorts of stuff with that for 40 cent in 100s

 

[sea moss]

What you don't have a crap google account for stuff like this?

--

Rick C.

I do for youtube, but on SED I drive without a seatbelt.

 

[sea moss]

What does SED have to do with academia.edu?

--

Rick C.

Never mind, I misunderstood what you were saying. Yeah I can use a throwaway for that.

 

[Ricky C]

On Sunday, April 12, 2020 at 4:49:53 PM UTC-4, sea moss wrote:

What you don't have a crap google account for stuff like this?

--

Rick C.

I do for youtube, but on SED I drive without a seatbelt.

What does SED have to do with academia.edu?

--

Rick C.

--+ Get 1,000 miles of free Supercharging
--+ Tesla referral code - https://ts.la/richard11209

 

[Klaus Kragelund]

On Sunday, April 12, 2020 at 4:50:59 PM UTC+2, sea moss wrote:

I once did a digital SMPS with what Lasse suggest here.

The error signal was fed into the ADC and DMA transferred the error voltage directly to the PWM timer (so not computation involved)

Around that was looped a peak current limit and soft start

In this can it only replaces the PWM part

You can do more by adding the loop, but then you get phase problems if you loop is not fast, wrt to the PWM frequency

Cheers

Klaus

Klaus, what was the advantage of using a microcontroller? Software-adjustable dead time?

We had a microcontroller there already, so it was just to save cost. Digital control is normally slower than analog, but can remove component tolerances and cut cost

In that design it controlled a half bridge converter, a boost converter, a maximum power point tracker and a backlight LED. Oh, and then also the main purpose of the product, a Modbus controller

Cheers

Klaus

Cheers

Klaus