load bank puzzle...

[John Larkin]

On Mon, 20 Mar 2023 05:49:46 -0000 (UTC), Jasen Betts
<usenet@revmaps.no-ip.org> wrote:

On 2023-03-19, John Larkin <jlarkin@highlandSNIPMEtechnology.com> wrote:
On Sun, 19 Mar 2023 17:28:35 +0000, piglet <erichpwagner@hotmail.com
wrote:

On 19/03/2023 15:24, John Larkin wrote:

There are some very cool 2 and 3-channel isolated ADCs intended for
use in electric meters. Cheap but not super fast; good for average
measurements but not for closing fast control loops.


So use the cool ADCs for metrology and do the control loop with op-amp?

piglet



Simulating a resistor needs multiplication. That could be done with an
MDAC but gets complicated.

Could be done with a digital pot, ah yeah, that\'s a type of MDAC I
guess.

Dpots are usually terrible MDACs, intended for audio use mostly. There
are good 12 and 16-bit fast 4Q MDACs around, but a multiply in an FPGA
is better and cheaper.

 

[legg]

On Sat, 18 Mar 2023 08:11:29 -0700, John Larkin
<jlarkin@highlandSNIPMEtechnology.com> wrote:

On Sat, 18 Mar 2023 07:13:33 -0400, legg <legg@nospam.magma.ca> wrote:

On Fri, 17 Mar 2023 14:41:51 -0700, John Larkin
jlarkin@highlandSNIPMEtechnology.com> wrote:

On Fri, 17 Mar 2023 17:14:59 -0400, legg <legg@nospam.magma.ca> wrote:

On Fri, 17 Mar 2023 08:21:23 -0700, John Larkin
jlarkin@highlandSNIPMEtechnology.com> wrote:


I\'ve been asked to design an 8-channel dummy load board. It\'s not very
challenging but somebody\'s got to do it. It will be used to simulate
small loads like solenoids or relays or torque motors. It needs some
inductance too, because the drivers often PWM. Maybe 10 or so watts
per channel.

I could do this electronically, but it would a lot easier and more
rugged if I use wirewound resistors. I was thinking of making a
conductance DAC, namely resistors R 2R 4R etc switched in parallel
across the inputs with an SSR per resistor.

But there is a history of clever load banks. When I was an EE student
at Tulane, two semisters of Electrical Machinery (with lab) was
mandatory. It was a pain but I learned a lot. We had a big load bank
in the machinery lab, a string of giant series resistors with a
3-position knife switch at each node. That made me think about using
series-parallel combinations to hit some target value.

https://www.dropbox.com/s/x3e7xxi13n6wd1o/Load_Banks_1.jpg?raw=1

The upper pic is the ancient Tulane load bank as I remember it.

I was thinking about the lower circuit for my gadget. I could use
wirewould resistors and kink the leads to space them maybe 3/4 inch
above my PCB, in the air stream. The higher value resistors might be
2512 surface mounts.

I think there is a tool to bend and kink resistor leads. Or we could
send a bunch out to a service maybe.




Active I=kV with a delay would give you an inductance.


A real inductor stores energy, which a synthesized inductor usually
doesn\'t.

Stored energy pumps current into flyback diodes or equivalent.


Does the driver really care?

It certainly might.


Where\'d the word \'dummy\' originate in the spec? If it reflects
the attitude of the buyer, I\'d say \'No Bid\'. I\'ve had it with
idiots like that.

Oh, they are just a biggish aerospace company.

No bid? Is that a good business model?


RL

https://en.wikipedia.org/wiki/Dummy_load


The minute you considered variable, multicase and
especially inductive, it ceased to be idiot-proof.

Use sockets and real loads, if it\'s that important.

RL

The customer has an existing design where they select and solder a
selected set of resistors and inductors per channel, unique to each
unit. We don\'t want to be in the business of doing that for them, with
every unit having its own dash number and BOM and test limits.

I\'d like to design a programmable dummy load board that we can
manufacture and stock and ship when we get an order.

It\'s looking like a straight conductance DAC is the way to go:
parallel N resistors (R, 2R, etc) with a solid-state switch per. Given
a binary control code K, net conductance is proportional to K so
resistance goes as 1/K. N=5 maybe; we\'re not simulating RTDs. That\'s
nice and simple and adds an open-circuit case for free. Two more SSRs
can add short and ground fault cases, selling points.

I like to add little goodies to products when it\'s not hard and
doesn\'t interfere with the base function. You never know if something
will appeal to someone and tip a basically emotional buy decision, as
in \"That thingie might be useful some day, let\'s buy theirs.\"

Colors matter too.

In my experience the biggest issue with the versatility
of even the simplest load is connecting to the DUT.

Every sodding device uses a different connector.

RL

 

[John Larkin]

On Mon, 20 Mar 2023 10:39:12 -0400, legg <legg@nospam.magma.ca> wrote:

On Sat, 18 Mar 2023 08:11:29 -0700, John Larkin
jlarkin@highlandSNIPMEtechnology.com> wrote:

On Sat, 18 Mar 2023 07:13:33 -0400, legg <legg@nospam.magma.ca> wrote:

On Fri, 17 Mar 2023 14:41:51 -0700, John Larkin
jlarkin@highlandSNIPMEtechnology.com> wrote:

On Fri, 17 Mar 2023 17:14:59 -0400, legg <legg@nospam.magma.ca> wrote:

On Fri, 17 Mar 2023 08:21:23 -0700, John Larkin
jlarkin@highlandSNIPMEtechnology.com> wrote:


I\'ve been asked to design an 8-channel dummy load board. It\'s not very
challenging but somebody\'s got to do it. It will be used to simulate
small loads like solenoids or relays or torque motors. It needs some
inductance too, because the drivers often PWM. Maybe 10 or so watts
per channel.

I could do this electronically, but it would a lot easier and more
rugged if I use wirewound resistors. I was thinking of making a
conductance DAC, namely resistors R 2R 4R etc switched in parallel
across the inputs with an SSR per resistor.

But there is a history of clever load banks. When I was an EE student
at Tulane, two semisters of Electrical Machinery (with lab) was
mandatory. It was a pain but I learned a lot. We had a big load bank
in the machinery lab, a string of giant series resistors with a
3-position knife switch at each node. That made me think about using
series-parallel combinations to hit some target value.

https://www.dropbox.com/s/x3e7xxi13n6wd1o/Load_Banks_1.jpg?raw=1

The upper pic is the ancient Tulane load bank as I remember it.

I was thinking about the lower circuit for my gadget. I could use
wirewould resistors and kink the leads to space them maybe 3/4 inch
above my PCB, in the air stream. The higher value resistors might be
2512 surface mounts.

I think there is a tool to bend and kink resistor leads. Or we could
send a bunch out to a service maybe.




Active I=kV with a delay would give you an inductance.


A real inductor stores energy, which a synthesized inductor usually
doesn\'t.

Stored energy pumps current into flyback diodes or equivalent.


Does the driver really care?

It certainly might.


Where\'d the word \'dummy\' originate in the spec? If it reflects
the attitude of the buyer, I\'d say \'No Bid\'. I\'ve had it with
idiots like that.

Oh, they are just a biggish aerospace company.

No bid? Is that a good business model?


RL

https://en.wikipedia.org/wiki/Dummy_load


The minute you considered variable, multicase and
especially inductive, it ceased to be idiot-proof.

Use sockets and real loads, if it\'s that important.

RL

The customer has an existing design where they select and solder a
selected set of resistors and inductors per channel, unique to each
unit. We don\'t want to be in the business of doing that for them, with
every unit having its own dash number and BOM and test limits.

I\'d like to design a programmable dummy load board that we can
manufacture and stock and ship when we get an order.

It\'s looking like a straight conductance DAC is the way to go:
parallel N resistors (R, 2R, etc) with a solid-state switch per. Given
a binary control code K, net conductance is proportional to K so
resistance goes as 1/K. N=5 maybe; we\'re not simulating RTDs. That\'s
nice and simple and adds an open-circuit case for free. Two more SSRs
can add short and ground fault cases, selling points.

I like to add little goodies to products when it\'s not hard and
doesn\'t interfere with the base function. You never know if something
will appeal to someone and tip a basically emotional buy decision, as
in \"That thingie might be useful some day, let\'s buy theirs.\"

Colors matter too.


In my experience the biggest issue with the versatility
of even the simplest load is connecting to the DUT.

Every sodding device uses a different connector.

RL

Yes. If we do a D25, we will also offer a small termination board, D25
to barrier strip.

My big single-channel load board will use a Phoenix connector where
the male has wire screws.

https://www.digikey.com/en/products/detail/phoenix-contact/1861263/9349684

 

[Phil Hobbs]

On 2023-03-20 10:31, John Larkin wrote:

On Mon, 20 Mar 2023 05:49:46 -0000 (UTC), Jasen Betts
usenet@revmaps.no-ip.org> wrote:

On 2023-03-19, John Larkin <jlarkin@highlandSNIPMEtechnology.com> wrote:
On Sun, 19 Mar 2023 17:28:35 +0000, piglet <erichpwagner@hotmail.com
wrote:

On 19/03/2023 15:24, John Larkin wrote:

There are some very cool 2 and 3-channel isolated ADCs intended for
use in electric meters. Cheap but not super fast; good for average
measurements but not for closing fast control loops.


So use the cool ADCs for metrology and do the control loop with op-amp?

piglet



Simulating a resistor needs multiplication. That could be done with an
MDAC but gets complicated.

Could be done with a digital pot, ah yeah, that\'s a type of MDAC I
guess.

Dpots are usually terrible MDACs, intended for audio use mostly. There
are good 12 and 16-bit fast 4Q MDACs around, but a multiply in an FPGA
is better and cheaper.

Yup. Dpots are mostly good for fairly coarse tweaks. There are some
reasonably quick ones (~6 MHz), but only in the lowest resistance range
(1k).

They also make really bad rheostats, on account of the 20% - 30%
resistance tolerance. (Monolithic matching makes them quite a lot
better as voltage dividers, as long as you don\'t pull too much current
out of the wiper.

The other main problem with both dpots and MDACs is the strong
dependence of bandwidth on the digital code.

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